behaviour of precast concrete floor slabs exposed to standar

*Corresponding author. International Fire Safety Consultant, 9 Lakis Close, Flast Walk, Hampstead,London NW3 1JX, UK. Tel.: #44-20-7431-5372; fax: #44-20-7431-5374.E-mail address: [email protected] (G.M.E. Cooke).

Fire Safety Journal 36 (2001) 459}475

Behaviour of precast concrete #oor slabs exposedto standardised "res

Gordon M.E. Cooke*Visiting Professor, Department of Civil Engineering, City University, London, UK

Received 2 February 2000; received in revised form 24 October 2000; accepted 10 January 2001

Abstract

This paper quanti"es the thermal movements of 14 simply supported precast reinforcedconcrete #oor slabs of 4.5m span and 900mm width exposed to two standardised heatingregimes used in "re resistance furnace tests. The tests were designed to show the e!ect of varyingthe slab thickness, type of concrete, imposed load, so$t protection and nature of "re exposureon the mid-span #exural de#ection and axial movements of the slab ends. Measured de#ectionsshowed that during the 90min design period of "re resistance thermal bowing was dominantand the e!ect of the 1.5 kN/m� design imposed load was small. The NPD hydrocarbon "reexposure caused a doubling of the #exural de#ections achieved using the standard BS 476: Part8 (now Part 20) "re exposure in the "rst 20min of exposure. � 2001 Elsevier Science Ltd. Allrights reserved.

Keywords: Floor slabs; Concrete; Standard "re tests; Thermal response; Structural response

1. Introduction

As part of the Building Research Establishment (BRE) Large Panel Structuresresearch programme, the Fire Research Station (FRS) undertook a full scale natural"re test in the Ronan Point high rise block of #ats in 1984. The test was terminatedbecause the "re exposed 4m long #oor slab spanning onto the external wall exhibitedan unexpected high rate of increase in the mid-span de#ection after only 12min fromignition. There was concern that, with further heating, the associated axial expansion

0379-7112/01/$ - see front matter � 2001 Elsevier Science Ltd. All rights reserved.PII: S 0 3 7 9 - 7 1 1 2 ( 0 1 ) 0 0 0 0 5 - 4

of the precast slab could push out the load bearing external wall panels at thewall/#oor junction causing an eccentric loading condition for the load-bearing ex-ternal wall panels which might precipitate a &pack of cards' type of progressivecollapse. A search of the literature revealed a paucity of experimental data on axialde#ections of concrete slabs exposed to "re: while many hundreds of standard "reresistance tests have been made on #oors there has never been a requirement tomeasure axial de#ections and such information was therefore very rare and unpub-lished.

The interdependence of thermal bowing, axial de#ection and axial restraint wasunclear. It was therefore decided to proceed with a two-part programme of "re tests.The "rst part, reported herein, would examine the "re behaviour of axially unre-strained, simply supported, precast concrete #oor slabs. The second part wouldexamine the e!ect of partial axial restraint on narrow strips of #oor construction. Forthe "rst part of the programme, the author, then working at FRS, proposed, designedand supervised seven "re tests on pairs of precast concrete slabs each nominally 4.7mlong�900mm wide. Six pairs were prepared by the BRE Civil Engineering Laborat-ory, Cardington. One pair was cut from a large #oor panel taken from the Ronanpoint #ats during demolition. The tests were made in a standard "re resistance #oortest furnace and were designed to determine the unrestrained mid-span de#ection andaxial de#ections of the slab ends at mid depth. The nature of the axial de#ectionsmeasured as time progressed is shown in Figs. 1(b) and (c).

It was assumed that the bowing behaviour of a large #oor panel which spans in onedirection will be similar to the behaviour of a narrower specimen if edge e!ects areguarded against so that unidirectional heat #ow was achieved in the narrowerspecimen. This assumption allowed two specimens to be tested side by side in the #oorfurnace, unrestrained by each other, in the simply supported condition with a span of4.5m. This also meant that the specimens could be easily manufactured, handled andtransported, and the cost of "re testing was reduced by more than 50%.

Precast #oor slabs of the kind tested are not used in modern #oor construction inmulti-storey buildings in UK which typically comprise hollow core prestressedconcrete planks with in situ topping or composite pro"led sheet steel/concrete #oordecks. In addition, the tested #oor slabs were simply supported which results inmaximum #exural defection representing the worst case scenario in which bene"cialrotational restraint generated by slab continuity over beams in multi-span #oors isignored. Nonetheless, the results have practical application to existing large panelprecast #oor construction which has little continuity at the supports and to newsingle-span conventional in situ reinforced concrete construction. The data can alsobe used to predict #exural de#ections of reinforced concrete #oor slabs havingrotational restraint but this requires an assessment of the positions of contra-#exure inthe #oor slab and any mitigating e!ect of membrane action arising from two-wayspanning, but this is beyond the scope of this paper. The test results are perhaps mostuseful for enabling both qualitative and quantitative comparisons to be made whilechanging important parameters such as the type of concrete and "re test severity.Some of these test results have been presented by Cooke and Morgan in a BREInformation Paper [1].

460 G.M.E. Cooke / Fire Safety Journal 36 (2001) 459}475

Fig. 1. Loading scheme and measured de#ections.

Numerical modelling of the thermal and structural response of "re-exposed com-posite steel and concrete structures has reached an advanced stage in the UK, much ofthe impetus coming from the recent full scale test work conducted on an eight-storeybuilding erected in the BRE large laboratory at Cardington. Universities involved inmodelling include Edinburgh, She$eld and City.

1.1. Fire test parameters

For the BRE slabs the following parameters were varied: slab thickness (150 and250mm), type of concrete (normal weight and light weight), live load (zero and1.5 kN/m� of #oor slab area), so$t protection (zero and two di!erent gypsum boardsystems) and severity of standard "re exposure (ISO 834 and the Norwegian Petro-leum Directorate (NPD) temperature-time curves). Table 1 lists the chosen combina-tions.

G.M.E. Cooke / Fire Safety Journal 36 (2001) 459}475 461

Fig. 2. Comparison of temperature}time curves used in tests.

The BRE slabs were designed to have a 90min "re resistance assuming a live load of1.5 kN/m� when exposed to the heating conditions speci"ed in BS 476Part 8: 1972(ISO 834), which is appropriate for structural elements in high rise blocks of residen-tial #ats in the UK. The time}temperature curve in Part 8 is the same as in the currentstandards i.e. BS 476Part 20, ISO 834 and the corresponding CEN standard. A com-parison of the NPD and ISO/BS temperature-time curves is given in Fig. 2. Thestructural design was based on BS 8110: Part 1: 1985 [2].

1.2. Fire test specimens

All the test specimens were 4.7m long by 925mm wide and were simply supportedat 4.5m centres. The "re exposed length was 4.0m.

The BRE slabs used concrete mixes designed to have a characteristic cube strengthof 30N/mm�. The normal weight concrete (NWC) used a siliceous (20mm #int gravel)aggregate with natural sand and had a nominal density of 2400kg/m�; the light weightconcrete (LWC) contained Lytag ( pulverised fuel ash) coarse aggregate and hada nominal density of 1800kg/m�. All the reinforcing steel bars were of high yieldribbed bar (Deformed Type 2 to BS 4449) having a nominal yield strength of


Table 1Fire test parameters�

Test Specimen Thickness(mm)

No. ofrebars

Concrete Liveload

Heating Comments

1 1 150 10 NWC No BS 476 BRE slab1 2 150 10 NWC Yes BS 476 BRE slab2 3 150 10 NWC No NPD BRE slab2 4 150 10 NWC Yes NPD BRE slab3 5 250 6 NWC No BS 476 BRE slab3 6 250 6 NWC Yes BS 476 BRE slab4 7 150 10 LWC No BS 476 BRE slab4 8 250 6 LWC No BS 476 BRE slab5 9 150 10 NWC No NPD BRE slab5 10 250 6 NWC No NPD BRE slab6 11 150 10 NWC No BS 476 BRE slab#so$t (1)6 12 150 10 NWC No BS 476 BRE slab#so$t (2)7 13 185 NWC No BS 476 Ronan Point slab7 14 185 NWC Yes BS 476 Ronan Point slab

�Notes: Live load " 1.5 kN/m�, So$t (1)"10mm thick glass reinforced gypsum board with 37mm airgap, So$t (2)"12.5mm gypsum Fireline board with 37mm air gap, NWC"Normal weight concrete,LWC"Light weight concrete, BS 476"BS 476: Part 8:1972 (ISO 834), NPD"Norwegian PetroleumDirectorate (hydrocarbon "re simulation).

460N/mm�. The primary (longitudinal ) steel was 8mm diameter. The concrete coverto the primary steel was 25 and 20mm for the NWC and LWC slabs, respectively,being appropriate for 90min "re resistance according to UK regulatory guidance [3].The concrete side cover was 25mm. The moisture content of the slabs varied between3.5 and 4.5% by weight.

The Ronan point slabs comprised a structural reinforced concrete slab of normalweight concrete nominally 180mm thick incorporating circular voids of 110mmdiameter running longitudinally at 150mm centres. This slab was overlaid witha non-composite 12.5mm thick layer of expanded polystyrene foam and a 65mmthick granolithic concrete screed. The screed and foam was present during the "retests.

Two kinds of proprietary boarded so$t protection were fabricated and installed atthe "re test laboratory by British Gypsum Ltd. One slab was protected with a 10mmthick glass reinforced gypsum (GRG) board. Another slab was protected with a12.5mm thick Fireline gypsum-based board. Both board protections were "xed to theconcrete so$t using cold formed steel members which resulted in an air gap of 37mm.These protection systems were included in the test programme as they had been usedin remedial work contracts on high rise #ats. Details of the tests are given in Table 1.

For the BRE slabs thermocouples were attached to 50mm diameter cylindricalcores of the appropriate concrete mix at a range of heights. The sensing ends werealigned horizontally so that they would lie on an isotherm and were adhered to thecore with an epoxy resin. The ends of the cores were lightly bonded to the plywood


Fig. 3. Longitudinal section through furnace showing transducer-support frame.

formwork before casting the slabs so that the position of all cores and hencethermocouples from the "re-exposed face were accurately known. The cores wereplaced at mid-span and quarter-span positions along the centreline of the slab.

1.3. Test apparatus

The tests were made in the "re resistance #oor furnace at the Warrington FireResearch Centre. The ends of each slab were simply supported. All de#ectionmeasurements were made relative to the ends of a slab using two purpose-madehollow steel frames which rested on the ends of the slab. Each frame was kept coolduring a test using a continuous #ow of water so it would not itself de#ect due toa change in ambient conditions. Linear displacement transducers (LDT's) were usedto measure vertical de#ections at mid-span and quarter-span positions. An LDT wasalso aligned horizontally at either end of the slab at mid depth so as to measure axialde#ection. The apparatus is shown in Fig. 3. Loads were applied using A-frames, twohydraulic jacks and a system of load spreaders to approximate uniformly distributedloading indicated in Fig. 1(a).

In each test, two slabs were laid side by side separated from each other and from thefurnace cover slabs with a #exible ceramic "bre seal so that the edges of the slabs wereprotected from "re and were free to de#ect during a test. Fig. 4 shows the arrangementin which a pair of slabs of di!erent thickness are being tested.

2. Test results

2.1. Temperature proxles in slabs

Averaged temperature pro"les within BRE slabs without so$t protection at 30minincrements are given in Figs. 5}7. The "gures show the e!ect of varying the "re


Fig. 4. Cross section of #oor slab specimen assembly.

Fig. 5. Temperature pro"les (150mm, NWC).

exposure and type of concrete. The pro"les in the 250mm slabs show a clearlypronounced moisture plateau at 1003C when steam is driven o! and also exhibitsteeper temperature gradients and higher temperatures near the exposed surface whencompared with the data for the 150mm slabs. Fig. 7 shows that the e!ect of the higherthermal insulation of Lytag LWC made little di!erence to the maximum temperaturesattained in the concrete, but the bene"t of LWC is considerable at the depth where thereinforcing steel is normally located: the temperature at 20mm after the 90min design


Fig. 6. Temperature pro"les (250mm, NWC).

Fig. 7. Temperature pro"les (250mm, ISO 834).


Fig. 8. E!ect of slab thickness and concrete type.

exposure to ISO 834 is approximately 3003C for the 250mm thick slab, and atthis temperature the reinforcing steel would have lost none of its room temperatureultimate tensile strength [4,5]. The e!ect of the hydrocarbon exposure whencompared to the ISO 834 exposure is, as expected, markedly to increase the temper-atures near the exposed face as shown in Figs. 5 and 6. It has to be recognised that it isdi$cult accurately to measure the temperature of concrete at the concrete/combus-tion gas interface because of the large temperature gradient.

All of the slabs tested resisted spalling throughout the full period of "re exposureand it can therefore be said that any aberrations in the temperature data are not dueto spalling; such aberrations can occur due to moisture removal during the heatingprocess and this e!ect, as previously mentioned, can and did have an e!ect. It shouldnot be concluded from this work that spalling is not a problem: it is generally acceptedthat spalling can occur where a large hogging moment is present (it was absent in thepresent tests because of the simple supports), and there is increasing evidence that highstrength concrete used in prestressed planks is prone to spalling in "re.


Fig. 9. E!ect of imposed load.

2.2. Mid-span deyections

Here some of the more important comparisons are made of measured mid-spande#ections. It is assumed that the #exural de#ection of a non-loaded slab is dominatedby thermal bowing, i.e. the self weight of the slab has negligible e!ect upon de#ectionsexcept near ultimate failure. With the exception of a 250mmNWC slab exposed to thehydrocarbon "re which su!ered runaway de#ection at 110min, none of the slabscollapsed during the exposure period of nominally 2 h so the assumption that thermalbowing is dominant within the 90min design period of "re exposure seems reasonable.

2.2.1. Ewect of slab thicknessFig. 8 shows the e!ect of slab thickness for three NWC and two LWC slabs,

respectively, when exposed to ISO 834. The thicker slab de#ects less which is what onemight expect intuitively and agrees with the theory of thermal bowing [1] whichshows that thermal bowing is inversely proportional to the slab thickness. The


Fig. 10. E!ect of heating rate.

de#ection curve for the non-loaded 180mm thick Ronan point slab suggests that therelative magnitudes of de#ection agree with the theory of thermal bowing up to80min*the curve for 180mm thickness lies between those for 150 and 250mm.

2.2.2. Ewect of concrete typeFig. 8 also shows the e!ect of concrete mix on de#ections of non-loaded slabs of

150 and 250mm thickness, respectively, when exposed to ISO 834. The lightweightconcrete slabs (incorporating Lytag aggregate made from pulverised fuel ash) de#ectmarkedly less than the NWC slabs. This may be attributed to (a) the lower thermalconductivity of the LWC which may result in lower temperatures and lower thermalexpansion in the "re-exposed structural layer, and (b) the lower coe$cient of linearthermal expansion of Lytag compared with dense aggregate, but further analysiswould be needed before the relative importance of thermal conductivity and thermalexpansion could be assessed. Certainly the di!erence in de#ections is large*at the90min design period of "re resistance the mid-span de#ections of the LWC slabs areroughly two}thirds those of the NWC slabs, and it should be remembered that theseare di!erences in thermal bowing since the slabs were not loaded.

2.2.3. Ewect of imposed loadFig. 9 shows the e!ect of imposed load on mid-span de#ections for NWC slabs 150

and 250mm thick, respectively, when exposed to ISO 834. Within the design period of


Fig. 11. E!ect of so$t protection.

90min "re resistance it is clear that the mid-span de#ections are a!ected to a negli-gible e!ect by the imposed load. The mid-span de#ection of the 180mm thick Ronanpoint slabs exposed to ISO 834 also demonstrated the negligible e!ect of the1.5 kN/m� imposed load although it should be noted that the design imposed load isnot known for the Ronan point building. Hence the de#ections are dominated bythermal bowing. This is an important "nding as it shows that a calculation of thermalbowing would su$ce in estimating the likely total de#ections before the onset ofcollapse.

2.2.4. Ewect of heating rateFig. 10 shows the e!ect of exposing non-loaded NWC slabs of 150 and 250mm

thickness, respectively, to the NPD (hydrocarbon) temperature}time curve which hasa higher heating rate than ISO 834. The NPD exposure results in much largerde#ections, especially in the early stage of exposure: at 20min the de#ections werealmost doubled in the 250mm thick slab. Although both slabs were designed to havea 90min "re resistance for ISO 834 exposure, the slabs were able to resist collapseunder the more severe NPD exposure for the 90min period and this suggests thatthere is a large measure of safety associated with present UK "re safety design practicefor reinforced concrete slabs.


Fig. 12. E!ect of slab thickness.

2.2.5. Ewect of sozt protectionFig. 11 shows de#ections of 150mm thick non-loaded slabs of NWC exposed to

ISO 834 with and without the two di!erent "re protecting boards. Details of theprotection are given earlier. The curves show that the addition of a plaster-based so$tgives much smaller de#ections provided the protection remains in place (part of theFireline fell down at approximately 55min). At the 90min design period of "reresistance the de#ection of the slab with GRG protection was roughly a quarter ofthat of the unprotected slab.

2.3. Axial deyections

Axial de#ection is here de"ned as the horizontal de#ection of one end of a slabrelative to the other end. All the measured de#ections were made at the mid depth ofthe slab, Fig. 1. A positive axial de#ection corresponds to an increase in the chordlength as occurs in the early stage of "re exposure due to thermal expansion. Later in



the "re exposure excessive bowing leads to a shortening of the chord length anda negative axial de#ection is obtained. Since the axial de#ection depends on themid-span de#ection in the later stages of "re exposure, an understanding of the axialde#ection curves requires reference to the corresponding mid-span de#ection curves.Measured axial de#ections are given in Figs. 12}15. Some general comments on theresults are as follows:

Fig. 12 shows axial de#ections for 150 and 250mm thick, non-loaded slabs of LWCexposed to ISO 834. The axial de#ection is larger for the thicker slab, and the time ofmaximum de#ection is di!erent, being delayed for the thicker slab. At 90min thede#ection was 5.5 and 9mm for the 150 and 250mm slabs, respectively. Maximumde#ections for 150 and 250mm thick, non-loaded slabs of NWC exposed to NPDwere 5.5 and 14mm, respectively.

Fig. 13 shows axial de#ections of the non-loaded and loaded 180mm thick Ronanpoint slabs and 150mm slabs of NWC exposed to ISO 834. The runaway axialde#ections of the Ronan point slabs are attributed to the runaway mid-span de#ec-tions. Fig. 14 shows, more than in any other test, the negligible e!ect of imposed load



on axial de#ection, and this can be attributed to the almost identical mid-spande#ection curves.

Fig. 14 shows that the largest axial de#ection of all the tests occurred with the250mm NWC slab exposed to ISO 834*the maximum positive de#ection was14mm. In contrast, the 150mm NWC slab exposed to NPD exhibited an earlymaximum positive axial de#ection of only 4mm followed by a reversal and largenegative de#ections. The "gure also shows the large e!ect of the rapid heatingachieved in the NPD exposure for 150mm thick NWC slabs.

Fig. 15 shows that the axial de#ection of the so$t-protected 150m NWC slab isapproximately three-quarters that of the unprotected slab at the 90min design periodof ISO 834 "re exposure.


Fig. 15. E!ect of so$t protection.

3. Conclusions

(1) A series of seven furnace tests has been successfully carried out giving accuratedata on mid-span de#ection and axial de#ection for 14 unrestrained reinforcedconcrete #oor slabs having a "re exposed length of 4000mm. The e!ect ofvarying the slab thickness, imposed load, heating rate, concrete type and so$tprotection has been established. These data enable numerical models to bevalidated.

(2) The method of test whereby two slabs nominally 900mm metre wide structurallyindependent of each other are exposed to the same heat #ux means that accuratecomparisons can be made. The method also means that specimens can easily bemanufactured, handled and transported, and the cost of testing is halved which isan important consideration for parametric research studies of the kind reportedin this paper.

(3) Mid-span de#ections were dominated by thermal bowing during the 90mindesign period of "re exposure; the e!ect on de#ections of imposing the design liveload of 1.5 kN/m�, Fig. 9, was very small and this suggests that BS 8110 isconservative.

(4) The higher rate of heating associated with the NPD hydrocarbon "re exposurecaused almost a doubling of mid-span de#ection obtained using the ISO 834 "reexposure in the "rst 20min, Fig. 10.


(5) The e!ect of using lightweight concrete employing Lytag coarse aggregate is toreduce the mid-span de#ection associated with normal weight concrete by rough-ly 30% at the 90min design period of "re resistance, Fig. 8. The reducedde#ections are probably due to the lower coe$cient of thermal expansion andlower thermal conductivity of light weight concrete.

(6) The e!ect of so$t protection is, as one might expect, to reduce the mid-spande#ection. Fig. 11 shows that the mid-span de#ection of the 150mm NWC slabwith so$t protection was roughly a quarter that of the unprotected slab. Fig. 15,however, shows that the slab having a so$t protection which remained in placesu!ers almost the same magnitude of axial de#ection at mid-slab depth as thesame slab with no so$t protection, although the peak de#ection occurred muchlater for the so$t-protected slab.

Acknowledgements

The author wishes to thank Professor G. Cox, Fire Research Station, BuildingResearch Establishment, for permission to publish the test results and Mr. R.L.Sawford and his team in the Civil Engineering Laboratory, BRE, Cardington formaking the precast concrete test specimens. Thanks are also due to British GypsumLtd for supplying and installing, free of charge, two so$t board protection systems.Again thanks are due to laboratory sta! of Warrington Fire Research Centre whoworked closely with the author in the preparation for and conduct of the tests. Theprovision of funds from the Construction Directorate of the Department of theEnvironment is gratefully acknowledged.

References

[1] Cooke GME, Morgan PBE. Thermal bowing in "re and how it a!ects building design. BREInformation Paper IP 21/88, Building Research Establishment, December 1988.

[2] BS 8110 Structural use of concrete, Part 1 Code of practice for design and construction. BritishStandards Institution, 1985.

[3] Morris WA, Read REH, Cooke GME. Guidelines for the construction of "re resisting structuralelements. Building Research Establishment Report BR 128, BRE, Garston, 1988.

[4] Holmes M, Anchor RD, Cooke GME, Crook RN. The e!ects of elevated temperatures on the strengthof reinforcing and prestressing steels. The Structural Engineer 1982;60(1):7}13.

[5] British Standards Institution, ENV 1993: Design of steel structures: DD ENV 1993-1-2: Structural "redesign (including UK NAD). Expected 2001.


behaviour of precast concrete floor slabs exposed to standar

Documents