performance of mechanical fasteners used with engineered ... · performance of mechanical fasteners...

Project No. UNB75 Value to Wood Project Research Report 2007

Performance of Mechanical Fasteners Used With Engineered Wood Products

by

Ian Smith, Andi Asiz, and Monica Snow

Faculty of Forestry and Environmental Management University of New Brunswick, Fredericton

This report was produced as part of the Value to Wood Program, Funded by Natural Resources Canada

April 2007

UNB75 Performance of Mechanical Fasteners Used with Engineering Wood Products of 116

Summary

The goal was to create information supporting the design of connections in structural systems employing modern engineered wood products (EWP), and to present that information in a format that can be adopted by regulatory agencies in Canada and elsewhere. Activities built on the NRCan funded Value-to-Wood project ‘UNB2 Design Methods for Connections in Engineered Wood Structures’ (2003-06). The approach taken was to recognize that there are two fundamental ways by which engineers design structural connections, and that building regulatory authorities permit. First, it is permitted to design connections based on rules that are detailed in documents like the Canadian Standards Association’s (CSA) Standard 086-01 ‘Engineering Design in Wood’. This is the traditional method and the rules are those that nearly all engineers follow when designing connections in commodity products like sawn lumber, glued-laminated-timber, plywood and Oriented Strand Board. Second, it is permitted to design connections based on information in third-party product assessment reports produced by agencies like the Canadian Construction Materials Centre. Such assessments reports and the design capacities they recommend are widely accepted by building control authorities. This approach is normally followed when proprietary products are involved (joined materials, fasteners, or both joined materials and fasteners). The research project emphasised: methods of establishing design properties that can be incorporated into the national or international timber design codes; and assessment methods that should be adopted by third-party agencies. Within this, attention was paid to developing methods that would yield consistency between the parallel routes by which engineers may design connections in commodity or specialty EWP materials. Essential to activities was close liaison with representatives of the Canadian wood products industry, and with national and international code development bodies. This was to ensure that emphasis was placed on supporting utilization of combinations of EWP and fastener products that are economically important to the production and consumption industries. There was implicit recognition that the project must dovetail with what code development bodies are doing now and may do in the future. Primary interactions were with producers of Laminated Veneer Lumber (LVL) which is the major type of EWP made in Canada, representatives of the Canadian Wood Council, the national CSA Technical Committee 086 ‘Engineering Design in Wood, International Organization for Standardization’s Technical Committee 165 ‘Timber Structures’, and the International Council for Building Research Studies and Documentation’s (CIB) Working Commission 18 ‘Timber Design’. Work centred mainly on validating and refining ideas developed during the UNB2 project, and interacting with the bodies that utilize the results of both the UNB2 and UNB75 projects. Tangible products of the activities include: • Identification of the range of failure mechanisms for EWP connections using commonly

adopted fastener like nails, screws and bolts. • Establishing a comprehensive database on characteristics of connections in Canadian

manufactured LVL. • Discussion documents and recommendations for consideration by code and standards

committees responsible for creation of new (or modification of old) design regulations, and definition of methods for test based assessment of connection systems involving use of proprietary products.


• Presentation of project results to bodies concerned with national and international regulation of design practices for constructions involving use of EWP.

It is important to recognize that the one year project discussed in this report was conceived to install one brick into a wall of technical know-how (and supporting databases) that supports engineers charged with the responsibility to design structural systems that are safe, serviceable and economic. The total process of building and maintaining the wall is far from finite. Ultimately it will be codes and standards committees, regulatory bodies and engineers in various jurisdictions who will mediate whether or not results of this project will be applied in practice. Project team members will continue to be involved in the necessary actions and hope that others share the view that Canadian made EWP are highly effective structural materials and that a framework has been established that has value beyond the national borders.


Acknowledgements

The University of New Brunswick acknowledges the financial support of this project by Natural Resources Canada. Thanks are due to the project champion Kenneth Koo of Jager Building Systems and industry liaisons Eric Jones and Peggy Lepper of the Canadian Wood Council. Much appreciation is extended to colleagues involved with Fastenings Subcommittee of Canadian Standards Association’s Technical Committee 086 ‘Engineering design in wood’, and especially Professor Dr. Pierre Quenneville from the Royal Military College of Canada who chairs the subcommittee. The project team also thanks Jager Building Systems for in-kind contributions of test materials. Many others too numerous to mention here have also provided valuable and much appreciated input.

Research Staff

• Dr. Ian Smith, Project Leader. • Dr. Andi Asiz, Research Associate. • Dr. Monica Snow, Former Graduate Research Assistant. • Dean McCarthy, Chief Technician. • Dale Lemon, Student Research Assistant.


Table of Contents

Summary ......................................................................................................................................... 2 Acknowledgements......................................................................................................................... 4 Research Staff ................................................................................................................................. 4 Table of Contents............................................................................................................................ 5 List of Tables .................................................................................................................................. 6 List of Figures ................................................................................................................................. 6 1. Project objectives ........................................................................................................................ 7 2. Introduction................................................................................................................................. 7 3. Test Program............................................................................................................................... 9

3.1 LVL connection tests ............................................................................................................ 9 3.3 Development of testing method appropriate for EWP connections.................................... 17 3.3 Fatigue tests on connections ............................................................................................... 23

4. Test results and prediction of connection strength ................................................................... 28 4.1 Connections with slender fasteners..................................................................................... 28 4.2 Connections with bolts........................................................................................................ 30

5. Development of code provisions for wood connection design................................................. 31 6. Summary of outputs and recommendations.............................................................................. 32 7. References................................................................................................................................. 33 Appendices.................................................................................................................................... 35 Appendix A - Selected Load-slip Curves of LVL Connections for Each Group Test Appendix B - EWP Bolted Connections Appendix C - Possible Canadian / ISO Approach to Deriving Design Values From Test Data Appendix D - Principles of Connection Design Appendix E - System Based Design of Structures


List of Tables

Table 1: Connections loaded parallel to strong axis of LVL........................................................ 11 Table 2: Connections loaded perpendicular to strong axis of LVL.............................................. 12 Table 3: Connection loaded off-axis at 30º to strong axis of LVL............................................... 13 Table 4: Test results for connections loaded parallel to strong axis of LVL................................ 16 Table 5: Test results for connections loaded perpendicular to strong axis of LVL...................... 16 Table 6: Test results for connection loaded off-axis at 30º to strong axis of LVL....................... 16 Table 7: Comparison of ultimate strengths of alternative perpendicular to the strong axis loading

of connections in EWP.......................................................................................................... 19 Table 8: Schedule of cyclic load (fatigue) tests on bolted specimens .......................................... 25 Table 9: Loading parallel to strong axis ....................................................................................... 29 Table 10: Loading perpendicular to strong axis ........................................................................... 30 Table 11: Loading 30º to strong axis ............................................................................................ 30

List of Figures Figure 1: Connection loaded parallel to strong axis of LVL with four fasteners under single shear

arrangement (44x146 mm member cross-sections) ....................................................... 10 Figure 2: Connection loaded perpendicular to strong axis of LVL with four fasteners under single

shear arrangement (44x146 mm member cross-sections).............................................. 11 Figure 3: LVL connections loaded 30º with respect to strong axis of LVL with two fasteners

under single shear arrangement (44x95 mm member cross-sections) ........................... 12 Figure 4: Connection failure mechanisms .................................................................................... 14 Figure 5: Illustrative load-slip curves for perpendicular to strong axis loading ........................... 15 Figure 6: Test setups for perpendicular to the strong axis loading of connections....................... 17 Figure 7: Connection configuration using new test method ......................................................... 18 Figure 8: Typical load deformation responses for perpendicular to the strong axis loading of

connections .................................................................................................................... 20 Figure 9: Failure mechanisms for perpendicular to the strong axis loading of connections ........ 21 Figure 10: Predicted progression of the failure process for the 12.7 mm bolt loading LSL

perpendicular to the strong axis (bolt loads the right-hand end) ................................... 23 Figure 11: Fatigue test specimens Figure 12: Schematic of cyclic load regime..... 24 Figure 13: Typical failed LVL and spruce connections with fatigue loading .............................. 26 Figure 14: Typical load-slip responses for un-failed fatigue specimens ...................................... 27 Figure 15: Typical load-slip responses for failed fatigue specimens............................................ 27 Figure 16: Load level vs. fatigue life ............................................................................................ 28


1. Project objectives The long-term goal of this project was to maintain competitiveness of Engineered Wood Products (EWP) in relation to other construction materials. The specific objectives were:

• To evaluate performance of traditional and modern fasteners used in Canadian manufactured EWP, with emphasis on Laminated Veneer Lumber (LVL).

• To establish principles and test protocols for evaluation of non-standard fasteners. These will set expectations for new products entering the marketplace via Product Evaluation Reports of the Canadian Construction Materials Centre at NRC (or equivalent).

• To implement the design information (for Canadian manufactured EWP) and develop protocols within the framework of Canadian standards falling under the jurisdiction of the CSAO86 ‘Engineering Design in Wood’ Technical Committee.

2. Introduction EWP currently represents about 5 percent of the overall structural wood product demand in North America, with a projected annual production of LVL, I-Joists and glued-laminated-timber (glulam) alone exceeding the equivalent of over 8 million cubic meters per annum (UNECE/FAO, 2003). The projection neglects the possibility of increased activity in the non-residential construction sector and is probably conservative. Survey information reveals that lack of high quality connections data, and design know-how and methods, for EWP is a primary inhibition to extended non-residential applications of such material (Snow et al, 2006). Removing that any other knowledge barriers are key to taking consumption levels beyond what has currently been envisioned. In Canada LVL is the prime type of EWP that is manufactured and used in building construction. However, despite its excellent structural properties LVL, and other EWP, is largely employed as a substitute for sawn lumber in residential and other ‘small building’ applications that do not fully exploit its potential as an engineering material. Knowledge of the strength and stiffness characteristics of mechanical connections, in any wood product, are essential to what engineers can achieve technically, and to creation of economically feasible design solutions (Madsen 1998). To date only very conservative approaches have been used to derive recommended design properties for connections in EWP, based on an assumption of equivalency between the performances of EWP and sawn lumber connections (Snow 2006). In fact, it is not an overstatement to say that the design practices that regulatory authorities and design codes currently regard as acceptable are crude and outdated. This is true in Canada and elsewhere. Historically there were some legitimate concerns about the quality control, and moisture and fatigue sensitivities, of wood-based composites. However, in the modern context, the discrepancy in what could be done with EWP reflects mostly that the store of research, data and know-how to draw on in the design of connections is very limited. The completed UNB2 project ‘Design Methods for Connections in Engineered Wood Structures’ demonstrated in principle that there are potentially large increases available in design capacities of connections in EWP (Smith et al, 2006). If such increases were to be accessed, the wood products industry would benefit because it would become technically possible to design larger


and more innovative structures that are permitted under existing practices. Also, taking the benefit of actual versus assumed performance will reduce costs of making connections in terms of reduced material, reduced labour and simpler design detailing. There are two primary ways by which engineers design structural connections with explicit approval of most building regulatory authorities in Canada. First, it is permitted to design connections based on rules that are detailed in the Canadian Standards Association’s (CSA) Standard 086-01 ‘Engineering Design in Wood’ (CSA 2004). This is the traditional method and the rules are those that nearly all engineers follow when designing connections in commodity products like sawn lumber, glulam, plywood and Oriented Strand Board. Second, it is permitted to design connections based on information in third-party product assessment reports produced by the National Research Council’s Canadian Construction Materials Centre (CCMC), or any other agency that the officials within a particular regulatory jurisdictions deem expert in the topic. This approach is normally followed when proprietary products are involved (joined materials, fasteners, or both joined materials and fasteners). Product assessments reports can contain suggested for design capacities and suitability of products for certain applications. There is a third method for designing structural systems and components like connections. Engineers who are licensed to practice within a province or territory may certify the acceptability of designs that are based on information other than that which is contained in Canadian design codes or CCMC evaluation reports (or equivalent). This is very rare for timber structures and involves liability issues that often are unacceptable to engineers. For the purposes of this project it is assumes that designs of EWP connections will in based on provisions of the CSA Standard 086-01 or product assessments by organizations like CCMC. The situation in the USA and other countries that are primary markets for Canadian manufactured EWP is equivalent to what has just been described. The research project emphasised: • Providing engineers with information needed to fully exploit the structural capabilities of

EWP materials like LVL, based on application of rational engineering methods. • Methods of establishing design properties that can be incorporated into timber design codes

and their supporting standards. • Assessment methods that should be adopted by third-party agencies. • Need for consistency between the parallel routes by which engineers may design connections

in commodity or specialty EWP materials. • Harmonization with international practice and leading development of those harmonized

practices. The general framework for all the above was established in the UNB2 project and the newer project is part of putting flesh on that framework and applying it in practice. Major project activities were: 1) testing of connections to supplement data from the UNB2 project; and 2) working with external experts and codes and standards bodies to translate that data for implementation. All activities were planned in consultation with the project champion, project liaisons, and other experts as appropriate. Specific project activities were: • Testing of connections with emphasis on LVL, using slender dowel-type fasteners including

common and spiral nails, lag screws, and modern self-taping screws. Connections tested


were single or double shear arrangements where fasteners load EWP either parallel or perpendicular to the strong axis of material symmetry1. The number of fasteners ranged from one to ten. Additionally tests were performed aimed at elucidating the performance of bolted LVL connections subjected to cyclic fatigue loading.

• Development of new test protocols, with emphasis on connections that cause bolt type fasteners to load EWP perpendicular to the strong axis of material symmetry.

• Technical information and documents in support of the activities of: CSA Technical Committee 086, International Organization for Standardization’s Technical Committee 165 ‘Timber Structures’ (ISO-TC165), and the International Council for Building Research Studies and Documentation’s Working Commission 18 ‘Timber Design’ (CIB-W18).

• Complex numerical and analytical models for predicting ultimate capacities and associated failure modes for connections in EWP.

• Use of complex models as an aid to evaluation of simpler ‘design level’ models that can reasonably be expected to be employed by practicing engineers.

3. Test Program 3.1 LVL connection tests Test setup: All specimens consisted of three LVL members connected using slender dowel type fasteners that were common nails, spiral nails, lag screws, and a proprietary type of self-taping screw. The nail fasteners used had regular bright coating and met the ASTM specification for driven fasteners (ASTM, 2001). The lag screws had cut thread and met the requirement of the ASME standard (ASME, 1996). Because of the epoxy-coating colour on their surfaces, the proprietary screws are referred to below as ‘red screws’ (Fasten Master, 2006). The LVL used in this project was Grade 3300Fb-2.0E with species composition primarily birch/aspen (CCMC, 2006). The LVL members of specimens were cut from billets with a 44mm x 298 mm cross-section by 2440 mm long. Members themselves were 44mm x 146mm or 44mm x 95mm in cross-section by 300 to 600 mm long. To mimic effects of inherent variability in LVL, test groups were created by randomly selecting pieces from billets. Figures 1 to 3 and Tables 1 to 3 show detailed connection configurations under parallel, perpendicular and off-axis loadings. Fasteners nominal diameters used were: 3.66 mm for common nails, 3.10 mm for spiral nails, 6.35 mm for lag screws, and 5.79 mm for red screws. Nails used were 76 mm long with non-coated and non-galvanized steel nails manufactured from the same batch. Lag screws and red screws of two lengths of were: 88 mm and 127 mm long. The number of fasteners was varied from 1 to 20 per connection in single and double shear loading arrangements. All specimens were conditioned in an environmental chamber at 20ºC and 65% RH, which produces an EMC of 12% in sawn lumber, but only an EMC of 7-8% in LVL (because of the gluing and hot pressing). LVL density, represented by specific gravity value, was 0.60 (CoV= 5%).

1 For LVL the term strong axis of material symmetry corresponds to the direction parallel to which axes of longitudinal wood cells in veneers layers are oriented. For other EWP it is the direction in which the material is stiffest and strongest and usually will be parallel to the longitudinal axis of a billet of material.


The nails were driven manually from the side to middle member and no pre-drilled holes were used to guide the nails. The lag screws were inserted through 3/16” (4.8 mm) pre-drilled holes in the members using an electric drill with hex driver bit, and the red screws were installed similarly but without pre-drilling. Tests: LVL connection tests followed procedure outlined in ASTM D1761 ‘Standard Test Methods for Mechanical Fasteners in Wood’ (ASTM 2000). Each test was replicated ten times in situations where members were loaded parallel or perpendicular to their axes. Six replicates were used for connections with off-axis loading of a member. Tests were conducted under static compression with fasteners loading members in parallel, perpendicular, and off-axis (at an angle other than 0º or 90º) to the strong axis of material symmetry (Figures 1 to 3). For perpendicular-to-strong axis loading only the middle member was loaded perpendicular to the strong direction. The two-side members were loaded parallel to the strong axis direction. For connections that load members at an angle to the strong axis direction, only the 30º from parallel condition was studied due to limits of LVL-billet dimensions. In such cases only the middle member was loaded off-axis (Figure 3). To prevent friction, plastic sheets were inserted between the side and middle members. All tests were carried out using an Instron universal testing machine with the rate of loading corresponding to crosshead movement of 2.54 mm/min. Relative movement of members in the direction of loading due to applied forces (slip measurements) were made. These movements were recorded at both of the joint planes using a Linear Variable Differential Transformer (LVDT), Figures 1-3. Tests were terminated either when the load dropped without potential for recovery, or when excessive deformation was observed. Excessive deformation that terminated a test was taken to be 20-40mm.

Figure 1: Connection loaded parallel to strong axis of LVL with four fasteners under single

shear arrangement (44x146 mm member cross-sections)

a 3a 1a 3

P

0.5P 0.5P

a 2a 4t a 4c

LVDT2 LVDT1

a 3a 1a 3

a 3a 1a 3

P

0.5P 0.5P

a 2a 4t a 4c a 2a 4t a 4c a 2a 4t a 4c

LVDT2 LVDT1

L


Table 1: Connections loaded parallel to strong axis of LVL

No Group ID Description

L (mm) N

a1 (mm)

a2 (mm)

a3 (mm)

a4c (mm)

a4t (mm)

1 LVL0-cn2 Common nail, one on each side 380 2 - - 100 73 73 2 LVL0-sn2 Spiral nail, one on each side 380 2 - - 100 73 73 3 LVL0-ls2 3.5" Lag screw, one on each side 380 2 - - 100 73 73 4 LVL0-rs2 3.5" Red screw, one on each side 380 2 - - 100 73 73 5 LVL0-ls8 3.5" Lag Screw, four on each side 380 8 50 40 70 46 60 6 LVL0-rs8 3.5" Red Screw, four on each side 380 8 50 40 70 46 60 7 LVL0-cn10 Common nail, ten on each side 600 20 70 70 50 35 41 8 LVL0-sn10 Spiral nail, ten on each side 600 20 65 70 45 35 41 9 LVL0-ls10 3.5" Lag screw, ten on each side 600 20 50 40 70 46 60

10 LVL0-rs10 3.5" Red screw, ten on each side 600 20 50 40 70 46 60 11 LVL0-5ls1 5.0" Lag screw* 380 1 - - 100 73 73 12 LVL0-5rs1 5.0" Red screw* 380 1 - - 100 73 73 13 LVL0-5ls4 5.0" Lag screw* 380 4 50 46 70 50 50 14 LVL0-5rs4 5.0" Red screw* 380 4 50 46 70 50 50

Notes: N= total number of fasteners per specimen. L = member length. * = double shear arrangement

Figure 2: Connection loaded perpendicular to strong axis of LVL with four fasteners under single shear arrangement (44x146 mm member cross-sections)

a2a4t a4c

600

P

300 a3

a1

a3

0.5P 0.5P

LVDT2LVDT1

a2a4t a4c

600

P

300 a3

a1

a3

a3

a1

a3

0.5P 0.5P

LVDT2LVDT1

L

H


Table 2: Connections loaded perpendicular to strong axis of LVL

No Group ID Description L

(mm)H

(mm) N a1

(mm)a2

(mm) a3

(mm) a4c

(mm)a4t

(mm)15 LVL90-cn2 Common nail, one on each side 600 300 2 - - 73 73 73 16 LVL90-sn2 Spiral nail, one on each side 600 300 2 - - 73 73 73 17 LVL90-ls2 3.5" Lag screw, one on each side 600 300 2 - - 73 73 73 18 LVL90-rs2 3.5" Red screw, one on each side 600 300 2 - - 73 73 73 19 LVL90-cn8 Common nail, four on each side 600 300 8 50 70 50 36 40 20 LVL90-sn8 Spiral nail, four on each side 600 300 8 50 60 50 36 50 21 LVL90-ls8 3.5" Lag screw, four on each side 600 300 8 50 40 50 46 60 22 LVL90-rs8 3.5" Red screw, four on each side 600 300 8 50 40 50 46 60 23 LVL90-ls1 5.0" Lag screw* 600 300 1 - - 73 73 73 24 LVL90-rs1 5.0" Red screw* 600 300 1 - - 73 73 73 25 LVL90-ls4 5.0" Lag screw* 600 300 4 50 40 50 46 60 26 LVL90-rs4 5.0" Red screw* 600 300 4 50 40 50 46 60

Notes: N= total number of fasteners per specimen. L=member length * = double shear arrangement

Figure 3: LVL connections loaded 30º with respect to strong axis of LVL with two fasteners under single shear arrangement (44x95 mm member cross-sections)

P

0.5P 0.5P

30 0

Strong axis direction

side members, loaded parallel

30 0 30 0



a 2

a 4c a 4t

L=381

a 3

a 3a 1

LVDT2

LVDT1

L

P

0.5P 0.5P

30 0



30 0 30 0



a 2

a 4c a 4t

a 3

a 3a 1

LVDT2

LVDT1

30 0



30 0 30 0



a 2

a 4c a 4t

30 0 30 0



30 0 30 0



a 2

a 4c a 4t

a 2

a 4c a 4t

a 3

a 3a 1

a 3

a 3a 1

LVDT2

LVDT1

L

L


Table 3: Connection loaded off-axis at 30º to strong axis of LVL

No Group ID Description L

(mm) N a1

(mm)a2

(mm) a3

(mm) a4c

(mm) a4t

(mm) 27 LVL30-ls2 3.5" Lag screw, one on each side 380 2 - - 60 48 48 28 LVL30-rs2 3.5" Red screw, one on each side 380 2 - - 60 48 48 29 LVL30-ls4 3.5" Lag screw, two on each side 380 4 35 35 60 30 30 30 LVL30-rs4 3.5" Red screw, two on each side 380 4 35 35 60 30 30

Notes: N= total number of fasteners per specimen. Pu = average capacity for the set of specimens. L=member length

Results: All specimens failed prior to reaching the maximum deformation permitted by the test arrangement, permitting the ultimate (maximum) loads to be determined from load-slip records. As expected, the majority of specimens failed via a ductile mechanism; either by wood (LVL) embedment failure, fastener yielding, or a combination of the two (Figure 4a). A few specimens failed by a brittle row-shear-out (Figure 4b). Figures 5(a) to 5(d) shows illustrative load-slip curves for perpendicular to strong axis loading with for each fastener. All representative load-slip curves for each group can be seen in Appendix A. Test results for each group of specimens are summarised in Tables 4 to 6. In general it can be seen that for loading parallel and perpendicular to the strong LVL axis, variability in ultimate load was small with coefficients of variation (CoV) less than 10%. For screw fastener connection where LVL was loaded at 30º to the strong LVL axis, relatively large variability was observed, CoV between 13 and 29%. This is attributed to the possibility that under off-axis loading alternative types of failure mechanism are possible at the dimensional scale of wood cells. The ultimate capacity of connections with ten nails per joint interface that loaded LVL parallel to the strong axis in single shear (20 nails per specimen), was considerably less than ten times the ultimate capacity of similar connections with one fastener per joint plane. The discount in capacity for the ‘group effect’ was about 50 to 60% per nail. The corresponding discount factor was about 67% for multiple nail fasteners loading LVL perpendicular to the strong axis in single shear (4 versus 1 nail per joint plane). The more significant discount factors obtained in this latter case is a well known phenomenon for timber connections (Quenneville 2006). Different strength performance was observed in the case of multiple screws versus a single screw per joint plane. Then the discount factor for ten screws loading LVL parallel to its strong axis was about 30%. For four screws versus one loading LVL perpendicular to its strong axis was about 10%. Off-axis loading caused a capacity reduction of about 45% per screw for two versus one screw per joint plane. For the parallel the strong axis loading direction, using four 127 mm (5”) screws does not produce higher strength than using eight 88 mm (3.5”) screws. However, the reverse situation was obtained for when the LVL was loaded perpendicular to the strong axis. It will be shown later in this report (section 4) that simplistic approaches to design of LVL connections with nails or screws (of various types) do not work. If simple methods are followed


to assign ultimate capacities, they must be very conservative in most instances. For example, adopting assumptions like assuming that the direction of loading relative to the materials strong axis, or that a connection with N fasteners had N time the capacity of a connection with one fastener would be overly conservative. Results given here are purely indicative of the type of information that could be extracted from the test data. Deciding how design rules should be framed and what, if any, simplifications are appropriate is primarily the business of relevant code writing bodies. The data collected during the project could be manipulated in many ways to support decision making of this type. Thus the raw test results are a valuable resource that will be made available in the first instance to the Fastenings Subcommittee of the CSA086 Technical Committee.

(a) Nails

(b) Screws

Figure 4: Connection failure mechanisms

spiral nails

common nails

Lag screw Red screw


(a) common nail (b) spiral nail

(c) lag screw (d) red screw

Figure 5: Illustrative load-slip curves for perpendicular to strong axis loading (four fasteners per side, single shear)

Slip (mm)

Slip (mm) Slip (mm)

Slip (mm)

Load (N) Load (N)

Load (N) Load (N)


Table 4: Test results for connections loaded parallel to strong axis of LVL

No Group ID Description Pu

(kN) CoV (%)

F1 (kN)

1 LVL0-cn2 Common nail, one on each side 7.8 2.5 3.9 2 LVL0-sn2 Spiral nail, one on each side 5.9 4.8 3.0 3 LVL0-ls2 3.5" Lag screw, one on each side 9.7 9.4 4.9 4 LVL0-rs2 3.5" Red screw, one on each side 9.5 5.6 4.8 5 LVL0-ls8 3.5" Lag Screw, four on each side 38.0 4.0 4.8 6 LVL0-rs8 3.5" Red Screw, four on each side 34.7 5.0 4.3 7 LVL0-cn10 Common nail, ten on each side 33.6 1.7 1.7 8 LVL0-sn10 Spiral nail, ten on each side 29.4 4.2 1.5 9 LVL0-ls10 3.5" Lag screw, ten on each side 68.5 7.1 3.4

10 LVL0-rs10 3.5" Red screw, ten on each side 58.8 5.7 2.9 11 LVL0-5ls1 5.0" Lag screw* 9.9 5.5 5.0 12 LVL0-5rs1 5.0" Red screw* 8.7 7.8 4.4 13 LVL0-5ls4 5.0" Lag screw* 32.2 5.4 4.1 14 LVL0-5rs4 5.0" Red screw* 35.9 3.0 4.5

Notes: Pu = average capacity for the set of specimens. F1= per fastener value (per shear plane). * = double shear arrangement

Table 5: Test results for connections loaded perpendicular to strong axis of LVL


Pu (kN)

CoV (%)

F1 (kN)

15 LVL90-cn2 Common nail, one on each side 9.7 5.0 4.9 16 LVL90-sn2 Spiral nail, one on each side 8.4 4.0 4.2 17 LVL90-ls2 3.5" Lag screw, one on each side 9.6 6.2 4.8 18 LVL90-rs2 3.5" Red screw, one on each side 11.4 4.1 5.7 19 LVL90-cn8 Common nail, four on each side 12.2 3.0 1.5 20 LVL90-sn8 Spiral nail, four on each side 11.7 4.2 1.5 21 LVL90-ls8 3.5" Lag screw, four on each side 30.6 6.9 3.8 22 LVL90-rs8 3.5" Red screw, four on each side 29.5 2.7 3.7 23 LVL90-ls1 5.0" Lag screw* 12.5 4. 9 3.1 24 LVL90-rs1 5.0" Red screw* 11.8 4.5 3.0 25 LVL90-ls4 5.0" Lag screw* 34.3 10.0 4.3 26 LVL90-rs4 5.0" Red screw* 34.7 4.0 4.3

Notes: Pu = average capacity for the set of specimens. F1= per fastener value (per shear plane). * = double shear arrangement

Table 6: Test results for connection loaded off-axis at 30º to strong axis of LVL


Pu (kN)

CoV (%)

F1 (kN)

27 LVL30-ls2 3.5" Lag screw, one on each side 15.8 13.3 28 LVL30-rs2 3.5" Red screw, one on each side 25.0 27.1 29 LVL30-ls4 3.5" Lag screw, two on each side 17.3 15.3 30 LVL30-rs4 3.5" Red screw, two on each side 14.7 28.8

Notes: Pu = average capacity for the set of specimens. F1= per fastener value


3.3 Development of testing method appropriate for EWP connections During the complete UNB2 project it was identified that the ASTM D5652-95 ‘Standard Test Methods for Bolted Connections in Wood and Wood Base Products’ (ASTM 2000) is unreliable. Specifically the problems relate to evaluation of capacities of EWP connections where dowel type fasteners load EWP perpendicular to the strong axis of material symmetry (Smith et al 2006, Snow 2006). The purpose of this activity was to develop a testing arrangement to replace the current ASTM three-point bending approach shown in Figure 6a. For EWP that arrangement can cause failure mechanisms and capacities that equate to the wood member failing as a beam, and that are unrelated to connection performance. Primary focus was on using the propped beam arrangement in Figure 6b. This new arrangement was expected to reliably produce connection failures at the right-hand end as seen in the diagram. It should be noted that the designation propped cantilever is intended to reflect that connections evaluated by the new approach can include those that transmit both shear and moment forces.

(a) ASTM arrangement (b) New arrangement

Figure 6: Test setups for perpendicular to the strong axis loading of connections New test setup: In the new test method, single and four bolted connections were used with double shear loading of a 12.7 mm bolt (SAE Grade 8). Wood member materials used were sawn pine lumber, LVL and Laminated Strand Lumber (LSL), with 12.7 mm steel plates as side members. The LVL used in this project was Temlam® Grade 3300Fb-2.0E with species composition primarily birch/aspen (CCMC, 2006). The LSL used was TimberStrand Grade 2250Fb-1.5E with primary species composition yellow poplar (CCMC, 2006b). Connection geometry specified (spacing, edge distance, and end distance) were the same as used in the ASTM based method as investigated previously (Smith et al, 2006; Snow, 2006), Figure 7. Test followed the same procedure as discussed in the UNB2 project. All pine, LVL, and LSL specimens were stored in an environment of 20ºC and 65% RH until they achieved an equilibrium MC. Subsidiary test indicated that pine specimens reached an Equilibrium Moisture Content (EMC) of between 12 and 13%, and LVL and LSL attained EMCs of 7 to 9%. Based on oven-dry volume and over-dry mass, specific gravity of pine, LVL, and LSL were 0.40 (CoV=5%), 0.60 (C0V=5%), 0.63 (CoV=6%), respectively. LVL and LSL specimens were

0.5LOAD 0.5LOAD

LOAD

SIDE MEMBERS

SUPPORT SUPPORTFASTENERS

LOAD

SIDE MEMBERS


MAIN MEMBER

LOAD

SUPPORT

SIDE MEMBERS

MAIN MEMBER

FASTENERS

LOAD

SUPPORT

SIDE MEMBERS

MAIN MEMBER

FASTENERS

0.5LOAD 0.5LOAD


44mm thick and 95mm deep, and pine specimens were 38mm x 89mm. All wood members in specimens were 610mm long. The clear span between supports in the test arrangement was 305mm. The member length and clear span matched those for the ASTM specimen. Bolt surface were unpolished. Bolts holes were drilled into the main members using stainless steel Forstner bits to ensure smoothness and straightness of drilling process. Drill holes were 20 mm for 19 mm bolts and 10 mm for the 9.5 mm bolts.

(a) single bolt

(b) four bolts

Figure 7: Connection configuration using new test method

609

95

12.7 mm steel side plates

19mm boltLoad

280

0.5xLoad 0.5xLoad

LVDT1LVDT2

609

95


19mm boltLoad

280

0.5xLoad 0.5xLoad

LVDT1LVDT2

609

38


280 40

3028.5

Load

0.5xLoad 0.5xLoad

6.4mm bolts38

609

38


280 40

3028.5

Load

0.5xLoad 0.5xLoad

6.4mm bolts38


Tests: Matched tests (i.e. matched wood materials) were performed using the ASTM and new test arrangements, with six replicates per arrangement. Tests were conducted under static load conditions using an Instron universal test machine with a crosshead movement of 2.54 mm/min. The testing speed matched that in the ASTM method. Two LVDT were placed to facilitate determination of failure related to the connection configurations. One LVDT was positioned at the middle span of main member to record flexural deflection, and another LVDT was place at the support to check shear deformation (Figure 7). Results: All specimens were loaded until failure was reached. Figure 8 shows typical load versus deformation records for specimens, and Figure 9 shows typical failed specimens. Failures for pine and LVL connections involved cleavage type splitting in the wood member, with crack propagation from the bolt hole(s) and parallel to the strong axis of material symmetry, Figures 9a&b. For LSL connections failures were due to bending failure of the wood member with most failure at tension edge beneath the central loading point, Figure 9c. Although the shapes of load-deformation curves are fairly similar for ASTM and the new arrangement, failures under the new arrangement occurred more rapid and at lower levels of deformation. Table 7 summarizes comparison of the ultimate loads between the two test methods. It should be noted here that the ultimate strengths measured for the new test were the mid-span loads that caused the connection failure, not the actual strength of the connection, which was positioned at one of the supports. Theoretically, by assuming symmetrical strength at the supports, the connection strength is simply half of the ultimate strength. It can be noted that the new method results in lower connection strength giving indication that lower bound values are not captured in the ASTM based test. This is no surprise since there is interaction between moment and shear forces at the mid span. While in the proposed test method, only shear at the member support is the main contributing force to failure. Based on the fail modes and observations during the tests, it is clear that both the ASTM and new test arrangements can create a failure that relates to connection performance, and simple are ways of testing the material loaded as a beam. However, it is believed that genuine connection failures are produced by the new method for EWP such as LVL, and for sawn lumber. Clearly design of a generally robust, yet practical, test method for all types of connections in EWP has still to be identified. The analytical modelling within the next subsection of this report is intended to elucidate the issues. Table 7: Comparison of ultimate strengths of alternative perpendicular to the strong axis loading

of connections in EWP

Type of wood

Fastener(s)

Connection strength, ASTM test (kN)*

Ultimate load, new test (kN),

Connection strength, new test

(kN

Pine-1 Single 19mm-bolt 8.6 (10) 9.8 (20) 4.9 Pine-2 Four 9.5mm-bolts 12.9 (4) 19.4 (17) 9.7 LVL-1 Single 19mm-bolt 16.6 (2) 16.2 (9) 8.1


LVL-2 Four 9.5mm-bolts 15.5 (7) 28.6 (9) 14.3 LSL-1 Single 19mm-bolt 28.4 (11) 30.9 (12) 15.5 LSL-2 Four 9.5mm-bolts 34.4 (13) 41.9 (6) 21.0

Note: * Data taken from the UNB2 project. Values in parentheses are CoV (%).

05

1015202530354045

0.00 5.00 10.00 15.00 20.00

Crosshead Displacement (mm)

Load

(kN

)

(a) ASTM test

(b) New test

Figure 8: Typical load deformation responses for perpendicular to the strong axis loading of connections

LVL

LSL

pine

-50-45-40-35-30-25-20-15-10

-50

0.0 -0.1 0.7 1.6 2.4 3.1 3.7 4.2 4.8

Mid-span displacement (mm)

Load

(kN

)

LSL

LVL

Pine

-50-45-40-35-30-25-20-15-10

-50

0.0 -0.1 0.7 1.6 2.4 3.1 3.7 4.2 4.8

Mid-span displacement (mm)

Load

(kN

)

LSL

LVL

Pine


(a) pine

(b) LVL

(c) LSL

Figure 9: Failure mechanisms for perpendicular to the strong axis loading of connections Numerical modelling: A numerical model was developed using the discrete lattice network approach that was reported previously (Smith et al 2006, Snow 2006). Focus was prediction and explanation of the failure mechanism under the new test configuration, based on connections in LSL with a single 12.7 mm bolt. A uniform 2-D lattice of elements was applied to form the geometry of the LSL member and bolt, Figure 10. In total the lattice contained 1818 pin-ended rod elements based on

ASTM test New test

ASTM test New test

New testASTM test


a lattice panel size of 5mm x 5mm. The interface between bolt and bolthole, which functions as a boundary support, was simulated assuming frictional contact behaviour (Snow 2006). A uniform pressure (over 3mm wide strip) represented the point load at mid-span. That central loading point was advanced incrementally to cause progressive damage accumulation in the lattice, up to a deformation of 15 mm. Figure 10 illustrates the predicted progression of damage at various amounts of movement of the central load. Numerical results indicate that failure is caused by excessive bending deformation (as in experiments). Initiation is by compression failure near the top surface of the LSL member, under the load point. Instability and global failure occurs due to tension induced damage in the member at mid-span. The model also predicts the presence of crushing beneath the bolt (in the connection). The study has so far not been carried far enough to fully elucidate what constitutes an ideal test arrangement. However, indications from the modelling are that applying the central load over a larger area and reducing that clear span will solve problems encountered, and yield a robustly reliable test method. Note: Although the project has ended it is intended to create numerical models for LVL and sawn lumber, and to perform experiments using arrangement predicted to be optimal. After that proposals for a ‘new method’ will be made to ASTM and other committees responsible for writing and maintaining written test standards.

(a) before loading

(b) deformation = 5 mm

(c) deformation =10 mm

(d) deformation =15 mm

Pfailure

P2

P1

Pinitial


Figure 10: Predicted progression of the failure process for the 12.7 mm bolt loading LSL perpendicular to the strong axis (bolt loads the right-hand end)

3.3 Fatigue tests on connections Fatigue tests were performed aimed at elucidating the likely importance of fatigue processes in wood connections subjected to cyclic loads, and especially whether that is a crucial factor in design of connections in EWP. Bolted connections loaded in tension were selected because those are governed by shear strength of wood members, which is known to be a fatigue sensitive stress situation (Smith et al 2003). Method: Connections had six 12.7 mm bolts (SAE Grade 8) loading a wood member parallel to the strong axis in a double shear arrangement, with 12.7 mm thick steel side plates, Figure 11. The wood members were 44mm x 164mm LVL or 38mm x 140mm sawn spruce lumber. The LVL used was Temlam® produced by the Jager Building Systems (CCMC, 2006). All specimens were conditioned in an environmental chamber at 20ºC and 65% RH, which produces an EMC of 12% in sawn lumber, but only an EMC of 7-8% in LVL. SG for LVL was 0.60 (CoV= 5%) and for spruce was 0.40 (CoV=7%). All cyclic load specimens had the connection geometry shown in Figure 11, and matched that of specimens in the completed UNB2 project. The load regime followed procedure illustrated in Figure 12. Each load cycle had a sinusoidal waveform and the loading frequency was 0.2 Hz. Maximum load levels for spruce ranged from 60 to 100% of the average short-term static strength (ultimate load) of matched specimens. While peak load levels for LVL ranged from 80 to 105% of the average short-term static strength of matched specimens2. The minimum load was always 10% of the average short-term static strength. Two LVDTs measured the slip of the wood member relative to the side plates. Ultimate load for matched specimens was 91.0 kN (CoV= 9.6%) for LVL, and 72.6 kN (CoV= 15.7%) for spruce (Smith et al 2006). Test schedule of cyclic loading tests is summarized in Table 8. Load was applied to specimens for either 10,000 cycles or until premature failure. Nominally it is assumed that if a specimen survived 10,000 load repetitions then the maximum load level applied had been in the vicinity of the fatigue limit. Because the tests took up to about 14 hours, a special cage made of plastic was designed around the test set up to prevent the specimens taking up excessive moisture from the surrounding environment (Figure 11b). 2 Because of inherent variability in capacities of nominally identical specimens, actual cyclic load levels (expressed as a percentage of the static strength) could not be determined. Therefore all load levels discussed here are nominal rather than true values.


(a) connection geometry (b) test environment

Figure 11: Fatigue test specimens

Figure 12: Schematic of cyclic load regime Results: All failure mechanisms for fatigue tests were of the brittle row-shear-out for LVL and spruce connections, without residual bending deformation of the bolts, Figure 13. This matched the mechanisms for static loading. Typical cyclic load-slip records are shown in Figures 14 and 15. For specimens that did not fail within 10,000 load cycles, responses were linear with very small amounts of drift in the slip amplitude (marching of hysteresis loops) as the number of applied load cycles increased. Individual hysteresis loops consumed little of the externally applied energy (Figure 14). For specimens that failed in less than 10,000 cycles, there was significant

Note: N=number of cycles

……

Minload

time

load

Max load

N=1000N=1

……

Minload

time

load

Max load

N=1000N=1

241

914

64

89

505050

LVL/spruce (2x6)

steel plates

bolts

241

914

64

89

505050

241

914

64

89

505050

LVL/spruce (2x6)

steel plates

bolts


marching of hysteresis loops, that became wider and consumed significant energy, near failure (Figure 15). Table 8 summarizes the test results. Figure 16 presents the relation between the nominal applied maximum load level and number of cycles to cause failure, for LVL and spruce connections. Fatigue life equations can be determined from this data using logarithmic fitting technique. It can be noted that the fitted fatigue equation for LVL is above that for solid wood spruce. Therefore, based on this study, it can be concluded that LVL connections have superior fatigue performance than spruce connections. This is attributed to LVL being a product that is characterised by a degree of cross-lamination between wood cells in adjacent veneer layers, which reinforces it in shear relative to sawn lumber. The findings are liable to be different for types of EWP with other structural features. Approximately, the results indicate that the fatigue limit for connections of the type investigated is about 75% of the short-term static strength.

Table 8: Schedule of cyclic load (fatigue) tests on bolted specimens LVL specimen (MC = 7 to 9%)

Spruce specimen (MC = 12 to 14%)

Group

N-failure (cycles)

Max slip (mm)

Group

N-failure (cycles)

Max slip (mm)

LVL-A-80% > 10,000 1.8 Spruce-A-65% > 10,000 4.0 LVL-A-85% > 10,000 2.0 Spruce-A-70% > 10,000 4.2 LVL-A-90% 9,000 - Spruce-A-75% 8,080 - Spruce-A-80% > 10,000 2.1 LVL-B-85% > 10,000 5.6 Spruce-A-85% 7,168 - LVL-B-90% > 10,000 1.7 LVL-B-95% > 10,000 2.0 Spruce-B-60% > 10,000 3.0 LVL-B-100% 2,026 3.0 Spruce-B-65% > 10,000 3.1 Spruce-B-70% > 10,000 1.2 LVL-C-80% > 10,000 0.3 Spruce-B-75% > 10,000 1.4 LVL-C-85% 4,430 - Spruce-B-80% > 10,000 1.6 LVL-C-90% 7976 - Spruce-B-85% > 10,000 2.1 LVL-C-95% >10,000 2.8 Spruce-B-90% > 10,000 3.2 LVL-C-100% 2,452 - Spruce-B-100% 2,952 - LVL-D-80% > 10,000 1.5 Spruce-C-70% > 10,000 2.1 LVL-D-85% 5,396 Spruce-C-75% > 10,000 2.3 LVL-D-90% > 10,000 1.7 Spruce-C-80% 1,022 - LVL-D-100% 1,644 Spruce-D-70% > 10,000 4.5 LVL-E-80% > 10,000 2.6 Spruce-D-75% 1,018 - LVL-E-85% > 10,000 2.7 LVL-E-90% > 10,000 2.9 Spruce-E-70% > 10,000 0.7 LVL-E-100% > 10,000 3.3 Spruce-E-75% > 10,000 0.9 LVL-E-105% 228 - Spruce-E-80% 6,824 -


Figure 13: Typical failed LVL and spruce connections with fatigue loading

(a) LVL (at 80% of the ultimate load)

(b) spruce (at 80% of the ultimate load)

Note: selected cycles plotted at N=2,4,10,100,1000, and 5000



Figure 14: Typical load-slip responses for un-failed fatigue specimens

(a) LVL (100% ultimate load)

(b) spruce (100% ultimate load)

Figure 15: Typical load-slip responses for failed fatigue specimens


Note: selected cycles plotted at N=2,4,10,100,1000, and


0

0.25

0.5

0.75

1

1.25

1 10 100 1000 10000 100000

N, number of cycles (log-scale)

Load

leve

l (%

)

LVLspruceLog. (LVL)Log. (spruce)

Figure 16: Load level vs. fatigue life 4. Test results and prediction of connection strength This discussion considers the ability of design level models to predict capacities of EWP connections employing mechanical fasteners. It draws on test results from this and the completed UNB2 projects. 4.1 Connections with slender fasteners ‘Slender fasteners’ for the purposes of this discussion are nails and screws of the types investigated in this project. Tables 6 to 8 summarize the experimental and predicted lateral strengths of connections with different dowel type fasteners. In these comparisons test capacities are taken to be the 5th percentile ultimate strength determined by fitting a two-parameter Weibull distribution to the data. This parallels the common practice that is the basis for estimation of reference design resistances in Canada and USA. Failures are classified as having ductile or brittle modes. Model predictions are based on the so-called European Yield Model (EYM) equations in the Canadian national timber design code (CSA 2004). LVL embedment strength that enters the EYM equations was estimated based on the equation provided in the code, which is function of density and fasteners diameter. The density of LVL is taken to be 0.6 (CoV=5%). Fasteners strengths bending strengths that also enter EYM equations were also calculated based on code values. For brittle failure modes, the row-shear-out mechanism was assumed for parallel-to-strong axis loading, and the wood splitting for mechanism was assumed for perpendicular to strong axis loading. The calculations were based on equations proposed by Quenneville (2006). Detail of connection geometries are given in Figures 1 to 3 and Tables 1 to 3.


The predicted values given in Tables 9-11 are all unfactored, i.e. no so-called resistance factor (know as the φ value) is introduced as in actual design. Design code resistances have been adjusted to a short-term static loading basis to match test conditions assuming KD = 1.25. For connections with load applied at 30º to the strong material axis direction, only ductile mode was calculated, since no design level brittle model exists or has been proposed. Comparisons here are an assessment of current or proposed design practices and not suggestions for what should be done. For parallel to strong axis loading (Table 9) models always predicted that a ductile mode would govern, which is correct. However, those ductile model predictions tended (but not always) to be conservative relative to test capacities. Closer examination indicates that the ductile model under-predicts strength for connection using single fasteners with a single shear arrangement. The ductile model predictions approach the test values for situations where fasteners are loaded in double shear. For multiple screw fasteners with single shear arrangements, predicted ductile capacities are almost equal to the test values. For perpendicular to strong axis loading (Table 10), predicted ductile capacities generally considerably lower than test resistances. However, for all single fasteners (singe and double shear arrangements) the predicted brittle capacities are not dissimilar to test values, but can be non-conservative. Clearly there are major deficiencies in the models because mostly test specimens exhibited ductile failures. The same observations apply to multiple nail connections with a single shear arrangement. For multiple screw fasteners (single and double shear arrangements), brittle mode predictions yield lower values than tests or ductile model predictions. Again there are clear model deficiencies. For off-axis (30º to the strong axis) loading, the predicted ductile mode values are lower that the test values except for the red screw connections.

Table 9: Loading parallel to strong axis Group Description Test Design models

5%tile (kN) Ductile (kN) Brittle (kN) LVL0-cn2 Common nail, one on each side 7.5 2.3 19.9 LVL0-sn2 Spiral nail, one on each side 5.4 1.7 19.9 LVL0-ls2 3.5" Lag screw, one on each side 8.2 6.1 19.9 LVL0-rs2 3.5" Red screw, one on each side 8.7 5.2 19.9 LVL0-ls8 3.5" Lag Screw, four on each side 35.2 24.2 39.8 LVL0-rs8 3.5" Red Screw, four on each side 31.8 20.6 39.8

LVL0-cn10 Common nail, ten on each side 32.7 23.2 99.6 LVL0-sn10 Spiral nail, ten on each side 27.3 17.0 89.6 LVL0-ls10 3.5" Lag screw, ten on each side 60.5 60.6 79.7 LVL0-rs10 3.5" Red screw, ten on each side 53.4 51.6 79.7 LVL0-5ls1 5.0" Lag screw, double shear 9.0 6.1 19.9 LVL0-5rs1 5.0" Red screw, double shear 7.6 5.2 19.9 LVL0-5ls4 5.0" Lag screw, double shear 29.3 24.2 39.8 LVL0-5rs4 5.0" Red screw, double shear 34.2 20.6 39.8


Table 10: Loading perpendicular to strong axis

Test Models

Group Description 5%tile (kN)Ductile

(kN) Brittle (kN)LVL90-cn2 Common nail, one on each side 8.9 1.8 10.6 LVL90-sn2 Spiral nail, one on each side 7.8 1.3 10.6 LVL90-ls2 3.5" Lag screw, one on each side 8.7 4.7 10.6 LVL90-rs2 3.5" Red screw, one on each side 10.6 4.0 10.6 LVL90-cn8 Common nail, four on each side 11.6 7.2 12.4 LVL90-sn8 Spiral nail, four on each side 10.9 5.3 12.4 LVL90-ls8 3.5" Lag screw, four on each side 27.1 18.9 12.4 LVL90-rs8 3.5" Red screw, four on each side 28.2 16.1 12.4 LVL90-ls1 5.0" Lag screw, double shear 11.5 4.7 10.6 LVL90-rs1 5.0" Red screw, double shear 10.9 4.0 10.6 LVL90-ls4 5.0" Lag screw, double shear 28.5 18.9 12.4 LVL90-rs4 5.0" Red screw, double shear 32.5 16.1 12.4

Table 11: Loading 30º to strong axis Test Model

Group Description 5%tile (kN) Ductile (kN) LVL30-ls2 3.5" Lag screw, one on each side 12.3 5.6 LVL30-rs2 3.5" Red screw, one on each side 13.9 4.8 LVL30-ls4 3.5" Lag screw, two on each side 13.0 11.2 LVL30-rs4 3.5" Red screw, two on each side 7.8 9.6

4.2 Connections with bolts In the UNB2 project, tests were carried out on high capacity bolted connections loaded parallel to strong axis of sawn lumber of spruce and pine, LVL, Parallel Strand Lumber (PSL) and LSL (Smith et al, 2006). The connection geometries followed the logic of Figure 11a but covered a range of combinations of bolt spacing, end distance, and edge distance. In that test program the failure modes were mostly brittle, with the brittle mechanisms being row-shear-out, group tear-out and net-section tension failures. Because governing mechanisms are variable, it is important to have reliable models for prediction of brittle and ductile capacities over the complete range of possible mechanisms. Dr. Pierre Quenneville has proposed a revision to the current Canadian design provision for bolted timber connection, including prediction of capacities for both brittle and ductile mechanisms (Quenneville, 2006 ; Quenneville et al, 2006). That proposal is based on study of connections in softwood glulam. As part of verification of his proposal, the UNB research team compared it (models in his proposal) to test results for EWP connections collected under the UNB2 project. Appendix B of this report contains the MS-PowerPoint presentation that summarises the comparison that was shown to the annual meeting of the Fastenings Subcommittee of the CSA086 Technical Committee, in November 2006. During this meeting, it


was agreed that future design provisions for bolted EWP connections will: (1) explicitly address the possibility of both ductile (EYM type mechanisms) and brittle mechanisms for all types of wood product member; and (2) recognize that the nature and range of brittle mechanisms to be considered differ depending on the types of ‘wood member’ (e.g. lumber, LVL). This reflected that, based on data from this and the UNB2 projects, the proposed ‘brittle’ mechanism design models are not accurate for connections in EWP. 5. Development of code provisions for wood connection design Working with various codes and standards committees was central to this project and there were a number of formal interactions with the CSA Technical Committee 086 and its subcommittees, the ISO Technical 165, and the CIB Working Commission 18. There have also been numerous informal contacts and discussions with members of those committees and other domestic and foreign experts on timber design. Appendices C to F contain documents and presentations related to the formal interactions with committees. Those draw on both this and the UNB2 projects, and on the long term expertise of the project leader and team members. The goal of all the interaction with code development committees and expert groups has been to overhaul what are clearly inadequate code provisions for design of connections in wood products. Doing that in a manner that accommodates introduction of modern EWP to the market place and the lexicon of structural engineers is also important. While attempting to do this it is important to recognize that each committee responsible for development or maintenance of a code at a domestic or international level is self governing in their operating methods and make independent decision on the basis of consensus. Therefore the appropriate approach is to provide committees with credible and soundly formulated data and suggestion, and thereafter to work with them iteratively until a satisfactory outcome is attained. The process is never swift. These are the considerations that defined the approach taken by the research team. Ideas discussed with code committees concerned: concepts important to overall structural systems and recognition of centrality of connection design to this; ideas related to derivation of design capacities by either mechanics based calculations or from tests data; and the format of resistance equations and calculation of resistance factors. Reaction to the ideas has been generally favourable so far. Project leader Professor Smith has been requested to lead efforts directed towards creation of an ISO standard for test-based determination of connection capacities. Appendix C is related to this. Professor Smith is also central to the ongoing revision of the national timber code CSA Standard 086-01 in respect of systems design principles and connection design. Appendix D deals with these aspects. Discussion on the similar topics was continued at the CIB-W18 Meeting with the hope that Canadian approaches will be the basis for code revisions in other countries. At the CIB-W18 meeting in 2006 (Florence, Italy) the UNB team presented papers on systems based design of structures, with emphasis on connection design issues and use of modern EWP (Appendix E); and on derivation of design capacities from test data (Appendix C). The UNB team has co-authored with Dr. Quenneville (chair of the CSA086 Fastenings Subcommittee) proposals for major revisions to connection design provisions in Canada. Those documents will be presented


and discussed at the November 2007 meeting of the CSA086 Technical Committee, with expected implementation by 2009. This one year project will install one, or perhaps a few, bricks into a wall of technical know-how (and supporting databases) that support work of code committees. The process of building and maintaining the wall is ongoing. Design codes constantly evolve and the project team members have every intention of remaining involved in the necessary actions. 6. Summary of outputs and recommendations Work centred mainly on validating and refining ideas developed during the UNB2 project, and interacting with the bodies that utilize the results of both the UNB2 and UNB75 projects. Tangible products of the activities include:

• Identification of the range of failure mechanisms for EWP connections using commonly adopted fastener like nails, screws and bolts.

• Establishing a comprehensive database on characteristics of connections in Canadian manufactured LVL.

• Discussion documents and recommendations for consideration by code and standards committees responsible for creation of new (or modification of old) design regulations, and definition of methods for test based assessment of connection systems involving use of proprietary products.

• Presentation of project results to bodies concerned with national and international regulation of design practices for constructions involving use of EWP.

Recommendations for future action include:

• Major revision of connection design practices applicable in Canada and elsewhere to encompass design practices applicable to use of modern EWP.

• Recognition by design codes of the central importance of thinking systemically about design of timber structures, and recognition of the systems approach in design rules applicable to connections and other components.

• Harmonization of practices for product assessments by agencies like the CCMC with approaches embodied in the national timber design code.

• Continuation of efforts to harmonize Canadian with international design standards and testing protocols.


7. References ASME (1996). Square and Hex Bolts and Screws. ANSI/ASME 18.2.1. ASME, New York. ASTM (2005). Standard Specification for Driven Fasteners: Nails, Spikes, and Staples, American Society for Testing and Material, West Conshohocken, PA. ASTM (1999). Standard Test Methods for Mechanical Fasteners in Wood, Standard D1761-95, American Society for Testing and Material, West Conshohocken, PA. ASTM (2000). Standard Test Methods for Bolted Connections in Wood and Wood Base Products, D5652-95, ASTM, West Conshohocken, PA. CCMC (2006). Registry Product Evaluations, Evaluation Report, 12719-R, Temlam® LVL, National Research Council, Institute for Research in Construction, Ottawa, Ontario. CCMC (2006b). Registry Product Evaluations, Evaluation Report, 12627-R, TimberStrand® LSL, National Research Council, Institute for Research in Construction, Ottawa, Ontario. CSA (2004). Engineering Design in Wood. Standard O86-01.2004, Canadian Standard Association, Rexdale, Ontario. Fasten Master (2006). Trusslok® - Engineered Wood Products, Product Specification, Agawam, MA. Madsen, B. (1998). Reliable Timber Connections, Progress in Structural Engineering and Materials, 1(3): 245-252. Quenneville, P (2006). Draft of code change proposal for wood connections, CSAO86-01, Meeting discussion document, November 17-18, Montreal. Quenneville, P., Smith, I., Asiz, A., Snow, M., and Chui, Y.H. (2006) Generalized Canadian Approach for Design of Connections with Dowel Fasteners, International Council For Research and Innovation in Building and Construction (CIB), Working Commission W18 – Timber Structures, Meeting 39, Paper No 39-7-6, Florence, Italy. Smith, I., Asiz, A., and Snow, M. (2006). Design Methods for Connections in Engineered Wood Structures, Final Report, Value-to-Wood Program, Natural Resources Canada, Ottawa. Smith, I., Asiz, A., Snow, M., and Chui, Y.H. (2006) Possible Canadian/ISO Approach to Deriving Design Values from Test Data, International Council For Research and Innovation in Building and Construction (CIB), Working Commission W18 – Timber Structures, Meeting 39, Paper No. 39-17-1, Florence, Italy.


Smith, I., and Chui, Y.H. (2006). Principles of Connection Design, International Standard Organization (ISO), Technical Committee 165 WG7, Meeting discussion document, Portland, Oregon. Smith, I., Landis, E. and Gong, M. (2003). Fracture and fatigue in wood. John Wiley and Sons, Chichester, UK. Snow, M. (2006). Fracture Development in Engineered Wood Product Bolted Connections, Doctoral Dissertation, University of New Brunswick, Fredericton, New Brunswick. Snow, M., Asiz, A., Chen, Z. and Chui, Y.H. (2006). North American Practices for Connections in Wood Construction, Progress in Structural Engineering and Materials, 8(2): 39-48. United Nations: Food and Agriculture Organization of the United Nations, Economic Commission for Europe (UNECE/FAO). 2003. “Forest Products Annual Market Analysis: 2002-2004”. Timber Bulletin - Volume LVI (2003), No. 3.


Appendices


Appendix A – Selected Load-slip Curves of LVL Connections for

Each Group Test


GROUP IDENTIFICATION

Connections loaded parallel to strong axis of LVL

No Group Description 1 LVL0-cn2 Common nail, one on each side 2 LVL0-sn2 Spiral nail, one on each side 3 LVL0-ls2 3.5" Lag screw, one on each side4 LVL0-rs2 3.5" Red screw, one on each side5 LVL0-ls8 3.5" Lag Screw, four on each side6 LVL0-rs8 3.5" Red Screw, four on each side7 LVL0-cn10 Common nail, ten on each side 8 LVL0-sn10 Spiral nail, ten on each side 9 LVL0-ls10 3.5" Lag screw, ten on each side

10 LVL0-rs10 3.5" Red screw, ten on each side 11 LVL0-5ls1 5.0" Lag screw, double shear 12 LVL0-5rs1 5.0" Red screw, double shear 13 LVL0-5ls4 5.0" Lag screw, double shear 14 LVL0-5rs4 5.0" Red screw, double shear

Connections loaded perpendicular to strong axis of LVL

No Group Description 15 LVL90-cn2 Common nail, one on each side 16 LVL90-sn2 Spiral nail, one on each side 17 LVL90-ls2 3.5" Lag screw, one on each side 18 LVL90-rs2 3.5" Red screw, one on each side 19 LVL90-cn8 Common nail, four on each side 20 LVL90-sn8 Spiral nail, four on each side 21 LVL90-ls8 3.5" Lag screw, four on each side22 LVL90-rs8 3.5" Red screw, four on each side23 LVL90-ls1 5.0" Lag screw, double shear 24 LVL90-rs1 5.0" Red screw, double shear 25 LVL90-ls4 5.0" Lag screw, double shear 26 LVL90-rs4 5.0" Red screw, double shear

Connection loaded off-axis at 30º to strong axis of LVL

No Group Description 27 LVL30-ls2 3.5" Lag screw, one on each side 28 LVL30-rs2 3.5" Red screw, one on each side 29 LVL30-ls4 3.5" Lag screw, two on each side 30 LVL30-rs4 3.5" Red screw, two on each side


LVL0-cn2

Slip (mm)

LVL0-sn2

Slip (mm)

Load

(kN

)Lo

ad (k

N)


LVL0-ls2

-20 -15-10-50 5

-12000

-10000

-8000

-6000

-4000

-2000

0

Slip (mm)

LVL0-rs2

Slip (mm)

Load

(kN

)Lo

ad (N

)


LVL0-ls8

-35 -30-25-20-15-10-5 0 5

-4.5

-4

-3.5

-3

-2.5

-2

-1.5

-1

-0.5

0

x 104

Slip (mm)

LVL0-rs8

Slip (mm)

Load

(kN

)Lo

ad (N

)


LVL0-cn10

0 5 10 15 20 25 30

-3

-2.5

-2

-1.5

-1

-0.5

0

x 10 4

Slip (mm)

LVL0-sn10

0 5 10 15 20 25 30

-3

-2.5

-2

-1.5

-1

-0.5

0

x 10 4

Slip (mm)

Load

(N)

Load

(N)


LVL0-ls10

0 5 10 15 20 25 30

-9

-8

-7

-6

-5

-4

-3

-2

-1

0

x 10 4

Slip (mm)

LVL0-rs10

0 2 4 6 8 10 12 14 16

-6

-5

-4

-3

-2

-1

0

x 10 4

Slip (mm)

Load

(N)

Load

(N)


LVL0-5ls1

-40 -35-30-25-20-15-10-5 0 5

-12000

-10000

-8000

-6000

-4000

-2000

0

2000

Slip (mm)

LVL0-5rs1

Slip (mm)

Load

(N)

Load

(kN

)


LVL0-5ls4

0 50 10 15 20 25 30 35 40

-4

-3.5

-3

-2.5

-2

-1.5

-1

-0.5

0

x 10 4

Slip (mm)

LVL0-5rs4

-30 -25-20-15-10-50 5

-4

-3.5

-3

-2.5

-2

-1.5

-1

-0.5

0

0.5

x 10 4

Slip (mm)

Load

(N)

Load

(N)


LVL90-cn2

-2 0 2 4 6 8 10 12 14

-14000

-12000

-10000

-8000

-6000

-4000

-2000

0

Slip (mm)

LVL90-sn2

-2 0 2 4 6 8 10 12 14 16 18

-12000

-10000

-8000

-6000

-4000

-2000

0

Slip (mm)

Load

(N)

Load

(N)


LVL90-ls2

Slip (mm)

LVL90-rs2

Slip (mm)

Load

(N)

Load

(N)


LVL90-cn8

Slip (mm)

LVL90-sn8

Slip (mm)

Load

(N)

Load

(N)


Slip (mm)

Slip (mm)

LVL90-rs8

LVL90-ls8Lo

ad (N

)

Load

(N)


LVL90-ls1

Slip (mm)

LVL90-rs1

Slip (mm)

Load

(N)

Load

(N)


Slip (mm)

Slip (mm)

Load

(N)

Load

(N)

LVL90-ls4

LVL90-rs4


LVL30-ls2

Slip (mm)

LVL30-rs2

Slip (mm)

Load

(N)

Load

(N)


Slip (mm)

Slip (mm)

Load

(N)

Load

(N)

LVL30-ls4

LVL30-rs4


Appendix B – EWP Bolted Connections


EWP Bolted ConnectionsPreliminary Assessment of Proposed

CSA086 Clause 10.2.4

Drs. Ian Smith, Andi Asiz, Monica SnowUNB Low-rise Construction Group

Objective

• To investigate applicability of the proposed ‘brittle’ connection failure clause to SCL connections.

• Initial focus on load parallel to SCL axis.


Scope• Connections loaded under tension parallel to the

strong axis.

• Double shear arrangement with steel side plates connecting two identical members.

• Members : LSL (44x146), LVL (44x146), pine (38x140), and spruce (38x140).

• 6 bolts, 12.7 mm diameter.

Brittle Connection Failure Mechanisms

PfPf Pf

Row shear (RS) Group tear-out (GT) Net Tension (NT)

(from proposed Figure 10.2.4)(Tension Parallel to the Strong Axis)


Factored Row Shear

RSr = φbrittle Σ 2 RSi (KD KSF KT)

φbrittle = 0.7, resistance factor for brittle failuresRSi = shear resistance along one shear plane of row “i”(N)

= 2.66 G0.85 Kls t nfi acrG = mean relative density of wood-based materials Kls = factor for member loaded surfaces

= 0.65 for side member= 1.00 for internal member

t = member thickness (mm)nfi = number of fasteners in row “i”acr = minimum of a3t (end distance) and a1 (spacing) (mm)

Factored Group Tear-out

GTr = φbrittle GT (KD KSF KT)

φbrittle = 0.7, resistance factor for brittle failuresGT = (RS1 + RSnr) + (27.3 G1.01 ?AGT-net)RS1 = shear resist along row 1 bounding the fastener group

= 2.66 G0.85 Kls t nfi acrRSnr = shear resist along row “nr” bounding the fastener group

= 2.66 G0.85 Kls t nfi acrAGT-net = critical area between the two outer rows (mm2)


Factored Net Tension

Tnr = φbrittle Tn (KD KSF KT)

φbrittle = 0.7, resistance factor for brittle failuresTn = ft AnAn = net area (mm2)ft = member specified tension strength

parallel-to-grain (MPa)

Ductile Failure Mechanism

Nr = φyield nyu ns nf

φyield = 0.8, resistance factor for ductile failuresnyu = lateral strength resistance for yielding (N)ns = number of shear planesnf = total number of fasteners in connection


Yielding - Two Shear Planes

2tdfn 22

a1yuθ=

11b1yu tdfn θ=

( ) ( )⎥⎥⎦

⎤

⎢⎢⎣

⎡β−

β+β+β+β

β+=

θ

θ211

y112yu tdf

M2412

)2(tdf

n

dfM212n 1y3yu θβ+β

=

where:t1 = side member thicknesst2 = main member thicknessβ = f2θ / f1θ

Mechanism

1a

1b

2

3

Test Method

Notes: - L =1200 mm; d =12.7 mm- varied a1, a2, and a3t- 10 samples for each group

L

d a3ta1

a2


Failure Mechanism - LSL

Net-tension(NT)

6.3190.350.876.276.2LSL-87.6204.2 50.850.876.2LSL-74.5207.8 50.876.250.8LSL-69.0195.9 50.850.850.8LSL-51.4212.7 63.576.276.2LSL-47.3204.8 63.550.876.2LSL-35.4211.6 63.576.250.8LSL-210.3208.1 63.550.850.8LSL-1

C.o.V (%)

Pu(kN)

a2(mm)

a1(mm)

a3t(mm)

Group

Test Result - LSL

Note: - moisture content 7 - 8%.


Key test observations - LSL:• Ultimate resistances only obtained after

considerable deformation.

• Failure mechanisms are complex, with many toughening mechanisms. LSL does not split like solid wood.

• None of a1, a2, and a3t have significant influence on failure mechanism or Pu (over the ranges considered).

• Variability in capacities was small (CoV ? 10%).

174.185.273.264.2136.5136.4LSL-8174.173.048.858.7136.5142.8LSL-7174.173.048.847.8136.5153.9LSL-6174.173.048.845.0136.5133.5LSL-5174.194.973.264.2136.5166.4LSL-4174.182.748.858.7136.5144.0LSL-3174.182.748.847.8136.5154.2LSL-2174.182.748.845.0136.5138.4LSL-1NTGTRS

EYMX JF-086

EYM(Mech. 1a)

5%tileTest

Group

Comparison: Test vs. Predictions – LSL(Unfactored Standard Term Capacities [kN])

Notes: G = 0.64 (C.o.V= 5%); embedment strength = 50.9 MPa (C.o.V=10%); spec. tensile strength = 32.8 MPa.

Proposed Clause: G = 0.64


Key comparison findings - LSL:

• Capacities are predicted well by EYM model (i.e. from only considering ductile mechanisms).

• Group geometry factor JF from current code does notapply.

• Proposed modifications to Clause 10.2.4 do not apply to situations investigated.

Failure Mechanisms - LVL

Row shear /Group-tear out

(RS & GT)


8.2111.950.876.2127LVL-813.492.050.850.8127LVL-715.696.350.876.288.9LVL-69.691.050.850.888.9LVL-58.6113.963.576.2127LVL-410.294.563.550.8127LVL-311.7102.363.576.288.9LVL-25.375.663.550.888.9LVL-1

C.o.V (%)

Pu(kN)

a2(mm)

a1(mm)

a3t(mm)

Group

Test Result - LVL

Note: moisture content 7 - 8%.

Key test observations - LVL:• Brittle failure modes dominated.

• Failure mechanisms were easily classified as RS or GT.

• All of a1, a2, and a3t could influence on the failure mechanism and Pu.

• Variability in capacities within arrangements ranged from low to moderately high (5% ? CoV ? 16%).


165.280.269.342.089.377.5LVL-8165.268.646.238.489.357.4LVL-7165.280.269.331.389.357.3LVL-6165.268.646.229.589.361.3LVL-5165.289.369.342.089.378.2LVL-4165.277.746.238.489.362.9LVL-3165.289.369.331.389.366.2LVL-2165.277.746.229.589.355.2LVL-1NTGTRS

EYMX JF-086

EYM(Mech. 1a)

5%tileTest

Group


Comparison: Test vs. Predictions – LVL(Unfactored Standard Term Capacities [kN])


Key comparison findings - LVL:• Group geometry factor JF from current code does not

apply.

• Proposed modifications to Clause 10.2.4 always predicts that RS will govern for situations investigated.

• Despite some inconsistencies, using G as a basis for calculating RS and GT capacities yielded reasonable prospective “code” capacities. Seem as close as has been reported before for bolts in glulam members.


Failure Mechanisms - Pine

Row shear /Group-tear out

(RS & GT)RS

RS

GT

GT

22.252.550.876.2127Pine-825.750.450.850.8127Pine-718.555.350.876.288.9Pine-612.845.550.850.888.9Pine-55.459.863.576.2127Pine-49.757.263.550.8127Pine-318.955.563.576.288.9Pine-218.859.063.550.888.9Pine-1

C.o.V (%)

Pu(kN)

a2(mm)

a1(mm)

a3t(mm)

Group

Test Result - Pine



Key test observations - Pine:• Brittle failure modes dominated.

• Failure mechanisms were often compound and not easily classified.

• All of a1, a2, and a3t could influence on the failure mechanism and Pu. However these were not very strong influences.

• Variability in capacities within arrangements ranged from low to high (5% ? CoV ? 26%).

47.845.140.732.068.026.6Pine- 847.838.327.129.568.023.3Pine- 747.845.140.724.068.030.8Pine- 647.838.327.122.168.028.7Pine- 547.850.140.732.068.043.6Pine- 447.843.227.129.568.038.4Pine- 347.850.140.724.068.030.6Pine- 247.843.227.122.168.032.6Pine-1NTGTRS

EYMX JF-086

EYM(Mech. 1a)

5%tileTest

Group

Comparison: Test vs. Predictions – Pine(Unfactored Standard Term Capacities [kN])




Key comparison findings - Pine:• Group geometry factor JF from current code is a

better representation for this case. Not really a big surprise as this situation is a closer situation to what that was based on.

• Proposed modifications to Clause 10.2.4 always predicts that RS will govern for situations investigated. The accuracy is spotty and the proposal is not always conservative. The current Code is closer overall.

• The long spacing in the row (a1 = 6d) cases yield the greatest inaccuracies in proposed code rules.

Failure Mechanism - Spruce

Row shear(RS)

1 row of 3 bolts(results not given)


13.478.563.576.2127Spr-215.772.663.550.888.9Spr-1

C.o.V (%)

Pu(kN)

a2(mm)

a1(mm)

a3t(mm)

Group

Test Result - Spruce


Key test observations - Spruce:

• Brittle failure modes only.

• Failure mechanisms were all classed as RS, but often were complex and strongly influenced by grain deviations.

• The test matrix was too limited to be conclusive about influences of a1, a2, and a3t. Pu was not dissimilar for the two cases studied.

• Variability in capacities within arrangements was moderate (CoV = 15%).


61.755.044.228.961.348.9Spr-261.747.729.520.061.343.0Spr-1NTGTRS

EYMX JF-086

EYM(Mech. 1a)

5%tileTest

Group

Notes: G=0.42 (C.o.V=5%); embedment strength =26.4 MPa (CoV=5%); spec. tensile strength= 14.2 MPa.

Comparison: Test vs. Predictions – Pine(Unfactored Standard Term Capacities [kN])


Key comparison findings - Spruce:• Group geometry factor JF from current code is not

appropriate.

• Proposed modifications to Clause 10.2.4 always predicts that RS will govern for situations investigated. The accuracy was good in one case and poor in the other.


Where This Leaves Uson the question of Proposed Clause 10.2.4

• This more proof that the existing code approach is flawed. It can be very inaccurate.

• Any new approach should recognize differences in material characteristics between various wood products. LSL and similar are a ‘class’ apart.

• For connections with more than one or two bolts brittle failures dominate. This is both a Connections Design and System Design Issue.

On-going Work at UNB

• Test of LVL connections using slender dowel type fasteners (traditional and modern).

• Development of test protocol for EWP connections.

• Development of advanced numerical and simplified design (code and applications) models.


Test on LVL Connections

• Stage-I: common and spiral nails, lag screws, self taping screws:– Parallel (comp. & tens.) and perpendicular (comp.)– Single and double shear – Number of fasteners: 1 to 10

• Stage-II: larger connections – > 10 bolts with large LVL members – Connections loaded off-axis.

Proposed Test Method(Perp. to axis)

LOAD

SIDE MEMBERS

MAIN MEMBERS


LOAD

SIDE MEMBERS

MAIN MEMBERS


LOAD

SUPPORT

SIDE MEMBERS

MAIN MEMBER

FASTENERS

LOAD

SUPPORT

SIDE MEMBERS

MAIN MEMBER

FASTENERS

Current ASTM

Possible new method


Advanced Numerical Model

LSL, single bolt

Solid wood, single bolt

Ex.

This work was undertaken withfinancial assistance from Natural Resources Canadan under the

Value-to-Wood prgramme

Thank you

This This workwork waswas undertakenundertaken withwithfinancialfinancial assistance assistance fromfrom Natural Natural ResourcesResources CanadanCanadan underunder the the

ValueValue--toto--Wood Wood prgrammeprgramme

ThankThank youyou


Appendix C - Possible Canadian / ISO Approach to Deriving Design

Values From Test Data


Possible Canadian / ISO Approach to Deriving Design Values From Test Data

Ian Smith, Andi Asiz, Monica Snow and Ying Hei Chui

Faculty of Forestry and Environmental Management, University of New Brunswick, Fredericton, NB, Canada

ABSTRACT This paper proposes principles for a standard practice for deriving factored design resistances of structural components or subsystems directly from test data. This applies to, for example, members, connections, trusses and shear-walls. The approach is intended as a fully consistent alternative to practices embodied in existing design codes. It is assumed that parent design standards (national loading and timber design codes) are based on Load and Resistance Factor Design (LRFD) concepts. The likely level of consistency between design solutions based on testing evidence and those based on existing LRFD codes is evaluated based on examples applicable to Canadian wood products. 1. Introduction The available range of proprietary products and subsystems, and special design situations involving structural use of wood products, has grown rapidly in recent years. Proprietary products are evaluated based on test data with the interpretations of evidence performed by product assessments organizations that are, in North America and elsewhere, independent of design code bodies. Assessment organizations produce documents containing suggested engineering design properties or equivalent information like limiting spans. The suggested properties only become validated within any regulatory jurisdiction if the appropriate building control authority accepts them. The purpose of this paper is to propose principles for a standardized approach that can be applied by product assessment organizations in Canada or elsewhere to derive engineering design values from test data. What is proposed is also thought suitable as guidance to engineers on how to make test based assessments of capacities of special components or subsystems. Here special means cases that do not lend themselves to ‘traditional design’ or are outside the scope of written design standards (e.g. use of components with non-standard dimensions). Ideas presented were initially conceived as a parallel alternative way of designing connections that fall outside the scope of the design code Canadian Standard 086-01 “Engineering design in wood” [1]. However, it became apparent that the principles could be applied generally to other components and subsystems, and that they may be a suitable basis for creation of an international standard under the International Standards Organization (ISO) system. Hence the broadened scope herein. Starting with connections was fortuitous in the sense that practices for establishing their design capacities have historically been inconsistent and background reasoning often


unknown. Thus there is no well framed or strongly conditioned thinking about what constitutes the best approach, as exists for example in the case of sawn lumber. This made it possible to start, so to speak, with an essentially blank piece of paper. This paper builds on past discussion in CIB-W18 Paper 38-102-1 ”New generation of timber design practices and code provisions linking system and connection design” by Asiz and Smith [2]. It also dovetails with the parallel 2006 meeting papers: “Generalised Canadian approach for design of connections with dowel fasteners” by Quenneville et al [3], and “Overview of a new approach to handling system effects in timber structures” by Smith et al [4]. The hope is that simultaneous discussion of these topics within CIB-W18, ISO/TC 165 Committee “Timber Engineering” and the CSA/TC 086 “Engineering Design in Wood” will facilitate progress and harmonization of practices. Focus below is, as a first step, on principles without attempt to specify all details or to create language suitable for a written standard. 2. Scope and Premises The scope is assessment of the behaviour of structural components and subsystems or special design situations through test evidence. Practices are intended to be adopted by product assessment agencies and be as far as is appropriate consistent with practices used to derive design properties specified in LRFD codes. It is recognized that test based product assessments will often be undertaken using quite limited amounts of test data, as compared to what underpins properties specified in design codes for some wood products (e.g. small dimension lumber). Practices described here are based on the following premises: Test based assessment will not be used as a means of circumventing applicable design codes. Parent design standards (national loading and timber design codes) are based on LRFD

concepts. Factored design resistances need to reflect target reliability indexes that relate to the nature of

the expected failure mechanism(s). There is no ambiguity regarding the structural function of any component or subsystem being

evaluated. Therefore all forces to be resisted can be defined. Interpolation is permissible in order to assign capacities to components or subsystems that

are similar to and intermediate between those assessed based on test evidence. The term similar implies that the wood products employed, material conditioning, boundary conditions, load components, loading regimes and workmanship are directly comparable. 3. Sampling and Number of Replicates 3.1 Sampling Specimens should be realistic and replicate expected field situations as closely as possible. Test materials and workmanship, and installation practices in the case of subsystems, must be representative of production situations. 3.2 Number of replicates


The number of test replicates should be sufficient for estimation of the average value and variance for each parameter used to characterize the structural response of any component or subsystem. The minimum acceptable number of replicates is six, and the Data Confidence Factor (Cf ) defined in Section 7 accounts for the possibility of unrepresentative sampling. It is intended that product assessment organizations or engineer be allowed to balance tradeoffs concerning relationships between the number of test replications, the level of variability in parameters that characterize the studied structural response, and the level of precision desired. 4. Test Practices and Loading Regimes Considerations to be taken into account will include, but may not be limited to: material conditioning prior to and after fabrication, long-term heating effects, boundary conditions, loading regime (e.g. time to failure, loading waveform and frequency), and workmanship.

An engineer should be responsible for design of the loading regimes best suited to procuring data from which to estimate properties defining the resistance of a component or subsystem to various applicable effects of loads. The necessary combination of loading regimes and other factors will vary depending on the type of component or subsystem and the intended field application. For components and subsystems manufactured using only traditional wood products (e.g. sawn lumber, plywood) assuming traditionally accepted methods for converting from the response under standardized static load tests of short duration to field situations can reasonably be considered reliable [e.g. 5]. Ditto if only well understood engineered wood products like Laminated-Veneer-Lumber (LVL) are used. In other circumstances it will be essential to establish data via a test matrix that properly addresses all of the issues itemised above, and possibly others. Professional judgement will apply. As a minimum, tests should be carried out using a regime under which monotonic load is increased steadily at a rate that causes failure in about 0.1 hours, i.e. static load tests. If only this minimum requirement is met then it must be assumed that any static or cyclic fatigue effects on strength will depend on the wood product components. In such circumstances the Service Factor ( kS ), as defined in Section 6.1, must take the most conservative possible values. For situations where a component or subsystem is intended to resist the effects of seismic or extreme wind forces it is advisable to incorporate cyclic loading tests into the testing schedule, based on ISO or similar methods. This will permit determination of a degraded load-deformation envelope as the basis for determining strength properties. For situations where sustained loads are expected to be of long duration it is highly advisable to conduct sustained load tests. A sufficient range of durations is one that permits establishment of a Load Level (LL) versus Time to Failure (Tf) relationship that will not be extrapolated by more than one order of magnitude based on a plot of LL versus log(Tf). Collected data should extend to at least three decades in Tf. 5. Interpretation of Test Data


5.1 General philosophical principle Because test data may be quite limited it is only realistic to expect to always accurately determine average design values for ultimate strength, or other properties that relate to assessment of a component or subsystem. The estimated variance associated with resistance properties may not be accurately characterized from available test data. Therefore it is proposed that average, rather than low exclusion level, characteristic test properties are the most appropriate base from which to derived design values. The determined Factored Design Resistance (φR ) according to equation (3), Section 6.1, can be made technically equivalent irrespective of whether an average or low exclusion level characteristic value is used. As has been demonstrated previously by Smith [6], randomness in low exclusion level characteristic values leads to noisy and inconsistent estimates of Resistance Factors (φ values) and Modification Factors (ki values). The result can be spuriously inconsistent design practices. The reason is easily demonstrated, in that, in reliability calibrations the process finds the value of φR that balances the equality in a prescribed design equation under certain conditions. Then the result is decomposed into φ , the Standardized Specified Resistance (RS) and ki values. Thus any uncertainty in variance of resistance introduces a first level of uncertainty into φR , and at a second level inconsistency into separated components of that product. RS is determined independently from analysis of test data, and that step causes second level noise in φ and ki values during any decomposition process. The second level noise is minimized and estimates of φ and ki are most robust if average characteristic resistance is used as the basis of establishing RS. As can be further elucidated, for systems where effects of design variables are coupled it is also most consistent to base calibration practices on RS that represents the average response. This is especially important when background data is from tests where the sample size is limited, which is itself the cause of noise in estimates of variance in resistance and therefore instability in low exclusion level estimates of resistance. There may be concern that the proposal to characterize resistance at the average level differs from past practice [7] where 5th-percentile resistance has been the basis for decomposingφR into component parts. The authors believe that such concern is unfounded because, as already indicated, the results can be technically equivalent. The primary difference in calibrations done as proposed here and those in past reliability studies is that what is proposed here results in the relationship:

φ

φR

R av

RR

av_ .

_ .

.

.

0 05

0 05

= ………… (1)

where: φR _ .0 05

= resistance factor based on R0.05, φR av_ .= resistance factor based on Rav., R0.05 =

estimated 5th-percentile resistance under standard conditions, and Rav = estimated average resistance under standard conditions. In Canada past approaches have introduced three additional manipulations into structural reliability calibration processes, as applied to design of isolated members [7]. First, the RS values do not relate directly to strengths observed under standardized test conditions, but are taken as


0.8 x short-term static strength. This adjustment has the purpose of converting observed strengths to those applicable under Standard Term Loads on roofs (dead plus snow loads) and floors (dead plus occupancy loads). This allows for expected damage due to static fatigue, which in timber engineering circles is better known as ‘the duration of loading effect on strength’. Second, there was smoothing across different sized products to allow for so-called size effects on apparent strengths of components, e.g. effect of member depth on apparent bending strength of sawn lumber. This was done for compact representation of data with design properties only being given for a reference member size, together with an adjustment rule for transforming properties so they apply to components of other sizes. This involved smoothing across reliability calculations in which component size was the variable of interest. Third, although nominally φ was consistently based on being used in conjunction with R0.05, it was in fact given a constant value, with RS rather than φ becoming the calibrated parameter. In North America this practice is termed Reliability Normalization of RS values. The result is that the relationship in equation (1) is replaced by:

φφ ψ

Code

R av

RR

RR

av

S

av

_ .

. .

..= =

0 8 0 05

………… (2)

where: φCode

= code specified resistance factor (typically 0.9), and Ψ = reliability normalization factor. The Ψ is a function of variance in the resistance and the target reliability index. Thus, neither φ nor RS have been determined directly from basic concepts. Despite what have been past practices, it is the view of the authors that for transparency it is important that RS values be properties that are directly reproducible from test data. Preferably they should be free from adjustments from test (short-term static strength) values to resistances applicable to standard term load conditions that are not reproducible in tests because the associated times to failure under sustained loads are not explicitly defined. It is important when determining design properties not to confuse the precision of calculation methods with correctness of output. For this reason although it is accepted that structural reliability concepts are the appropriate basis for determining φ , this does not necessarily imply that the reliability analysis methods need be complex and numerically based. What is critical, as with standardized test methods, is that processes be repeatable and consistent. Calibration of the ratio of the Factored Effects of Loads to the Factored Resistance (ratio of right to left side of the design equation - equation (3)) is the main mechanism for controlling safety levels. Choice of a suitable Safety Index (β ) is the most important decision, Section 7. 5.2 Classification of failure modes It is proposed that failure modes be classified according to the observed level of apparent ductility under static load conditions, determined based on the Ductility Ratio (D) calculated according to Figure 1. Table 1 details the proposed categorization of D. Because failure mechanisms and the extent of ductility may be mixed between test replications, it is suggested that when more than 25 percent of D values imply more brittleness than Dav. does, then the assigned classification should be ‘reduced’ by one category.


Figure 1 – Definition of characteristic deformations ∆U and ∆Y

Table 1 – Classification of failure mode Classification Average ductility ratio, Dav. Brittle D ≤ 2 Low-ductility 2 < D ≤ 4 Moderate-ductility

4 < D ≤ 6

High-ductility D > 6

NOTES: It is recognized that this system of classification may prove too challenging for practical implementation. An alternative is to adopt classifications such as ‘No, Low, Moderate and High Reserve Capacity’. Various options are being investigated.

5.3 Determination of yield and ultimate loads A range of options exist regarding the process for estimating the level of load at which damage and irreversible deformations start to accumulate. This level of load is commonly termed the Yield Load (RY) as for metals, even though wood products do not actually deform plastically. Subsystems may exhibit true plastic deformations if incorporated metal parts and fasteners are loaded beyond their yield point. Typically wood components and subsystems do not exhibit a clear transition from elastic to in-elastic response and unambiguous definition of RY can be impossible. In such cases the Offset Distance approach illustrated in Figure 1 is an acceptable method for estimating RY. Acceptable values for the offset distance will depend on the type of component or subsystem, but as a rough guide they can be expected to be in the order of 0.1mm if the measured deformation is a translation, and 0.002 radians if the deformation is a rotation. Many types of component and subsystem exhibit extensive post yield-point inelastic deformations, and in some cases apparent post-yielding hardening in the response. Under such circumstances the ultimate load is often not reached before termination of a test for reasons such as reaching the maximum deformation permitted by the test apparatus. In such circumstances an acceptable conservative method for estimating RU, and the associated deformation ∆U, is to substitute for those the maximum load applied and the associated deformation. Resulting classifications of failure mode according to Table 1 would result in conservative design. For some situations it may be desired to also estimate a so-called Failure Load (RF) from the post-peak part of a load deformation envelope curve, and the associated deformation ∆ F . 6. Form of LRFD Design Equations


6.1 Strength Limit States For strength controlled limit states the design equation is assumed to take the generic form:

≥Rφ effects of factored loads ………… (3) where: φ = resistance factor, R = resistance adjusted for all design specific considerations. The adjusted resistance is calculated according to the generic form: R = R k k k kS S T N H ………… (4) where: RS = standardized specified resistance, kS = service factor, kT = treatment factor, kN = number of components or subsystems in series factor, k H = system failure mechanism factor. The factor kS accounts for the coupled cumulative effects of loading and service conditions. Specifics of equations (3) and (4) will vary depending on sensitivities of resistances to design variables, and exactly what parameters need appear will vary from situation to situation. It is of course necessary to have prior applicable evidence, or to collect such evidence by testing as already indicated in Section 4. The companion paper by Smith et al [4] also discusses the concepts underpinning kN and k H . 6.2 Deformation Limit States Whether there is need to consider deformation limit states directly depends on the purpose for which any component or subsystem is intended, i.e. is end use dependent. The term deformation here is used in a very broad sense and includes the effects of all types of recoverable and non-recoverable movements. Although deformation limits are mostly applicable to systems and subsystems, there are situations where components have to be designed against the possibility of excessive deformability. The most likely applicable deformation limit states for components are associated with connections made by carpentry joints or using mechanical fasteners, or are deformations in linear elements that will be incorporated into plate or shell systems. As discussed in more detail by Smith et al in the context of systems [4], there are two reasons for such limits: Need to avoid basing design capacities on levels of resistances that are only attainable at

deformations that the parent system will not permit until stages well past initiation of global failure. This need is linked to the system failure mechanism factor ( k H ) in equation (4).

Need to avoid none collapse related, commonly referred to as Serviceability Limit States. For systems and subsystems deformation related limit states can fulfil a number of functions related to performance under normal use conditions. Examples include: avoidance of collision of adjacent buildings due to excessive inter-storey drift during strong wind or seismic events; avoidance of damage to the building envelope that leads to long term durability problems; and avoidance of annoying floor vibrations. Typically timber design codes will not be source documents for prescribing deformation limit state related criteria, even though some stray into that role [e.g. 1]. Therefore directives in any standard for test based assessment of design values should be restricted to guidance about collection of deformation data as ancillary information to the collection of strength information. The limit states and associated causal forces and excitations are best defined via building and loading codes, or if applicable from first principle.


7. Resistance Factors As is widely documented, society’s tolerance of building failures and associated risk to life is low, compared with risks for other life activities. The acceptable risk of death due to building failures has been estimated to be 0.14 x 10-6 per annum [8]. Consequently tolerated failure rates for components and subsystems of buildings are low. Assigned combinations of resistance and load factors within design equations are the tool used to control (in an informal sense) the likelihood of failures in components of systems. Even though structural reliability methods have been used as support to decisions concerning those factors, those methods primarily support relative scaling of φ values assigned for design of different types of component. Current knowledge of causal relationships is inadequate for true understanding of how choice of one value of φ versus another choice will map to failure rates for components, or to failure rates for subsystems and systems that incorporate the components. Nevertheless, reliability methods are valuable tools allied to and supporting code committee decisions. The Safety Index (β) that is incorporated into reliability based calibrations of design equations is the parameter that is most widely accepted as a compact indicator of relative security again failure that different selections of φ will yield. Failure here, and elsewhere, is taken to mean that the effects of applied loads have equaled the available resistance of any component, presuming that both loads and resistances can be sensibly assumed to be the result of random processes. Estimates of acceptable β have been made based on inverse reliability analysis that shows what β values typify design solutions judged to be acceptable under historical engineering approaches like Allowable Stress Design. Acceptable in this sense implies that there is a desirable balance between the security against failure and the amount of materials required. Values of β that have been judged appropriate fall within the range of 2.5 to 4.5 for components within timber and steel structures [e.g. 7, 9]. Here that is the range assumed appropriate for standardized practice for deriving design resistances directly from test data. For simplicity it is proposed that the standardized practices be restricted to determination of the average characteristic resistance against failure and the necessary companion value of φ , Section 6. Taking the simplest possible formulation, the resistance factor can be calculated as [9, 10]:

fCV )exp( αβφ −= ………… (5) where: α = calibration coefficient = 0.75, β = safety index applicable to the type of component or subsystem, V = coefficient of variation in resistance, and Cf = data confidence factor. The calibration coefficient is given the value 0.75 based on Smith [6]. Suitable choices for β are suggested within Table 2. There it is proposed that the safety index be a function of both the type of failure mode observed in tests, and whether or not the component or subsystem is to be part of a parent system that permits development of alternative load paths were a component or subsystem to fail [4]. The coefficient of variation in resistance should be calculated as:

V Vdata= +2 201. ………… (6)


where Vdata = coefficient of variation calculated from the test data. The term 012. is included to allow for general uncertainties in estimation of V. Data confidence factor C f is given by [10]:

If sample size n ≥ 10: C Vnf = −1 2 7. ………… (7)

If sample size n < 10: Cf = V

data

nR

R⎟⎠⎞

⎜⎝⎛

27,05.0

min ………… (8)

where: Rmin = minimum strength of n test values, and R0.05,data = raw data estimate of 5th percentile strength (which can be estimated by fitting a statistical distribution). Table 2 – Proposed values for Safety index (β )

Brittle Low-ductility Moderate-ductility

High-ductility

Classification of failure (Table 1)

β value No alternative load paths possible

4.5

4.0

3.5

3.0

Type of parent system

Alternative load paths possible

4.0

3.5

3.0

2.5

In the above the there is no adjustment included to change the basis of the reference condition for the Standardized Specified Resistance ( Rs ) from short-term loading, as is the common standardized test situation, to some other basis. Allowing the option for that adjustment and other possible deviation of test conditions from those for standardized resistance, the final proposed standardized approach for determining ‘factored standardized resistances’ directly from test data is:

DoLtestitest

fdataavS kk

CVRR

,,

, 1)75.0exp(

∏−

=β

φ ………… (9)

where: Rav data, = average strength of n test values, ktest i,∏ = product of modification factors adjusting from the Standard Conditions associated with RS to test conditions, and ktest DoL, = modification factors adjusting from the Standard Term Loading Duration associated with RS to the test duration. Based on past practices, for Canada ktest DoL, = 1.25 [7]. As will be noted, there is no actual need to separate φ from RS. If that step is included to yield information in a more familiar LRFD format, it can be achieved using equation (5) as the basis for decomposing equation (9). 8. Expected Level of Consistence with Current Practice Examples are given here to illustrate the likely level of consistency between design solutions


based on the proposed approach for determining design resistances directly from test data and existing properties in the current edition of the Canadian timber design code [1]. Table 3 summarizes Standardized Factored Resistance ( SRφ values) for data representing different levels of variability in resistance (apparent bending strength of relatively small dimension sawn lumber loaded as joists in a dry condition). In calculations for the proposed method β is taken equal to 3.5. This corresponds to a situation where the response exhibits low ductility (alternatively thought of as a component with Low Reserve Capacity) and members are employed in a parent system that can develop an alternative load path(s). That is the most typical application of such materials. All tabulated values are applicable to Standard Term Loading as defined in Canada ( ktest DoL, = 1.25). The observed member strength data is taken from the literature [7] and is the same ‘in-grade’ data on which the design properties in the Canadian code are based. Factor Cf in calculations to the proposed method are based on n = 400 [7]. Table 3 – Comparison of Standardized Factored Resistances ( SRφ values) – Apparent bending strength

Product definition Test data – bending strength (MPa)

Standardized Factored Resistance (MPa)

Species

Grade

Nominal size

(mm)

Average

CoV

5 %tile

Canadian timber design

code [1]

Proposed Approach

38 x 241 40.44 0.27 21.60 16.3 14.6 38 x 191 47.91 0.28 25.47 17.8 16.9

Select Struct.

38 x 89 56.02 0.21 35.45 25.2 23.4 38 x 241 27.55 0.43 9.55 4.55 3.86 38 x 191 20.99 0.32 9.96 4.97 6.65

Douglas Fir - Larch

No. 3

38 x 89 46.19 0.46 14.47 7.04 10.1 38 x 241 37.07 0.24 21.72 15.8 14.5 38 x 191 50.41 0.29 25.69 17.3 17.3

Select Struct.

38 x 89 67.88 0.28 35.83 24.5 23.9 38 x 241 27.11 0.36 11.49 6.93 7.80 38 x 191 27.30 0.32 12.86 7.56 8.65

Hem-Fir

No. 3

38 x 89 43.15 0.35 18.64 10.7 12.6 38 x 241 35.34 0.23 21.08 16.3 14.1 38 x 191 39.87 0.25 23.01 17.8 15.2

Select Struct.

38 x 89 52.46 0.21 33.44 25.2 22.1 38 x 241 16.50 0.26 9.11 6.93 6.11 38 x 191 24.29 0.30 12.08 7.56 8.11

Spruce-Pine-Fir

No. 3

38 x 89 32.13 0.31 15.64 10.7 10.5 Based on tabulated comparisons, there is in general quite good consistency between design properties determined by the proposed approach and properties in the design code. Some discrepancies exist, but that has to be the case because the proposed approach is not a copy of the other one. It is deduced that designs proven based on test evidence according to the present proposal would yield acceptable levels of safety. The essence of this deduction would not change were more complex approaches to the reliability analysis to be inserted. An argument against use


of simple reliability based approaches, like equation (5), could be that they do not accurately account for interactions between load and resistance sides of the design relationship. However, the authors believe that it is important not to overcomplicate processes that involve engineering judgment, and that calculations be treated as an aid to informed judgment. Unlike numerical methods, simple approaches do not obscure the relationship between calculated standardized factored resistances and the raw data on which they are based. Employing a kernel equation like equation (9), which is a single step process, maximizes the transparency of interrelationships. 9. Conclusions The approach proposed here for determining design properties for structural components and subsystems directly from test data is very simple, yet it appears to provide good consistency with provisions of existing code rules. Although created within the context of design practices in Canada it is believed that what is suggested is also the suitable basis for creating an ISO standard for data evaluation. A key concept embedded within the proposed approach is that the reliability index used within any analysis of test data be related to classification of the failure mode and the end use situation. Classification of failure modes would be according to a brittle – ductile scale. Definition of the intended end use would distinguish between whether a parent structural system incorporating components or subsystems of the type investigated is of a type capable to developing an alternative load path were any component or subsystem to fail. This dovetails with suggestions in a companion CIB-W18 meeting paper on Systems Level Design. Acknowledgement The authors gratefully acknowledge financial support from Natural Resources Canada under the project ’UNB2 - Design Methods for Connections in Engineered Wood Structures’ (2003-06). References 1. Canadian Standards Association (CSA). 2005. “Engineering Design in Wood”, CAN/CSA

Standard 086-01, CSA, Toronto, ON, Canada. 2. Asiz, A. and Smith, I. 2005. “New generation of timber design practices and code provisions

linking system and connection design”. CIB-W18 Meeting in Karlsruhe, Paper 38-102-1. 3. Quenneville, P., Smith, I., Asiz, A., Snow, M. and Chui, YH. 2006. “Generalised Canadian

approach for design of connections with dowel fasteners”. CIB-W18 Meeting in Florence (in press).

4. Smith, I., Chui, Y.H., and Quenneville, P. 2006. “Overview of a new approach to handling system effects in timber structures”. CIB-W18 Meeting in Florence (in press).

5. Forest Products Laboratory. 1999. “Wood handbook--Wood as an engineering material”. General Technical Report FPL-GTR-113. U.S. Department of Agriculture, Forest Service, Forest Products Laboratory, Madison, WI, USA.

6. Smith, I. 1985. “Methods of calibrating design factors in partial coefficients limit states design codes for structural timberwork: with special reference to mechanical timber joints”. Research Report 1/85, Timber Research and Development Association, High Wycombe, Bucks, UK.


7. Foschi, R.O., Folz, B.R. and Yao, F.Z. 1989. “Reliability-based design of wood structures”. Structural Research Series Report No. 34, Department of Civil Engineering, University of British Columbia, Vancouver, BC, Canada.

8. Reid, S.G. 2000 “Acceptable risk criteria”. Progress in Structural Engineering and Materials, 2(2): 254-262.

9. Ravinda M.K. and Galambos T.V. 1978. “Load and resistance factor design for steel”, ASCE Journal of the Structural Division, 104(ST9): 1337-1353.

10. Leicester R.H. 1986. “Confidence in estimates of characteristic values”, CIB-W18 Meeting in Florence, Paper 19-6-2.


Appendix D - Principles of Connection Design


DISCUSSION PAPER

ISO TC165 Working Group 7 “Joints made with mechanical fasteners” ISO TC 165 Meeting Portland, Oregon August 1-4, 2006

Principles of Connection Design

Ian Smith and Ying Hei Chui

University of New Brunswick, Fredericton, NB, Canada

ABSTRACT This document is produced in response to ISO TC165 resolution 260 (E) and discusses a proposed radical overhaul of Canadian design provisions related to connections. Changes are expected to encompass both the underpinning principles and detailed design rules, and be consistent with parallel changes intended to facilitate introduction of a systems base approach to selection and sizing components of all types. The primary approach for assigning design capacities will become mechanics based models that explicitly address each potential joint or connection failure mechanism. A secondary permitted approach will be to derive connection capacities directly from test data. Resistance factors will be estimated based on reliability concepts that recognise factors specific to connections, rather than mimic practices previously developed for design of isolated wood product members. Additionally consideration is being given to the possibility of allowing designers to choose between two levels of methodology in design of connections and parent structural systems. Level 1 design methodology is envisioned as predicated on elastic analysis of systems, being relatively uncomplicated, and being presented in a format very similar to existing Limit States Design (LSD) code provisions in Canada. By contrast Level 2 methodology is envisioned as permitting options of traditional type LSD or Capacity Based Design (CBD), and explicitly linking design of connections to the characteristics of parent structural systems. Although more complex, the upper level approach would lead to more efficient design solutions and enhanced structural performance. If CBD were implemented it could be based on either elastic or plastic system analysis. Any solution based on plastic analysis would require selection of connection hardware capable of achieving a ‘high-ductility’ rating. This paper outlines the current state of thinking on how various pertinent factors can be addressed. Its scope is restricted to systems fabricated with mechanical inter member connections.


1. Introduction This document is produced in response to ISO TC165 resolution 260 (E) and discusses a proposed radical overhaul of Canadian design provisions related to connections. It is believed that the issues being confronted are the same as those pertaining in other countries represented within ISO TC165 and that what is discussed can become the basis of harmonized international design practices. All structural design is a technical art, and is not the mere application of rules and formulae. Timber design codes provide designers with information about how components manufactured from wood products are to be sized, and how to size connections joining them together. Those documents do not provide any guidance on how to link material utilization with decisions about structural form. Therefore all primary design decisions have been made, and often system structural analysis performed, before the timber design codes need to be consulted. Based on a recent survey of North American structural engineers [1], typical practice is to size members of timber systems before selecting the connection methods. Many structural engineers report that they do not size the connections at all and instead simply add notes to the design drawing that specify that the connections are to be designed by the system fabricator. Obviously connection design is often assigned limited importance. On the contrary, for timber structures consideration of the connections should be assigned paramount importance, and should enter at the beginning of the design decision tree. Current timber design codes provide rules for design of isolated components (e.g. members, shear-walls, joints) based on their ‘free body’ diagram, together with limited guidance on control of system deformation. Free body component forces are determined based on applicable factored design loads, determined in the case of components from statically indeterminate systems based on assuming an elastic system response. It depends upon the skill of the individual designer whether or not compatibility between deformations in the parent system and deformation of components is considered. Not all designers will consider whether the global system would permit a component to achieve the deformation necessary to mobilize its assumed resistance under overload conditions, i.e. components in composite subsystems or statically indeterminate global systems must satisfy deformation compatibilities as well as resist forces as a free-body. Failure to consider deformation compatibility will result in poor design solutions. Amongst the various classes of components in timber structures, connections are undoubtedly frequently poorly designed. Because of the array of variable design parameters connection design is inherently the most complex matter that any timber design code can address. Radical overhaul of the Canadian Standard CAN/CSA 086-01 “Engineering design in wood” [2] is intended in respect of system, subsystem and connection design provisions, with the expectation that changes will be implemented by 2009. The goal is a code that fosters selection of correct combinations of structural form, subsystems and components, with the intention that design of connections not be relegated to a matter of afterthought. Also it is the aim to create provisions that promote timely introduction of new products (e.g. Structural Composite Lumber, connection hardware) into the range of options according to methods that yield consistency in design of systems using traditional and new products. This paper discusses the nature and rationale for possible Canadian code changes related to the design of mechanical connections. A


companion paper produced for the 2006 CIB-W18 meeting in Florence in August 28-31 addresses the broader issues [3]. 2. Design Methodologies and Choices 2.1 Scope and Basis of Design Capacities So far, for connections it is agreed in principle for Canada that mechanics based analysis methods will be the primary approach for assigning capacities to connections within a Limit States Design (LSD) framework. Currently the Canadian timber design code adopts such a framework applied with Load and Resistance Factor Design (LRFD) nomenclature. With a few specific exceptions the source basis of design capacities for connections is empirically interpreted test data, with embedded inconsistencies between connections made using various types of fasteners. Test data collected in Canada specifically for the purpose of updating CAN/CSA 086-01 will be the preferred reference point for calibrating mechanics based analysis methods (see Acknowledgements). Account will also be taken of archival North American and foreign data, to fill gaps. A key aspect of data collected by the authors, co-workers and Canadian colleagues is that it addresses use of both traditional solid wood products like sawn lumber and glued-laminated-timber (glulam), and modern Engineered Wood Products. Some EWP have very different characteristics from solid wood. The initial scope will be connections involving members manufactured from softwood sawn lumber and glulam, Structural Composite Lumber manufactured from softwood and hardwood species, structural wood panels (plywood and OSB), and light gauge or plate steel. Fastener types to be included will be nails, timber rivets, screws (wood, self-tapping and lag), plain dowels (also known as drift pins), bolts and tubes. There exist in the present code methods for assigning design capacities for certain types of proprietary connection hardware (ditto proprietary wood products used as linear members). So far this only covers connections made using punched metal plate fasteners, and joist hanger connections. The methods differ for the two types of hardware, and no actual design capacities are given in the code for such products. Engineers can calculate design capacities based on characteristic properties supplied by the hardware manufacturers and verified by third parties. Presently, there is no consistency in the methodologies adopted by third party agencies to derive characteristic properties for proprietary fastenings, although it is claimed that these procedures meet the intent of the Canadian timber design code [2]. Because there are in the marketplace a rapidly expanding range of products, it is now proposed to use a consistent approach of determining characteristic properties and therefore design capacities of connections based on third party verified test data. This will create a uniform basis across different combinations of wood product and connection hardware. Proposed practices are outlined in a companion paper [4]. Since design codes typically shy away from providing design properties for proprietary products, the expectation is that the new approach will guide staff of Product Assessment Organizations (like the Canadian Construction Materials Centre) who evaluate products. 2.2 Allowing Capacity Based Design Two conceptual approaches are practically applicable to normal design of structural systems and


components (members and connections), irrespective of the types of materials used: 1. Ensuring that each component and subsystem is strong enough to resist the estimated effect

of applied design forces without sustaining irreversible damage. Here this is termed Strong Component Design (SCD) and essentially it corresponds to the approach that all existing timber design codes employ, assuming an elastic system response. The qualification “essentially” is added because in the case of design of joints many timber design codes permit the estimation of the resistive capabilities based on post-elastic response [5]. For statically determinate systems this inconsistency does not lead to erroneous force analyses and therefore no error occurs in sizing of components. However, for statically indeterminate systems the inconsistency will result in erroneous force analyses, with possibly serious inaccuracies in sizing of components. To draw a parallel, it is like designing structural steel systems by mixing estimates of elastic internal forces with plastic resistances of components. That is of course a universally prohibited practice. It is doubtful that structural engineers widely appreciate the inconsistency exists in structural timber design.

2. Controlling how systems will fail by applying what is referred to as the Capacity Based

Design (CBD). The concept is that some elements be located in systems to channel the flow of forces, with those control elements designed so that they and not other components will be the weak elements. If the control elements are designed to fail with high-ductility, non-brittle system failure can be enforced. CBD was originally developed for seismic design of reinforced concrete structures [6]. However the underlying principles can be applied to all statically determinate and indeterminate systems where assumption of pseudo static loads is reasonable. For indeterminate systems the internal forces can be determined by elastic or plastic analysis methods. For timber structures mechanical connections are the only type of elements that can be made to fail predictably with high-ductility. Therefore the essential caveat with CBD of timber systems is that only certain types of connections (i.e. those with predictable high-ductility) be permitted as the control elements. Also, if design is based on plastic analysis, an additional caveat must be that there be enough such control connections to create a plastic system collapse mechanism. Previously Chui and Smith [7] have studied the possibility of applying CBD to timber structures and found that it is feasible. The biggest challenge undoubtedly lies in identifying the combination of wood based products (members), mechanical fasteners and reinforcing strategies capable of yielding highly ductile responses. However, as recent research illustrates, it is possible using properly selected combinations of existing products to achieve the objective without need for any member reinforcement [8, 9]. Undoubtedly allowing CBD would foster development of new efficient connection hardware.


Figure 1 – Logic of Capacity Based Design: based on elements in series

(where: DU = factored effect of loads, LCLC R ,,φ = lower factored connection resistance,

UCUC R ,,φ = upper factored connection resistance, MM Rφ = factored member resistance) As illustrated in Figure 1, code implementation of CBD requires that both low and high exclusion level estimates of connection capacities be made. That necessitates reliable estimation of the variance in strength distributions, as opposed to the traditional attempt to only characterise the lower tail of the distribution. For design solutions to be efficient the variance in connection resistances must be small, otherwise member resistances will be grossly underutilized. Clearly predictability and control of variance in connection capacities means using hardware for which metal fasteners or other metal parts will be the control elements. It is interesting to note that accomplishing true consistency between the design of systems and connections could only be attained with CBD. That is unless connection design capacities are restricted to within the elastic response range, but that is not viable as connection capacities would be unreasonably low and also failures would always initiate in members that are quite brittle. The fast fracture phenomenon would become an issue and any system collapses would be more likely to be disastrous, see Section 6.4 [3]. Because of the need for seamless transition from old to new design methodologies, it would be impractical to simply change from SCD to CBD whatever the technical arguments are. So far no decision has been made in Canada regarding whether both the existing (SCD) approach and CBD will actually be permitted. Therefore issues associated with both options are examined within the remainder of this discourse. 2.3 Reliability Concepts The intent is that factored design resistances will be estimated based on reliability concepts. Details of how have not been finalised but the methods will not simply be an attempt to replicate past practices applied to isolated wood product members. The reliability methods must be those appropriate to connection design. They should reflect general considerations outlined in the companion paper by Smith et al [4], plus special considerations applicable to connections. Even for restricted situations like design of connections employing dowel like fasteners there is an enormous possible range of acceptable design solutions. Also except for the rare case of a single dowel fastener connection there is always ‘system effects’ present. Geometric and service condition related variables strongly affect how connections behave and the influences can be strongly coupled. Thus, the variance in resistances of nominally identical connections cannot be precisely defined. Archival of test data is usually limited in scope and expert judgement has to be employed by both design code committees and design practitioners. A premium needs to be put on avoiding, or at least minimizing, artificial instability and incoherence in estimates of characteristic design properties and associated resistance factors, or modification factors accounting for service conditions that differ from reference conditions associated with base specified resistances. Such instability and incoherence exists in the present connection design equations employed in Canada, and elsewhere, and mainly results from the desire to implement Factored Design Resistances (R values) that are forced to fit a design


relationship wherein they are the products of sequential multiplication adjustments to so-called Standardized Specified Resistances (RS values). Nominally RS values are 5th percentile resistances. Failure to recognize various factors having coupled interactions, i.e. where significant cross-products in effects of the variables exists, amplifies noise in any underlying test data [10]. Design equations and their calibration need to embrace rather than ignore coupled influences (see Section 6.2). Reliability methods applicable to connections should recognise that the best possible available information concerns average response under any design situation. It cannot be supposed that relatively massive data sets will exist, as for example is often the situation for small dimension sawn softwood lumber products, again because the number of variables is too vast. Estimates of variance in resistances of nominally identical connections are typically only hand-waving estimates. Data and expert judgement are needed in order to arrive at an acceptable estimate (see Section 6.2). Complex calculations (statistical and reliability analysis) are helpful aids, but only when judgement is capable of sorting the valuable from the misleading results. 2.4 Dual Level Design Methodology A major issue still under debate in Canada concerns the possibility of permitting two levels of connection design methodology. If a dual level approach is decided on, Level 1 methodology would be uncomplicated and envisaged for use by, for example, engineers without specialist knowledge of structural timber design. By contrast Level 2 methodology would be relatively complex and suitable for use by, for example, engineers with specialized knowledge of timber design. Consultations with design practitioners [1] has revealed that most engineers in North America only occasionally design timber structures, with usually only simple components or subsystems involved. Therefore a Level 1 alternative should be robust and simple to comprehend. Solutions would tend toward conservatism. Level 2 designs would involve explicit account of interrelationships between the behaviour of complete structural systems and selection of connection methods. Its advantage would be that it allows experts to utilize material efficiently, and facilitates the incorporation of innovative products or systems in timber structures. 3. Decisions on Terminology and Implied Consequences In the literature the terms connection and joint have been used synonymously by many authors, and timber design codes only directly address design of joints [e.g. 2]. Here a connection is taken to mean a complete assembly for linking together two or more structural components. The term joint is taken to mean the arrangement needed to attach one end or one surface of a single structural component to the rest of structural system to which it belongs. Joint and connection can be synonymous if the latter contains only one joint. Design of joints involves identification of all possible local failure mechanisms, selection of fasteners or other necessary hardware (type, size and number), assessment of the joined materials against adverse effect of stresses caused by attachment to fastening hardware, and if appropriate avoidance of excessive deformation across the joint plane. As an illustration, Figure 2 shows some mechanisms that occur in simple bolted tension joints. In the photographs wood members


are being viewed during tests through a transparent plastic plate (Lexam®) that replaced a normal steel splice plate in a three member joint with a central wood member [9].

Figure 2 – Example of possible failure joint mechanisms: bolted tension joint Design of connections involves aggregation of effects of mechanisms for joints, and identification of possible global failure mechanisms associated with the overall connection assembly. As an illustration, Figure 3 shows some possible global failure mechanisms for connections made with dowel fasteners. Failure mechanisms individually or in combination define the failure mode of a connection.

Figure 3 – Example of possible global failure

mechanisms: dowel fastener connections Figure 4 – Definition of ductility ratio

Failure mode is the term applied to the gross nature of the failure in a complete connection. It is proposed that the mode be defined on four class rating scale that encompasses brittle and ductile behaviours. The mechanism or composite effect of mechanisms with the lowest capacity will control the nature of the failure mode. Table 1 gives a suggested rating scale for failure modes. To avoid subjectivity each classification is related to the quantitative parameter Ductility Ratio (D). The table also lists suggested associated values of the Reliability Index (β) which is the parameter that implements relative scaling of acceptable failure probabilities for the different types of mode. Ductility ratio is determined as illustrated in Figure 4 and proposed practices are consistent with those in companion meeting papers [3, 4]. It is recommended that only connection types capable of achieving the high-ductility classification with minimal degradation be permitted for use in structures designed according to CBD.


It is recognized that the widths of D bands within the proposed system for classification of failure modes could be too narrow. It is also possible that a classification system on brittle-ductile scale might be too difficult to implement in practice. One alternative is to adopt ratings of No, Low, Medium and High Reserve Capacity. Various options are being investigated. Table 1 – Classification of failure modes based on D and associated β values

Brittle Low-ductility

Moderate-ductility

High-ductility Classification of failure mode D ≤ 2 2 < D ≤ 4 4 < D ≤ 6 D > 6 Type of parent system (Table 2)

β value

No alternative load paths possible (light-frame, heavy frame, other)

4.5

4.0

3.5

3.0

Alternative load paths possible (light-frame, heavy frame, shell, plate)

4.0

3.5

3.0

2.5

4. Sequencing of Design Decisions

Figure 5 – Sequence of decision and interrelationship of system and connection design

Earlier discussion in this document refers to the design of connections, because that is the target outcome. However, to date design code rules have been formulated to address design of joints. Consequently how and how well broader issues of connection design are handled is completely a


function of the extent of a design engineer’s personal knowledge. As has been found in a recent survey of North American structural engineers, their background knowledge of timber is often shallow [1]. Therefore the authors believe that design codes need to be formulated in a format that aids correct decision making, and with practice assist designers to build up their expertise in structural timber design. The philosophical approach that underpins the remainder of this discourse is that both Level 1 and Level 2 design methodologies should contain specific guidance on the selection of compatible combinations of structural form, wood products and types of connection hardware. Additionally, Level 2 design methodology should result in design capacities of joints and connections that are conditional on both their nature and to what type of parent system they belong. Level 2 would in essence reward designers for correct decision making. Figure 5 illustrates the structuring of decisions that it is proposed be embedded in design codes irrespective of whether Level 1 or Level 2 methodologies are followed. Decisions about types of connections would be integral to both general design clauses and those dealing specifically with connection design, and part of revisions necessary to implement system design [3]. Tables 2 to 4 illustrate preliminary thinking on how decisions in steps 2 to 5 (in Figure 5) should be based. There are only two classes of materials proposed for wood based products in Table 3, because ability to achieve the more refined four levels classifications scheme of Table 1 for connections is most usually a function of the metal hardware (e.g. metal fasteners, metal parts). The notion that member materials can be divided into splitting and no-splitting categories reflects direct experience in tests of connections made with a range of sawn lumber and EWP members [8, 9]. In reality the classification of failure mode of joints is not as simple as dividing the jointing materials into splitting and non-splitting type. Researchers have long recognized that, in addition to jointing material, joint failure mode depends on fastening details (e.g. slender ratio of fastener, end fixity and ‘hardness’ of fastener) and loading characteristics (e.g. static vs dynamic). Our ability to predict ductile failure is well proven as numerous research studies have verified the reliability of the so-called European Yield Model (EYM) to predict failure load and mode of failure of mechanical fastened joints, provided brittle fracture of material is suppressed. Even so this is only true for behaviour under static loading conditions, and there has been evidence to suggest that the same failure modes do not necessarily apply under dynamic loading conditions. In contrast predictive models for brittle failure of joints are still in their infancy, and much research is still required to further develop some of the concepts currently being proposed [add Pierre’s paper as reference??] Having predictive tools for all possible modes of failure is a pre-requisite for implementing the approach discussed in this document for design of timber connections with dowel like fasteners. Table 2 – Classification of parent structural systems

Classification Description Light frame without capability to develop alternative load paths

Systems where less than four linked and parallel framing members or subsystems act together to resist the effects of loads.

Light frame with capability to develop alternative load paths

Systems where four or more linked and parallel framing members or subsystems act together to resist the effects of loads.


Heavy frame without capability to develop alternative load paths

Statically determinate systems, or other systems, where failure of a single component would cause disproportional damage to the whole system.

Heavy frame with capability to develop alternative load paths

Statically indeterminate systems where failure of a single component would not cause disproportional damage to the whole system.

Shell and plate structures Those structures depending on two-way curvature and folded plate action for stability.

Other Any system not within another classification. Table 3 – Classification and expected behaviours of structural wood products

Classification Expected behaviour Examples Splitting Brittle response sawn lumber, un-reinforced glulam, LVL, PSL Non-splitting Ductile response reinforced-glulam, plywood, OSB, LSL

The logic that underpins Table 4 is that in recent studies it has been found that a correlation often exists between the type of failure mechanism and the ductility ratio [8, 11]. Thus when mechanism capacities are estimated using mechanics based model analysis the nature of the mechanism can be employed as a surrogate in lieu of direct evidence concerning the magnitude of D. Obviously, when connection capacities are determined directly from test data, classification of the response on the basis of Table 1 would be straightforward [4].

Table 4 – Examples of classification of failure modes for connections made with dowel like fasteners: capacities estimated using mechanics based models

Member Classification (Table 3) (for least ductile joined material)

Connection Classification

(Table 1) Splitting Non-splitting

Governing mechanism (Figures 2 and 3) (Examples only)

Brittle X row tear out, block shear, net section, member interference

X bearing failure beneath fasteners with only rigid body movements of fasteners

Low-ductility

X X all connection level mechanisms Moderate-ductility

X plastic hinges formed in fasteners

High-ductility X all joint mechanisms 5. Limit States Design Equations Only the LSD formulas applicable under SCD are given here. The parallel forms for CBD can be deduced from Figure 1. Whether serviceability should be explicitly addressed by LSD codes is a contentious issue. In the context of connection design, serviceability will equate to occurrence of excessive deformation. The authors favour the approach that codes give general guidance on estimation of likely levels of connection deformation under various service situations, with application of that information being a matter of system rather than component design.


The SCD design strength checking equation is assumed to take the generic form:

≥Rφ effects of factored loads ………… (1) where: φ = resistance factor, R = resistance adjusted for all design specific considerations. The adjusted resistance is calculated according to the generic form: R = R k k k ks S T N H ………… (2) where: Rs = standardized specified resistance, kS = service factor, kT = treatment factor, kN = number of connections in a system factor, k H = system failure mechanism factor. The factor kS accounts for the coupled cumulative effects of loading and service conditions. Specifics of equations (1) and (2) will vary depending on sensitivities of resistances to design variables, and exactly what parameters need appear will vary from situation to situation. It is of course necessary to have prior applicable evidence, or to collect such evidence by testing as already indicated in earlier sections of this paper. The companion paper by Smith et al [3] discusses the concepts underpinning kN and k H . As discussed in Section 8, it is suggested that kN and k H need not appear within a Level 1 design methodology. 6. Other Key Issues Within the relatively constrained discussion possible in these pages many details are skirted and in some instances ignored. Despite what this may be taken to imply, in fact much work has already been done and many detailed discussions held in Canada, particularly within the Fastenings Subcommittee of the CSAO86 Technical Committee. Proposals have been formulated regarding specific code changes, in some instances extending to creation of code language. The remainder of this document concentrates on giving some sense of approaches taken, and elucidation of current thinking regarding what are the key outstanding ‘difficult questions’ and how to move forward on them. 6.1 Transparent Mechanics Models for Mechanisms


Figure 6 – Illustration of failure mechanism: row tear-out mode for dowel fasteners

Even in these days when there is computer software that automatically applies code design rules to standard types of joint designs [12], many engineers still regard the existing code provisions as being confusing and too complex. Commonly they complain that it is not possible to distinguish design solutions resulting in ductile joint failure from those resulting in brittle joint failure [1]. Therefore it is intended that mechanics based models used to analyse various failure mechanisms, at joint or connection level, will be given in a format that makes the nature of the associated failure explicitly transparent. Also it is intended to avoid complexity even if that sacrifices accuracy to some extent, and to illustrate each mechanism diagrammatically so that it will be intuitively apparent why various terms appear within the associated mechanism equations. Figure 6 gives one simple illustration of the level of complexity intended and what is meant by transparency. 6.2 Handling Uncertainty in Source Data and Estimating Effects of Coupled Variables Because of the complexity inherent to the behaviour of connections in timber structures, and the impracticality of attempting to assess that behaviour by full factorial experimental evaluations, there will always be uncertainty associated with the precision and robustness of source data. Another problematic issue is accounting for the effects of service related variables. It may be known that some factors are strongly coupled, but typically the quantification necessary to assign values to modification factors (notably kS and kT ) will be lacking. Estimates of the likely variance in connection capacities will always pose difficulty. Based on test data collected by the authors and co-workers, and expert judgement, it is proposed that coefficients of variation in resistance used to set resistance factors (φ values) be those given in Table 5. Table 5 – Proposed coefficients of variation to be assumed for setting φ values

Member Classification (Table 3) (for least ductile joined material) Splitting Non-splitting


(Table 1) Coefficient of Variation, V

Brittle 0.35 0.25 Low-ductility 0.30 0.20 Moderate-ductility 0.25 0.15 High-ductility 0.20 0.10

Assessment of variance (characterised by the coefficient of variation) in connection resistances has double edged implications under CBD, because it influences both choice of φ values for connections and the efficiency with which member capacities are utilized. So far the authors have not developed ideas on that topic to a state ready for open discussion. Eurocode 5 [13] incorporates a design condition related modification factor that couples the influences of the expected cumulative duration of design loads and the service environment. This accounts for mechano-sorptive influences on the static fatigue response of wood products. It is


based on the combination of test data, experience and code committee judgement. The current approach in Canada is that the effect of ‘duration of the loading’ is assumed to be uncoupled from other influences. Values of the associated modification factor are derived from data for dimension sawn lumber, but are also applied to all other wood products and all connections. That is clearly a technically inappropriate practice but simplifies design calculations. The authors favour aligning the Canadian approach with Eurocode 5 or a closely related variant. As discussed in more detail elsewhere [4] a related issue is that in Canada currently standardized specified resistances ( Rs values) have a basis that adjusts from short-term test strength to that associated with so-called standard term load duration. Like the duration of loading modification factor the basis is test data for dimension sawn lumber with standard duration resistance being taken as 0.8 time short-term test resistance. There is no precise temporal definition of standard term duration but it corresponds to typical combined dead plus snow roof loads and dead plus occupancy floor loads in Canada. The calibration factor 0.8 is based on results of comparative reliability analysis with and without incorporation of static fatigue over the design lifetime [14]. Thus any attempt to clean up the basis modification factors cannot be disassociated from the notion of having standardized specified resistances that are based on standard term loading. 6.3 Assignment of Resistance Factors As an illustration, the case of assigning resistance factors to connections under SBD is considered here. Taking the simplest possible formulation, the resistance factor can be calculated as [15]:

)exp( Vαβφ −= ………… (3)

where: α = calibration coefficient = 0.75 [10], β = safety index [Table 1], and V = coefficient of variation in resistance [Table 5]. Table 6 lists the suggested φ values with results rounded on the basis of 0.05 increments. It should be noted that these φ are intended to be implemented within equation (1) in conjunction with resistance R that is the average resistance adjusted to design conditions. The reasons are those outlined in Section 2 concerning practical impossibility to reliably determine R on another basis (except if there is adequate test data specific to the design conditions [4]). Table 5 – Proposed resistance factors, applied within equation (1)

Member Classification (Table 3) (for least ductile joined material) Splitting Non-splitting

Type of parent system (Table 2)


(Table 1) Resistance Factor, φ

Brittle 0.30 0.40 Low-ductility 0.40 0.55 Moderate-ductility 0.50 0.70

No alternative load paths possible (light-frame, heavy frame, other)

High-ductility 0.65 0.80 Brittle 0.35 0.45 Alternative load paths

possible (light-frame, Low-ductility 0.45 0.60


Moderate-ductility 0.55 0.70 possible (light-frame, heavy frame, shell, plate) High-ductility 0.70 0.80

As can be deduced from combined consideration of Tables 1 and 5, and the resulting φ values, the intent is to militate strongly against designs which result in brittle connection failures. Readers may desire to compare the suggested φ values with those already adopted, for example in Canada [2]. That is not recommended because in themselves such comparisons are meaningless. Having said that, we do have evidence that the high ductility fasteners, such as nails, perform well under the current design provisions in Canada. A recent code change proposal submitted by the authors to the CSA O86 Technical Committee to adopt mechanics-based design for wood screws and nails used the proposed approach to derive resistance factor. It was shown that the new factored lateral resistances for nails are similar to the current design solutions for the majority of design cases. This is evidence that at least the upper limit of the values presented in Table 5 for ductile connections are reasonable. Impacts of proposed new codes can only be sensibly assessed by comparing proposed design solutions with those resulting from application of past design codes. This said, it can be stated that approximately there would be relatively minor changes from existing solutions for what here are defined as brittle mechanisms. Solutions would be liberalized for situations where high-ductility mechanisms govern. Such an outcome is believed entirely justifiable, based on what as yet are mostly unpublished test findings by the authors and co-workers under the research project titled ‘UNB2 - Design Methods for Connections in Engineered Wood Structures’ (2003-06), and follow up work in progress (see Acknowledgements). 6.4 Size of System Effects on Failure Capacity There has for many years been discussion on the topic of the relationship between the sizes of wood product components and their apparent strengths. Various empirical explanations and even some theories have been advanced in this respect. Clearly various influences are at play including that the sizes of components influence the ratio of strain energy released during brittle failures relative to the energy required to create new fracture surfaces. From this alone, it is to be expected that components will often appear weaker and more brittle (in terms of material strength rather than component capacity) the larger they are. Severity of the effect depends on constraints applied to specimens and the type of loading control. The same general phenomenon will apply to connection tests. This may not mean however that size of system effects are an issue important enough to be addressed during design of timber structures. It will depend on the constraint applied to real structural members and subsystems by the parent system and the type of loads that do the external work. The nature of any structural system determines whether fast fracture effects will be an issue, with the collapse of the Twin Towers in New York being the two most famous fast fracture collapses. Historically timber structures do not seem to have been unduly susceptible to fast fracture phenomena, but that does not mean that future timber structures will be of types that are immune especially if new materials and construction forms are employed. Weakest link theories have been frequently applied to timber members, most probably as a


device that mimics effects of many factors. Factors incorporated will range from fast fracture to alterations of the material quality (e.g. regarding by crosscutting sawn lumber), to effects of boundary conditions by changing specimen aspect ratios, etc, etc. This is not to say there will not be some inherent material variability leading to ‘weak link effects’ on apparent strength. Wood product strength properties can be conservatively assigned by basing them on maximum available product dimensions. However for systems containing multiple connections such an easy strategy is not available. The number of connections that are loaded in series will effect the capacity of a structural system, and the number of connections in a system factor ( kN ) is included in equation (2) as a recognition that weak link effects could be an important factor. An analogous situation is the effect that finger joint frequency has been found to have on the apparent moment capacity of wood I-joists [16]. The authors and co-workers are giving consideration to the above but at present there is no specific suggestion on the nature of kN , beyond the deduction that it be a function of the number of connections within a system or subsystem. The whole system would be the domain of concern for a system without ability to develop alternative load paths following a component (connection) failure, and subsystems the relevant domain of concern if alternative load paths can be developed. Tentatively, a suitable design form could be:

)exp( BNAkN −= ………… (4) where A and B are parameters calibrated based on numerical analyses of collapsing structural systems (or test data if any exists), and N is the number of connections in series. Undoubtedly A and B would need to be a function of the type of system (Table 2) because that influences how forces will redistribute and accounts for unequal initial forces transferred by individual connections. The authors and co-workers have made some tentative first steps towards undertaking such collapse analyses. 7. Level 1 versus Level 2 Design Approaches It is proposed that Level 1 design provisions will be similar in nature to those in the present Canadian code [2] and permit only SCD. Focus would remain on design of joints with only indirect recognition of connection failure mechanisms. There would be explicit recognition of the difference in failure characteristics and inherent levels of variability exhibited by wood products classed as non-splitting as opposed to those classed as splitting types. Also there would be recognition in assignment of φ values regarding the nature of the governing failure mechanism and necessary safety levels. The content of Tables 1 and 3 to 5 would apply to Level 1 designs. However there would be no account taken of the nature of the parent structural system, beyond recognizing whether or not it was capable of creating alternative load paths were a connection to fail. Capacities of joints would therefore be mostly uncoupled from system design. Capacities of connections would be permitted to be determined in Level 1 based on analysis employing mechanics models, or test data using supplemental approaches (most likely contained in a document separate from the main code document). Essentially, the changes proposed for Level 1 implement modern knowledge about how connections behave and should


be designed. The approach would permit integration of new wood products into the regulated design framework. It would assist designers to create better structural systems, but not directly link design of systems and connections as being the same or reward designers for selecting the most appropriate combinations of structural form and materials. The Level 2 approach as proposed would fully implement all possible enhancements of design discussed above. Therefore CBD would be allowed and the design of connections would be intimately linked to design of systems as indicated in the set of design steps shown in Figure 5. Level 2 would very explicitly allow designers to profit from being skillful at their trade in the sense that they could achieve solutions not possible before. The difference between design approaches can be summarized through the applicable generic SCD equations for strength limit states (employing the definitions applicable to terms in equations (1) and (2)):

Level 1: ≥= TSS kkRR φφ effects of factored loads ………… (5)

Level 2: ≥= HNTSS kkkkRR φφ effects of factored loads ………… (6) Although outside the discussion here, both levels of code document should concern themselves with important matters like guidance on attaining good fire performance and durability of structural connections and parent systems. 8. Concluding Comments The design of connections in timber structures has often been relegated to a secondary role by code committees and designers. Whether or not it is intentional, that is a great error and hopefully this paper makes some contribution toward rectifying the situation. Elucidation of how design code provisions related to connection design could be improved requires a broad based shift in the philosophical basis of code provisions. Incremental improvement does not seem an adequate response to what is necessary and therefore it is hoped that code committees will be bold in their actions. Although there is strong attention here to the timber design code situation in Canada, it is believed that the issues are the same elsewhere and therefore much cross-fertilization is possible. Despite national or regional difference in details of design practices for connections, the authors believe much harmonization is possible and should be strongly encouraged. In Canada discussion is at the stage where serious decisions need to be made and it is expected that the key ones will be made by the end of 2006. To avoid unnecessary incremental adjustments to the Canadian code, proposed changes need to be coordinated with changes affecting system design of timber structures. Although not discussed in detail here, the driving force leading to consideration of major change across the scope of the Canadian timber design code is transition of the National Building Code of Canada (NBCC) [17]. The NBCC is moving away from prescriptive design toward so-called Objective Based Design (OBD), which is a variant of performance based design. There is consequently impetus to create tools (material design codes) that recognise that construction products of all types are evolving, construction methods are evolving, and the types of structures society in general requires architects and engineers to create is evolving. Materials codes that do not maintain currency will largely


become irrelevant as OBD only regards them as one avenue to towards creation of acceptable solutions. The Canadian situation is not unique and parallel broad adjustments in the regulation of building design are occurring around the globe. Timber design codes that continue to be obsessively focussed on use of traditional products, like sawn lumber, glulam or generic nails and bolts, will quickly become obsolete. A big question related to this discourse is whether colleagues and code drafting committees will regard some of the changes suggested as too radical. The authors believe that it has been known for a long time that connection design practices need to be radically overhauled [e.g. 18], and “keeping putting off the evil day” will not help matters. In respect of design complexity, nothing that is proposed would raise the required level of designer expertise beyond what is required to design, for example, reinforced concrete components. It might well enhance the status of timber engineering if designers are given tools (codes) that permit them greater scope than now to achieve innovative structural solutions. Acknowledgements The authors thank the Natural Sciences and Engineering Research Council of Canada, The Canadian Wood Council and Forintek Canada Corp. for financial support over the last decade that enabled essential background knowledge to be developed. More recent support from Natural Resources Canada under the project ’UNB2 - Design Methods for Connections in Engineered Wood Structures’ (2003-06) has facilitated extension data to encompass connections in relatively new types of wood products, and has assisted creation of what is proposed here. Many graduate students, postdoctoral fellows and others have played important parts and that is acknowledged. Patience of colleagues in the CSA086 Technical Committee is appreciated because the ideas expressed here are in large part the result of many long debates under its auspices. References 1. Snow, M., Asiz, A., Chen, Z. and Chui, Y.H. (2006). North American Practices for

Connections in Wood Construction, Progress in Structural Engineering and Materials, 8(2): 39-48.

2. Canadian Standards Association (CSA). 2005. “Engineering design in wood”, CAN/CSA Standard 086-01, CSA, Toronto, ON, Canada.

3. Smith, I., Chui, Y.H., and Quenneville, P. 2006. “Overview of a new approach to handling system effects in timber structures”. CIB-W18 Meeting in Florence (in press).

4. Smith, I., Asiz, A., Snow, M. and Chui, Y.H. 2006. “Possible Canadian / ISO approach to deriving design values from test data”, CIB-W18 Meeting in Florence (in press).

5. Smith, I. and Foliente, G. 2002. “LRFD of timber joints: International practice and future direction”, ASCE Journal of Structural Engineering, 128(1): 48-59.

6. Paulay T. 1981. “Developments in the seismic design of reinforced concrete frames in New Zealand”, Canadian Journal of Civil Engineering, 8: 91-113.

7. Chui Y.H. and Smith, I. 1993. “Capacity design of wood structures”, Proceedings of Annual Conference of Canadian Society for Civil Engineering, II: 365-374.

8. Murty B. 2005. “Wood and engineered wood connections using slotted-in steel plate(s) and tight-fitting small steel tube fasteners”, MSc thesis, University of New Brunswick,


Fredericton, NB. 9. Snow M. 2006. “Fracture development in engineered wood product bolted connections”, PhD

thesis, University of New Brunswick, Fredericton, NB. 10. Smith, I. 1985. “Methods of calibrating design factors in partial coefficients limit states

design codes for structural timberwork: with special reference to mechanical timber joints”. Research Report 1/85, Timber Research and Development Association, High Wycombe, Bucks, UK.

11. Smith, I. Foliente, G. Nguyen, M and Syme, M. 2005. “Capacities of dowel-type fastener joints in Australian pine”, ASCE Journal of Materials in Civil Engineering, 17 (6): 664-675.

12. CWC. 2005. WoodWorks® Structural Design Software. Canadain Wood Council, Ottawa, ON.

13. CEN. 2004. “Eurocode 5_Design of timber structures: Part 1-1: General - Common rules and rules for buildings”, Standard EN 1995-1-1(E), Commission for European Normalization, Brussels, Belgium.

14. Foschi, R.O., Folz, B.R. and Yao, F.Z. 1989. “Reliability-based design of wood structures”. Structural Research Series Report No. 34, Department of Civil Engineering, University of British Columbia, Vancouver, BC, Canada.

15. Ravinda M.K. and Galambos T.V. 1978. “Load and resistance factor design for steel”, ASCE Journal of the Structural Division, 104(ST9): 1337-1353.

16. Shapr, D. J., Suddarth, S. K. and Beaulieu, C. 2000. “Length effect in prefabricated wood I-joists”. Forest Products Journal, 50(5):29-41. 17. NRC. 2005. “National building code”, National Research Council, Ottawa, ON. 18. ASCE. 1996. Mechanical connections in wood structures. ASCE Publication No. 84. American Society of Civil Engineers, New York, NY.


Appendix E – System Based Design of Structures


Overview of a New Approach to Handling System Effects in Timber Structures

Ian Smith and Ying Hei Chui

University of New Brunswick, Fredericton, NB, Canada

Pierre Quenneville Royal Military College, Kingston, ON, Canada

Abstract

In Canada ideas are being formulated about how system design can be implemented within the national timber design code. Momentum for this comes from need to address the changing nature of the construction industry and allied professional engineering practice. The end product needs to be something that can be seamlessly transitioned into design practice. System design is not envisaged as an approach that will yield unfamiliar solutions for familiar problems. Rather it is seen as supporting creation of timber solution in new situations. This paper attempts to identify the major framework issues, starting with the question of what the term system should mean. There is no attempt to follow faddish trends that fly under the same label. It is thought that landscape level changes will shape system design in Canada, and presumably elsewhere, and will concern evolution of architectural solutions, rapid broadening in the available range of structural products (wood, non-wood and composite varieties), changing construction methods and practices, and evolution in regulatory regimes. Detailed technical issues such as what constitute acceptable ‘search tree’ and ‘structural reliability’ algorithms will need to be addressed at some point, but that should be after framework issues have been resolved. 1. Introduction Currently the timber design code in Canada gives cursory guidance on system level issues and detailed guidance on ‘free body sizing’ of components. The focus of this discussion is a transition from that to a document that gives integrated guidance on design of structural systems constructed mainly from wood products. The transition should mostly be about philosophical shifting of emphasis, with the intent that a restructured code will support systemic thinking. This will not, of course, diminish the need for technical updating of detailed component sizing rules, but it will modify the range of factors that enter them. Although emphasis here is on the broad issues, it is recognized that fine details and the mathematical intricacies must be addressed at some stage. However, doing that now would as the old British saying goes “to put the cart before the horse”. Formulation of design codes is (like design itself) an essentially subjective process founded as much on professional judgment as exact data and precise formulae. Legendary statistician


George Box is credited with saying “All models are wrong, some models are useful”. By analogy, all codes are based on inexact concepts, but despite this, they usually are useful aids to engineering practice. One can add that codes remain useful as long as they support contemporary needs. There can never be any best way of formulating a design code, just as no single design solution is correct. Criteria for measuring perfection are subjective and discourse, and suggestions here are simply one of many parallel tracks that could lead to creation of effective code provisions. Under existing timber design codes the procedures result in what have been proven, by practical experience, to be acceptable design solutions for recurring types of design problem. Therefore, if the nature of timber construction were never to evolve further practices formulated in those documents would continue to result in adequate solutions. With adequate solutions being those acceptable to society in terms of balancing structural safety, material conservation and cost. However, as is abundantly clear in Canada and other parts of the westernized world, the nature of construction of all types is changing rapidly. The driving forces are well beyond the control of the timber community and include aging and more urbanization populations, resource scarcity, climate change and international trade practices [1, 2]. Evolving tastes, needs of society and business, and technological development drive architectural design. Future practices will not replicate past utilization of construction materials. Architecture of buildings is changing, e.g. more wall openings, more irregularity in plans and roofs. There is rapid broadening in available types of structural products and connection hardware (not just wood products). Construction methods and practices are changing, as are regulatory regimes that control design and construction practices. What in Canada is called Objective Based Design [3] is helping change the landscape. Although not yet impacting design in Canada, it is clear from international trends that future design practices will need to address more than traditional loading scenarios. For example, avoidance of disproportional collapse under accidental or deliberate blast loads has becoming an important issue [4]. Avoidance of disproportionate collapse whatever the causal agent of damage is an excellent example of an issue that can only be handled properly via systemic design approaches. Systems design methods are believed necessary mainly because they embody the approach most likely to result in robust ways of doing new things under new regulatory regimes. In practical terms, systems level design becomes necessary when: o non-traditional loads must be resisted (e.g. bomb blasts), o non-traditional structural forms are employed (e.g. very large wall openings, complex roofs), o non-traditional combinations of materials are employed (e.g. wood-plastic composites), or o non-traditional intra or inter component, or system support, connections are employed (e.g.

proprietary hardware). As already indicated, when only common traditional timber products are employed within traditional structural forms existing approaches are adequate. This is because the free body sizing rules for components implicitly acknowledge, i.e. were calibrated to incorporate, the system context. For example, it is implicit that small dimension sawn lumber joists will be used in load-sharing arrangements. For traditional combinations of materials and traditional connections it is known by experience that component deformation characteristics will be compatible. Also, that problems due to factors like excessive mechanosorptive creep and local instabilities are unlikely.


However, to again illustrate in terms of floors, when wood I-joist and open-web joist products with tall cross-sections were first employed, codified design practices based on experience with systems having sawn lumber joists broke down. This was because the rules no longer implicitly reflected the appropriate system issues. Usually the problem was that the new products altered the two-way structural action within floor systems, or that warping and instability of joists became issues. The strategy to achieve satisfactory design solutions using new joist products was to substitute proprietary design methods in lieu of code rules. Rapid evolution is occurring in many countries with respect to the range of available: construction products and materials, construction methods and structural design tools. Concurrently, the range of options for how new products and technologies can enter the marketplace is being opened up. Often this is for trade reasons, and because of a broad public and political perception that a more open regime will provide society with better purchasing choices. Regulatory frameworks are moving towards greater choice concerning how acceptability of engineering designs can be demonstrated. In Canada under Objective Base Design using rules in material codes, such as the CAN/CSA timber design code [5], has become just one of several optional parallel approaches to proving acceptability of any design. So far the ‘other approaches’ are only selectively employed to justifying acceptability of proprietary components or systems. If in even the medium term (circa > 5 years from now) the timber design code in Canada, and presumably those elsewhere, is to remain relevant to needs of structural designers that document will have to be subjected to more than cosmetic or incremental surgery. This discussion is predicated on the assumption that timber design codes need to be reshaped so that they focus more directly on how wood products are utilized. Option 1: Make systems level design mandatory. Option 2: Maintain existing style provisions for traditional wood products (sawn lumber, glulam and panel products) in traditional applications employing traditional connections; in parallel with new systems level provisions that can be applied to all wood products. 2. What is a Structural System? It is proposed that the term structural system means any arrangement of components capable of developing collapse mechanisms (internal to its domain) that involve deformation or failure of more than one component. In essence, any stable arrangement with inter-component connections would be a system. A secondary consideration is what constitutes a component. To be consistent with the proposed definition of a system, a component must be any manufactured or prefabricated part that can develop failure mechanisms that are completely internal to its domain. Types of parts easily classed as components include sawn timber framing and glued-laminated timber members in ‘post and beam’ construction; wall, floor and roof panels; and connections. Other parts may not be easily classified because their end use will determine whether they behave as systems or components (e.g. trusses, spaced columns).


Depending on the complexity and size of a building structure, a design engineer might need to define and analyse a nested series of systems, with higher level systems incorporating the lower level systems (subsystems). The mechanisms considered would differ between the nested systems. At the purely practical level a system would be any arrangement for which an engineer decides that a separate structural analysis is required. Engineers will resolve dilemmas through professional judgement, as at present. Reference above to free body sizing of components means the consideration of only the effects that internal forces have on any component that has been abstracted from its parent system. Under the free body approach (which is the basis of existing timber design codes [e.g. 5]) sizing of components is independent of the nature of the parent system once the internal forces have been estimated. Thus the essential difference between what the authors call systemic design and current practice is that under system level design both the nature of system mechanisms and the consequences of any particular component failing would enter component sizing rules. Thus necessary sizes of components would not just reflect the magnitude of factored free body forces. Recommendation 1: Classify systems according to the structural form, and components according to the type of materials. Recommendation 2: Provide explicit guidance on the necessity to consider both system and component level failure mechanisms. Recommendation 3: Make sizing rules a function of system and component classifications and the expected consequences of component failures. 3. Seamless and Shock Free Transition Complaints from practicing engineers have arisen following previous revisions of timber design and other codes. Concerns focussed on: o Whether revisions result in meaningful differences in design solutions. o The extent of retraining necessary to learn about content and how to apply new codes. These two aspects can be traded off because practitioners realise that retraining is worthwhile provided there are substantive gains in terms of material economies and flexibility in specifying design solutions. The last major restructuring of the Canadian timber design codes took place in 1984 when the Allowable Stress Design (ASD) code was supplemented by a Load and Resistance Factor Design (LRFD) code. The ASD code was not revised between 1984 and when it was theoretically withdrawn in 1994. In practice the ASD code remained in wide use until at least about 2001. During the same period the parallel LRFD code was revised three times. Initially only a relatively short transition period was intended before the ASD code was eliminated. Use of ASD persisted in Canada because many engineers preferred the older option, and in response building authorities in provinces, territories and cities permitted the practice. Anecdotally it is understood that many older engineers still use the ASD code! Clearly there exists the perception that


inadequate benefits accrued from code changes. The experience is understood to have been replicated elsewhere (e.g. UK, USA). Recommendation 4: Major modifications should only be made to design codes if that results in substantive alterations to design solutions, and/or they expand the range of what can be done with wood products. Recommendation 5: To be consistent with the philosophy behind Objective Based or Performance Based Code and minimize the ‘shock factor’, new provisions should be transparent with respect to intent, intuitively consistent with engineering principles (avoid empiricism), and minimize format and content changes to design equations. 4. Scope of System Design Provisions Figure 1 summarizes, in the Canadian context, the proposed scope of system level design provisions within the timber design code, and how those should compliment those in the national building code.

Figure 1 – Proposed scope of system design provisions Figure 2 shows the proposed sequence of design decisions around which code rules should be shaped. The intent is that the timber design code would explicitly define how choice of any particular combinations of structural form, wood product construction material(s) and connection methods affects the expected system failure mechanisms. The nature of the system and its expected failure mechanism would govern the assignment of system related factors entering component design equations. This should encourage designers to make direct links between design decisions and efficiency with which materials can be utilized. For example, if designers select statically determinate structural forms and splitting prone members, and therefore failure of any one component would result in system collapse, component sizing rules should be relatively conservative (i.e. aimed at attaining a relatively high reliability level). By contrast, if statically indeterminate structural forms are selected and failure of one or more components would not result in system collapse, or if ductile failure mode was ensured through the capacity design approach, then sizing rules could be relatively liberal. For all systems the level of ductility achievable in connections should also enter the sizing rules for other components [6]. Recommendation 6: Embed ‘rewards’ into design code rules so that there is incentive to refine design practices, i.e. make it possible to achieve higher component resistances and to do new things by adopting systemic design practices.


Figure 2 – Proposed sequence of design decisions Within step 5 in Figure 2 choice of design method refers to the possibility of what here is called Strong Component Design (SCD) or Capacity Based Design (CBD), with the former being equivalent to current practice [5]. CBD concepts were originally developed for design of reinforced concrete structural systems [7] but could be applied to timber structures [8]. The choice of Level 1 versus Level 2 methodologies in step 6 relates essentially to the notion of permitting designers to select between alternatives of simple or complex sizing methods for components. 5. Format of Design Equations For consistency with Recommendations 4 to 6 it is proposed that under the systemic approach the design equations for sizing component be simple modifications of existing LRFD equations. Below only the strength LRFD equations applicable under SCD are given. For other situations like CBD or serviceability calculations the approach would be consistent with what is shown here. Whether serviceability should be addressed as part of component design is a contentious issue. The authors favour the approach that codes simply give guidance on assigning load-deformation characteristics to components for various service situations. Design engineers could then exercise professional judgement about serviceability issues. Preferably serviceability would be assessed at the system rather that component level. The SCD design strength checking equation should take the generic form:

≥Rφ effects of factored loads ………… (1) where: φ = resistance factor, R = resistance adjusted for all design specific considerations. The adjusted resistance is calculated according to the generic form:


R = R k k k ks S T N H ………… (2) where: Rs = standardized specified resistance, kS = service factor, kT = treatment factor, kN = number of components in series factor, k H = system failure factor (Sections 7 and 8). The factor kS accounts for the coupled cumulative effects of loading and service conditions. Specifics of equations (1) and (2) will vary depending on sensitivities of resistances to design variables, and exactly what parameters need appear will vary from situation to situation. It is of course necessary to have prior applicable evidence, or to collect such evidence by testing. A companion paper by Smith et al [9] discusses some related issues. Recommendation 7: Introduce into design equations for component sizing modification factors accounting for the number of components and the system characteristics. As suggested under Option 2 one possibility is to create dual level design methods. If this was done Level 1 provisions could be similar in nature to those in the present Canadian code [5] and permit only SCD. There could still be explicit recognition of the difference in failure characteristics and inherent levels of variability exhibited by various wood products and connection methods (Section 6 and 7). Also there could be recognition in assignment of Level 1 φ values of the natures of the governing failure mechanisms for components. However there would be no account taken of the nature of the parent structural system. Sizing of components would therefore be uncoupled from system design. Familiar solutions would result for familiar problems. The Level 2 approach would fully implement all possible enhancements of design discussed in this paper and companion papers [6, 9]. Recommendation 7: The structured decision sequence illustrated in Figure 2 be embedded directly in design codes irrespective of whether Level 1 or Level 2 methodologies are followed. Option 3: Permit dual level practices that give designers the option of ignoring or explicitly embracing new systemic concepts. The applicable generic SCD factored resistances for strength limit states would be: Level 1: ≤= TSS kkRR φφ effects of factored loads ………… (3) Level 2: ≤= HNTSS kkkkRR φφ effects of factored loads ………… (4) 6. Classification of Systems, Materials and Connections Tables 1 to 4 illustrate preliminary thinking about classification decisions embedded in the decision sequence of Figure 2. Only two classes of wood products are suggested, Table 3, because ability of systems to achieve the more refined four-level classifications scheme of Table 1 is mostly a function of the connection characteristics. The notion that member materials can be divided into splitting and no-splitting categories reflects direct experience in tests conducted by the authors for a broad range of wood products (e.g. sawn lumber, glulam, Engineered Wood Products, Structural Panels).


Table 1 – Tentative classification of failure mechanisms based on Ductility Ratio (D), and the associated Reliability Index (β value)

Brittle Low-ductility

Moderate-ductility

High-ductility

Classification of failure mechanism of a component

D ≤ 2 2 < D ≤ 4 4 < D ≤ 6 D > 6 Type of parent system (Table 2) β value No alternative load paths possible 4.5 4.0 3.5 3.0 Alternative load paths possible 4.0 3.5 3.0 2.5

Table 2 – Proposed classification of structural systems

Classification Description Light frame without capability to develop alternative load paths

Systems where less than four linked and parallel framing members or subsystems act together.

Light frame with capability to develop alternative load paths

Systems where four or more linked and parallel framing members or subsystems act together.

Heavy frame without capability to develop alternative load paths

Statically determinate, or other, systems where failure of one component would cause disproportional system damage.

Heavy frame with capability to develop alternative load paths

Statically indeterminate systems where failure of one component would not cause disproportional system damage.

Shell and plate structures Those structures depending on two-way curvature and folded plate action for stability.

Other Any system not within another classification. Table 3 – Proposed classification and expected behaviours of structural wood products

Classification Expected behaviour Examples Splitting Brittle response sawn timber, un-reinforced glulam, LVL, PSL Non-splitting Ductile response reinforced-glulam, plywood, OSB, LSL

Table 4 – Example: Proposed classification of failure mechanisms of connections (when capacities and failure mode are estimated using mechanics based models)

Member Classification (Table 3)

(for least ductile joined material)


(Table 1)

Splitting Non-splitting

Governing mechanism (Examples only)

Brittle X row tear out, block shear, net section, member interference

X bearing failure beneath fasteners with only rigid body movements of fasteners

Low-ductility

X X all connection level mechanisms Moderate-ductility

X plastic hinges formed in fasteners

High-ductility X all joint mechanisms


The intent is that classifications in Table 1 will be used within reliability calculation processes that scale resistance factors (φ values) for components, and classifications in Table 2 would be used during assignment of system failure factors ( k H values) for components. Classifications in Tables 3 and 4 would be subsidiary information entering detailed sizing rules for components. The logic that underpins Table 4 is that in recent studies the authors have found that a correlation often exists between the type of failure mechanism for components and their ductility ratios. Thus when capacities for various mechanisms are estimated using mechanics based models, the nature of the mechanism can be employed as a surrogate in lieu of direct evidence concerning the magnitude of the ductility ratio (D). Obviously, when mechanisms and capacities are determined directly from test data [9], classification of the response on the basis of Table 1 would be straightforward. So far the authors have not formulated firm ideas about how system ductility can be estimated without recourse to complex analyses. One possibility is to create an indexing method that assigns system ductility as a function of component ductility, and employing classifications in Tables 3 and 4. Possibly contributions of individual components could be weighted as is commonly done in post-disaster ‘use assessments’ of buildings. Numerical case studies are expected to be an appropriate means for focussing insight and clarifying suitable approaches. Recommendation 8: Reliability index values, and consequently resistance factors, for component design should be a function of the classification of the expected failure mechanism and whether a system is capable of developing alternative load paths following component failure. 7. System Failure Mode Design of systems involves aggregation of effects of mechanisms for components and identification of possible global failure mechanisms associated with the overall system assembly. Failure mechanisms individually or in combination define the failure mode of a system. It is proposed that the term failure mode be reserved for describing the gross nature of the failure in any complete system. Currently in Canada the timber design code [5] together with the national building code [3] specifies global design requirements and the loads to be considered. Provisions recognize that structural engineers must pay attention to whole system behaviour. System level design checks required by the building code relate to maintenance of equilibrium under pseudo-static design rules, avoidance of instability, and avoidance of excessive deformations. Member design rules distinguish between heavy-framed and light-framed construction systems. Light-frame systems are assumed capable of load sharing behaviour if certain prescriptively defined caveats are met (e.g. spacing of structural members). No load-sharing capability in heavy-framed systems is acknowledged by the code. The stipulations do not reflect how light or heavy frame systems actually behave. Relatively recently, circa the last 10 years, the Canadian design code has had major modification to its provisions related to the design of shear walls. New design rules treat shear walls as systems that correspond to complete length walls. However as yet there is no linking of shear wall design to the nature of the parent system. Design of connections totally ignores systemic issues.


Recommendation 9: System design level design guidance and required design checks should remain the prerogative of the parent building code. Recommendation 10: Timber design codes should draw attention to and amplify, if appropriate to use of wood products, system design level requirements in the parent building code. The first of these recommendations reflects the opinion that the broad guidance on systemic aspects of design that is contained in the Canadian building code is adequate. Also that it does not (and should not) impinge on ability of designers to exercise their professional skills. The second recommendation recognises that currently the timber design code does not go far enough in contextualizing guidance on the systemic aspects of design using wood products. 8. System Related Modification Factors (kN and kH) The notion behind the modification factor kN which accounts for the number of components in a system is that the apparent resistances of components decrease in proportion to the ‘stressed volume’ of material in structures. There has been discussion for many years in timber engineering circles about the inverse relationship between the amount of material that members contain and their apparent material strength. Various empirical explanations and even some theories have been advanced in respect of size of member effects on apparent strength, especially for sawn timber, structural composite lumber and glulam. Many influences are at play, but the most important factor is that as sizes of components and systems increase the ratio of strain energy released during component failures is increased relative to the energy required to create new fracture surface [4, 10]. The apparent brittleness of components and systems tends to increase with size, and hence apparent strength reduces, because the ratio of energy released as the result of a part failing increases compared with the energy consumed in breaking that part [4, 10]. Brittleness and reductions in apparent strength do not (cannot) occur indefinitely as systems get bigger [10] and that needs to be recognized. Tentatively it is suggested that the number of components in a system be considered a surrogate for introducing a relationships between system size and apparent decreases in material resistances of components. How significant the issue is will depend on the nature of the component failure and the characteristics of systems [4]. Recommendation 11: Detailed case studies be performed across a range of structural systems to determine the significance of system size effects on brittleness and apparent strength. The notion behind the modification factor k H which accounts for the nature of the parent system and the expected classification of the system failure mode is that not all systems are equally capable of permitting components to attain their free body capacity. For members this could possibly be an extension of the current approach in the Canadian design code [5], wherein factors are applied to enhance isolated component capacities provided that the system meets certain criteria (Section 7). For connections considerations and approaches can be expected to be quite different from current approaches for members. Incorporating connections into systems can


either enhance or diminish their isolated component capacities, depending on the basis of those capacities. To illustrate, assume that capacities of connections are based on post-yield resistance (ultimate capacity) when loaded in isolation. However, intra system compatibility constraints resulting from the structural forms of systems would not always allow connections to attain displacement levels necessary to mobilize their ultimate resistances, prior to the complete system collapsing. Such a possible situation occurs for nailed joints in shell structures. Similar situations can be envisaged for members. The system classification scheme in Table 2 could be a surrogate for linking attainable component resistances to system type. Recommendation 12: Detailed case studies should be performed across a range of structural systems to determine how best to handle effects of system type on the ability of components to achieve free body capacities. Consideration should also be given to what constitutes the most appropriate interpretation of the resistance responses of components. 9. Adoption of Reliability Concepts So far reliability concepts have supported code committee decision making in North America in respect of load and resistance factors under LRFD formats. This applies only to component design (members and structural panels). Approaches adopted are based on so-called nominal rather than true reliability. Analyses imply nominal failure rates of about 1 per 1000 over the design lifetime (2.4 ≤ β ≤ 3.0). Of course this is far from realistic and reflects that neither structural representations of components nor the reliability techniques are fully realistic. Nevertheless they are useful, as George Box eloquently implied. For systems, or indeed some types of components (certainly connections) currently applied reliability methods will undoubtedly need to be supplemented or even totally changed [9]. Undoubtedly it will be necessary to get closer than now to dealing in terms of true reliability, and with realistically acceptable failure rates. Recommendation 13: Options for system level reliability concepts need to be fully investigated. It is important that this last recommendation not be thought of as only addressing the need for debate on mathematical intricacies. As was stated at the outset of this discourse, it is necessary to adopt a much broader perception of what is needed, and to create something that engineers will want to use. 10. Concluding Comments The authors do not imagine that they have identified all the relevant issues, nor do they imagine that the approaches suggested will meet with universal support even in Canada. They hope that this “Aunt Sally” will promote debate on issues surrounding the introduction of systems level design into timber design codes. Not debating this creates the risk that such documents become obsolete and that a changing world passes them by. References


1. Brown, L.B. 2006. “Plan B 2.0: Rescuing a planet under stress and a civilization in trouble”, W. W. Norton & Company, New York, p. 365.

2. Smith, I. 2006. “Reaching for the limits with timber construction”, Symposium - Responding to

Tomorrow's Challenges in Structural Engineering, Budapest, September 13-15, International Association for Bridge and Structural Engineering (in press).

3. NRC. 2005. “National building code”, National Research Council, Ottawa, ON. 4. Smith, J.W. 2006. “Structural robustness analysis and the fast fracture analogy”, Structural

Engineering International, 16(2): 118-123. 5. Canadian Standards Association (CSA). 2005. “Engineering design in wood”, CAN/CSA

Standard 086-01, CSA, Toronto, ON, Canada. 6. Quenneville, P., Smith, I., Asiz, A., Snow, M. and Chui, Y.H. 2006. “Generalised Canadian

approach for design of connections”. CIB-W18 Meeting in Florence (in press). 7. Paulay T. 1981. “Developments in the seismic design of reinforced concrete frames in New

Zealand”, Canadian Journal of Civil Engineering, 8: 91-113. 8. Chui Y.H. and Smith, I. 1993. “Capacity design of wood structures”, Proceedings of Annual

Conference of Canadian Society for Civil Engineering, II: 365-374. 9. Smith, I., Asiz, A., Snow, M. and Chui, Y.H. 2006. “Possible Canadian / ISO approach to

deriving design values from test data”, CIB-W18 Meeting in Florence (in press). 10. Smith, I., Landis, E. and Gong, M. 2003. “Fracture and fatigue in wood”, John Wiley and

Sons, Chichester, UK.

performance of mechanical fasteners used with engineered ... · performance of mechanical fasteners...

Documents