Proceedings of DETC `04 ASME 2004 Design Engineering Technical Conferences and

Computers and Information in Engineering Conference 30th Design Automation Conference

Salt Lake City, Utah, USA, September 28, October 2, 2004

DETC2004-57082

AN OPTIMIZATION STUDY OF THE ULTRASONIC WELDING OF THIN FILM POLYMERS

Alejandro A. Espinoza Orías∗, John E. Renaud‡ Department of Aerospace and Mechanical Engineering

University of Notre Dame Notre Dame, IN 46556 Email: jrenaud@nd.edu

∗ Graduate Research Assistant, ASME Student Member ‡ Professor, ASME Fellow

ABSTRACT In manufacturing industries, ultrasonic welding has

established itself as one of the most effective techniques for fusing plastic assemblies due to its rapid performance and the absence of filler material. In this report, a thermoplastic polyurethane prototype of an orthopedics device that requires a hermetic seal is joined using ultrasonic welding. A robust design approach is used to study the manufacturing control factors that influence the process. The welding process factors and their interactions are used to characterize the resulting seal. Burst testing is used to assess weld strength. Optical microscopy in addition to SEM images, are used for a qualitative evaluation of the welded joint. Optimum process parameter settings from the robust design study deliver a strong and leak-proof weld.

Keywords: Design of Experiments, Plastics welding, Ultrasonic Welding, Hermetic Seals, Thin Films, Burst Testing.

NOMENCLATURE Aconv Converter amplitude. Aoutput Output amplitude. DoE Design of Experiments. Gbooster Booster gain. Ghorn Horn gain. HAZ Heat Affected Zone. N Number of replications. MSD Mean Square Deviation. Pb Burst pressure (test output). Pb

max Highest burst pressure. Pb

mean Mean burst pressure. PI

mean Mean burst pressure from matrix I. PII/III

mean Mean burst pressure from matrices II and III. P Welding Pressure.

t Film thickness. Tg Glass transition temperature. Tm Melting temperature. Th Holding time. Tw Welding time. TPU Thermoplastic Polyurethane Yn n-th experimental result. Ymean Average of all the recorded yn for each level.

Y Average for the three levels representing each factor.

η Signal to noise ratio.

INTRODUCTION Orthopedics has grown tremendously over the last half of

the twentieth century and into this century and its overall goal is to improve the quality of life for the aging and those affected by disease and trauma. For example, techniques using bone graft to repair or replace bone tissue in cases of accidents or corrective surgeries benefit from research. More than 150,000 total hip replacement surgeries are performed in the U.S. every year [1]. Bone grafts have reached 500,000 procedures annually and are rising [2]. The Osteoporosis and Related Bone Disease National Resource Center of the National Institutes of Health, estimates that osteoporosis costs Americans 38 million U.S. dollars every day, based on data from nursing homes and hospitals. Older people are frail and cannot withstand the trauma of surgery as well as younger persons, therefore novel ways to perform the surgery must be developed.

Two types of hip surgical interventions are practical: the hip fixation, which attempts to recover normal function of a hip after fracture and a total hip replacement where the greater trochanter region of the femur is completely removed and a

1 Copyright © 2004 by ASME

new acetabular cup is attached to the pelvis of the patient. Both types are shown in Figure 1.

Hip replacement surgery techniques and instrumentation are a growing sector for research in orthopedics. In a research partnership with a local orthopedics company, Notre Dame Biomaterials and Biomechanics Group (www.nd.edu/~amebio) is developing a minimally invasive hip replacement technique that among the new improvements it introduces, involves the sealing of a thin polymer film device. This research is part of this new product development effort.

Figure 1 LEFT: STAKE TYPE TOTAL HIP REPLACEMENT IMPLANT. RIGHT: HIP FIXATION DEVICES POSITIONED

INSIDE A FEMUR. This new thin film sealing application requires a sealing

process to produce a leak-proof seal. Among many current plastics joining processes, ultrasonic welding was chosen due to its swift operation, absence of a filler material and a proven track record as a serial manufacturing method. Due to the use of proprietary shapes and designs at this product development stage, a simplified test geometry was employed in this investigation to perform an initial study of the strength characteristics of the ultrasonic weld. Two square sheets of thin film thermoplastic polyurethane (TPU) were joined with a hollow circular sonotrode. This produced a circular pouch, which was cut in half to obtain two burst pressure test specimens.

In order to find the best conditions and parameters for the sealing procedure, a robust design approach was used to assess the influence of the identified on-line factors associated with this process. Our goal in this work is to perform a manufacturing process optimization of an effective thin polymer film joining technique. The robustness of this welding procedure will be dictated by proper selection of the manufacturing process parameters. To achieve this in a scientific and repeatable manner, statistical methods currently used in industry, namely the Taguchi robust design methodology, will be used to evaluate the influence of the factors affecting the welding process. The goal is to identify the proper combination of factors that yield the strongest and best weld attainable using ultrasonic welding. As such, the identified parameters used in this investigation were three, namely: welding time tw, holding time th, and welding pressure P.

MATERIALS AND METHODS

FUSION WELDING OF POLYMERS The term fusion welding applies to all techniques employed to join polymer parts where the fusion and solidification takes place at the joining interface. The majority of these methods do not require the intervention of a filler material. Widespread fusion welding techniques make use of a specific means to heat the joining interface to a temperature that allows melting. Subsequently, by holding the pieces together the fusion of the parts is completed.

ULTRASONIC WELDING

The use of ultrasound is old enough for it to claim to be a mature technology. The first ultrasonic welding was performed in the mid 1960s. Several landmark papers, such as the one by Benatar et al. [3] investigated the effects of near field and far field ultrasonic welding on various types of thermoplastics. Among others, Benatar [3], Michaeli and Korte [4], Van Wijk et al. [5], Liu et al. [6], and Volkov [7] also studied the maximization of welding strength of solid parts. However, few references to the study of near field welding of pouches and polymer film applications are found in the literature. A recent industrial type description of ultrasonics applications to packaging can be found in Herrmann and Lynch [8]. Their account illustrates ultrasonics use in food and pharmaceuticals industries.

Ultrasonic energy is applied to the parts in the process and such energy generates enough heat to melt and fuse the parts together. The most conventional way to achieve successful conveyance of the sonic energy is to employ the 20, 30 or 40 kHz frequencies. Figure 2 explains the process details as it is applied to a part being welded. To weld an assembly, sufficient energy is required to melt the interface. Once the molten state of the interface is achieved, the adjoining parts are kept tightly pressed until solidification, producing thus a consistent and uniform weld. The hardware arrangement known as the stack conveys the energy to the interface and includes (in order), first the converter, which is connected to the booster, which is in turn attached to the last component (in contact with the part to be welded): the acoustic tool or sonotrode.

Figure 2 THE ULTRASONIC WELDING PROCESS [9].

2 Copyright © 2004 by ASME

CIRCULAR POUCHES To weld a closed contour, a circular seal was employed. This ensures the simplest geometry that can be used in this case. Thermoplastic Polyurethane (TPU) was the material to be welded. Thickness values ranging from 0.254 mm to 1.016 mm were tested. Squares of 127 mm per side were welded one on top of another. A circular pouch or blister of 82.55 mm diameter was produced when two square sheets were sealed together.

Figure 3 SCHEMATIC DEPICTING THE WELDED PARTS

BEFORE AND AFTER THE PROCEDURE.

TOOLING The chosen horn was, a circular tool with an I.D. of 82.55 mm shaped as a hollow cylinder and made of titanium without hardening surface treatments. A 2kW Branson Ultrasonics IW 920+ welder was used to weld these pouches. The thin edge of the cutting anvil was machined to a 0.5 mm wide sharp edge. Since an energy director cannot be attached to thin films, a common practice is to build them into the holding fixture or anvil. A sectional drawing of the anvil can be seen in Figure 4. For testing purposes of the seal, a pneumatic apparatus was developed to conduct burst testing in a similar fashion to the ASTM F1140 standard [10].

Figure 4 CROSS SECTIONAL DETAILS FROM THE WELDING FIXTURE

WELDING FACTORS

The following are the controllable factors available in the Branson 920 IW+ ultrasonic welding station. Some of them were left as fixed settings so they were not included in the Design of Experiments approach.

Down Speed

It is the vertical velocity of the carriage as it approaches the part to be welded from its rest position. The slower, the better for applications involving thin films. Too fast of a velocity can impact the horn against the fixture. In this case, down speed was set at setting 1.5 from a scale of 0 to 10. The 0 setting does not allow movement of the carriage.

Amplitude

The most important factor is the welding amplitude. Modification is achieved through the use of the appropriate booster combination together with the gain associated to the horn. The approximate output amplitude of the ultrasonic wave it can be calculated knowing the average output amplitude of the converter Aconv (a value given by the manufacturer) multiplying it by the booster gain Gbooster, and finally multiplying that by the horn gain Ghorn using Equation 1:

Aoutput = Aconv × Gbooster × Ghorn Equation 1

Using Aconv = 17 µm, Gbooster = 1.5 booster and Ghorn, an output amplitude at the horn of 56.1 µm is found.

Welding Time

This factor sets the amount of time that the ultrasonic stack will be transmitting axial vibration to the part being welded. In this investigation, a range of 0.5 to 2.0 seconds was explored.

Holding Time

This factor sets the length of time force will be applied to the part while it undergoes re-solidification after being subjected to the ultrasonic energy. The appropriate holding time will prevent welding flash and will provide a uniform bonding. Holding time was studied in the range of 0.5 to 3 seconds.

Welding Pressure

The means for securing the part being welded is the pressure applied by the pneumatic carriage. In this particular welder, pressure is a factor set only once. Some more modern equipment can increase or decrease the pressure being applied according to the vertical collapse distance of the part. It works in conjunction with the holding time as it helps the fusion bonding process after the plastic has been softened or melted by the application of ultrasonic energy. For this investigation a wide range of pressures were used, ranging from 1.72 to 4.82 bar.

FEASIBLE FACTORS AND LEVELS

After several trial and error attempts, three particular combinations of factors were chosen. This was accomplished through patient observation and recurring testing of the values found. This being an exploratory investigation in ultrasonic

3 Copyright © 2004 by ASME

welding, it was necessary to outline the design space where acceptable and optimum designs can be found. The three levels found to work best are listed in table 1:

Table 1 CHOSEN OPERATING LEVELS FOR EACH FACTOR

BURST TESTING

The destructive testing evaluation method pursued in this investigation was burst testing. It proved to be a straightforward procedure that directly relates the weld quality to seal integrity. Tensile tests were not pursued because they do not describe the desired product property of maximum allowable burst pressure. The sealed pouches were cut in half, thus providing two semicircular test specimens (Fig. 6). A LabView® PC virtual instrument interface was developed to record the pressurization data and provide a real-time plot of the pressure vs. elapsed time curve.

Figure 5 IN-HOUSE DEVELOPED ELLIPTICAL SHAPE CLAMP INSERT. ARROW SHOWS MESHED AIR CONDUCT.

As for the connector, it was realized that the open-package testing mode from the F1140 standard could be adapted to the geometry of the samples being studied. Pressure regulation was achieved by the use of a Matheson® manual pressure regulator that in turn was connected to a pressure transducer sending an analog 0-5 V dc signal to the LabView® data logging application.

An original leak-proof clamping and connecting device was developed in-house to carry out the burst testing. The comparatively smaller size and different geometry of the present application required a special clamp developed specifically for it. The solution found consisted of an ellipse contour shaped clamp that holds an ellipsoid shape insert containing the air hose connector (Fig. 5). Special care was taken to conform this shape to the test sample in an opened but not strain causing position, i.e. making the opening small enough so that the ends of the welded strip were not subjected to tearing stresses that could influence the sample when it underwent inflation.

Diffusion of the incoming airflow was deemed important to avoid a direct jet impinging the weld. This was achieved by applying a wire mesh insert at the tip of the connector. This is also shown in Figure 5. A picture of the finished pressure clamp can be seen in Figure 6.

DESIGN OF EXPERIMENTS

Robust design provides a methodology for tuning manufacturing process parameters. This technique is of widespread acceptance in the manufacturing industry. It is referred as an “off-line” quality control method since it is normally applied off the production line. Its main feature is to achieve the maximum possible quality with minimum cost and maximum product performance. The low sensitivity to the effects of variability often found in production or operation environments is the main feature of a robust design. That is, even with adverse manufacturing conditions or uncontrollable operating conditions, it still is possible to achieve a robust product or a robust process.

Levels Weld Time Tw(s) Hold time Th (s) Pressure P (bar)

1 0.5 1.1 1.72

2 0.6 1.5 3.45

3 1.2 2.5 4.14

TTSS

TTCC

AA

Figure 6 TWO VIEWS OF THE IN-HOUSE DEVELOPED PRESSURE CLAMP. TC: TOP HALF CLAMP; TS: TEST

SPECIMEN; A: AIR INLET. The main tool employed by DoE is the orthogonal array.

The orthogonality property implies that for every level of a factor, all levels of the other factors occur the same number of times [11]. This unique feature of the arrays is used to organize the experiment such that the least number of combinations are run to gather the most information about the process or product. Each level of the orthogonal array is termed a level of the experiment. The number of levels n is denoted in the name of the array Ln. Arrays are constructed in several different ways and can include factors with two or three levels. Most of this information can be found in works by Phadke [12], Peace [11] and Condra [13], among others. A particular array can be chosen for a specific experiment according to the number of levels and factors to be studied. The number of columns in the array denotes the number of factors that can be studied at the same time. Sometimes, according to the linear graph corresponding to a given array, interactions between the main effect factors are examined and assigned to the corresponding column.

SIGNAL TO NOISE RATIO η

Quality engineering procedures use the concept of signal to noise ratios as a measure to gauge the effects of the studied factors and their interactions. It represents the ratio of the mean (the signal) to the standard deviation (noise) [14]. The expression to find η involves the calculation of a specific mean square deviation (MSD) for the kind of quality characteristic that is appropriate for the case in study. For this application, the larger the better quality characteristic is the suitable option, since the goal of the experiment is to maximize the weld strength.

4 Copyright © 2004 by ASME

The general expression for η in dB units is:

[dB] log10 MSD⋅−=η Equation 2

And for the larger-the better case MSD is given by:

nMSD nYYY 22

22

1

111 +++=

K Equation 3

Where Yn is the experimental result obtained for the n-th replication of the experimental run. The number of replications for each level is denoted by n. Response tables and plots are assembled with this information.

EXPERIMENTAL MATRIX

Three factors with three levels each were defined for study with the matrix experiment. An orthogonal array that facilitates the use of three, three-level factors is the L9 orthogonal array listed in Table 2, along with its corresponding linear graph in Figure 7.

Run Factor A Factor B Factor C Factor D

1 1 1 1 1

2 1 2 2 2

3 1 3 3 3

4 2 1 2 3

5 2 2 3 1

6 2 3 1 2

7 3 1 3 2

8 3 2 1 3

9 3 3 2 1

Table 2 TAGUCHI L9 (34) ORTHOGONAL ARRAY

FIGURE 7 LINEAR GRAPH FOR THE L9 ARRAY The first three columns were implemented in the matrix experiment. The linear graph indicates that factors A and B are the main effect factors, and columns C or D can be used for other factors or as in this case empty to gauge possible interactions between them. Interacting factors are in this case, factors A and B. The interaction is denoted by AxB and was assigned to column D since column C was used for the third factor. It was decided that in order to have more consistent information from the welding process, three matrix experiments were to be run. Table 3 lists all three combinations to produce the different experimental matrices.

This can be interpreted as running a full factorial experiment, however it is not quite the case, since it is about three different fractional factorial experiments that provide the most information possible in only nine runs per experiment. The column assignment, again, following the linear graph, used columns A and B for the main effect factors. Column C was given to a third factor and Column D was set to empty to gauge

the interactions between factors set in columns A and B. Result plots do indicate this situation. Eight replications were carried out for each level, completing a total sample size of 72 samples per matrix. The complete matrices are shown in tables 4 to 6.

Matrix No. Factor A Factor B Factor C Factor D

1 Tw Tw Th Empty

2 Th Th P Empty

3 P P Tw Empty

Table 3 COMBINATIONS OF FACTORS TO DEFINE

THE EXPERIMENTAL MATRICES

Run Welding time Tw Hold Time Th Pressure P Empty

1 1.2 1.1 1.72 1

2 1.2 1.5 3.45 2

3 1.2 2.5 4.14 3

4 0.6 1.1 3.45 3

5 0.6 1.5 4.14 1

6 0.6 2.5 1.72 2

7 0.5 1.1 4.14 2

8 0.5 1.5 1.72 3

9 0.5 2.5 3.45 1

Table 4 MATRIX FOR EXPERIMENT I

Run Welding time Tw Hold Time Th Pressure P Empty

1 1.2 1.1 1.72 1

2 1.2 1.5 3.45 2

3 1.2 2.5 4.14 3

4 0.6 1.1 3.45 3

5 0.6 1.5 4.14 1

6 0.6 2.5 1.72 2

7 0.5 1.1 4.14 2

8 0.5 1.5 1.72 3

9 0.5 2.5 3.45 1

Table 5 MATRIX FOR EXPERIMENT II

Run Welding time Tw Hold Time Th Pressure P Empty

1 1.2 1.1 1.72 1

2 1.2 1.5 3.45 2

3 1.2 2.5 4.14 3

4 0.6 1.1 3.45 3

5 0.6 1.5 4.14 1

6 0.6 2.5 1.72 2

7 0.5 1.1 4.14 2

8 0.5 1.5 1.72 3

9 0.5 2.5 3.45 1

Table 6 MATRIX FOR EXPERIMENT III

A C,D B

5 Copyright © 2004 by ASME

RESULTS

The DoE study provides results in the form of the collected data and the calculations of the mean values for each level, the MSD and η collected in a response table. The response tables for each matrix are shown next in table 7. Column ΣY is the cumulative sum of the quality characteristics of the level that comprises a specific factor. The number of replications differs from n in that n is the number of levels that contain the specific factor under study as opposed to the number of replications, that is the number of times the specific level has been repeatedly run. This number should be 1/n of the total sample. In this case 24 is one third of 72. Y and η are in this table the averages for the three levels representing each factor.

RESPONSE PLOTS Two types of response plots are provided. The first one is a response plot of all the factors studied in each experiment. The second kind examines the interactions between factors by lumping together individual response plots for each experiment matrix.

Figure 8 RESPONSE PLOTS FOR EACH INDIVIDUAL EXPERIMENT

Table 7 RESPONSE TABLES FOR EXPERIMENTS I, II AND

III, FROM RESPECTIVELY

6 Copyright © 2004 by ASME

INTERACTION PLOTS The interaction plots created from the data in the response tables are shown in Figures 9 to 11. These plots examine the interactions that were defined by the linear graph corresponding to the L9 array. In each case, the nomenclature AnBj of the bottom legend refers to the pair combination of the n-th level of factor A with the j-th level of factor B. The trends originally observed in the combined plots of Fig. 8 are somehow corroborated by the trends shown in the subsequent plots. That is to say, that the results are inclined towards the higher energy direction, namely, more weld time. Higher pressures are also favored in these results.

Tw x Th Interaction - Matrix I

24

25

26

27

28

29

Tw1 Tw2 Tw3

S/N

rat

io [

dB]

AnB1 AnB2 AnB3

Tw x P interaction - Matrix II

24

25

26

27

28

Tw1 Tw2 Tw3

S/N

Rat

io [

dB]

AnB1 AnB2 AnB3

Figure 9 INTERACTION PLOT FOR MATRIX I

Figure 10 INTERACTION PLOT FOR MATRIX II

OBSERVED FAILURE MODES

Two distinct burst modes were observed depending on the yielding and failure characteristics of the weld. Both produce a sudden burst, however the pathway leading to the burst event was different in each case. A first mode (from now on referred to as Mode I) was characterized by a small puncture or separation in a small circular arc at the moment of bursting. These samples did not undergo swelling nor did they sustain the severe plastic deformation of samples that failed under

mode II characteristics. Typically this small separation was achieved at the welded region, meaning insufficient or poor welding, even though the samples did not separate when pulled by hand. Most of the lower recorded burst pressure values were attained with Mode I. Weaker weld consistency and uneven seal integrity were typical attributes of this failure mode. A characteristic curve for Mode I is shown in figure 12.

Th x P Interaction - Matrix III

26.5

27.0

27.5

28.0

28.5

Th1 Th2 Th3

S/N

Rat

io [

dB]

AnB1 AnB2 AnB3

Figure 11 INTERACTION PLOT FOR MATRIX III

Failure Mode I

0

5

10

15

20

25

0 5 10 15 20 25Elapsed Time (sec)

Burs

t Pr

essu

re (p

si)

Figure 12 CHARACTERISTIC PRESSURE VS. ELAPSED TIME CURVE FOR FAILURE MODE I

Figure 13 SAMPLE SHOWING SMALL SEPARATION AND THE TRANSITION FROM A WELDED TO SEPARATED

REGIONS (MODE I)

7 Copyright © 2004 by ASME

Figure 14 CHARACTERISTIC DEFORMATION CURVE

EXHIBITING THE VISCOELASTIC BEHAVIOR OF TPU FILM

Conversely, Mode II (Figure 14) did achieve some of the highest recorded values. Viscoelastic behavior of the TPU material was observed in this mode since the film stretched up to an approximate three-fold increase in volume of the test bag. After a maximum pressure point was reached, severe swelling of the bag ensued with a reduction in pressure up to a maximum of 0.69 bar down from the maximum pressure value. The observed failure mode involved as in mode I, separation of the welded region. However it did not reach the same proportions. Since weld integrity in these cases was of a more robust nature, minimal separation was noticed as an initiation of the burst event, followed by a rupture of the material side not the weld itself. This follows the already known fact that in any weld, the weld itself attains higher strength than the immediate surroundings known as the heat affected zone.

Figure 15 BURST TEST SEQUENCE DEPICTING SEVERE

PLASTIC DEFORMATION AND EXCEPTIONAL WELD RESISTANCE (MODE II)

CHARACTERIZATION OF THE HEAT AFFECTED ZONE Failure Mode II

0

10

20

30

0 10 20 30 40 50 60

Elapsed Time (sec)

Burs

t Pr

essu

re (p

si)

Optical and electron microscopy were used in an attempt to assess qualitatively the HAZ as a way to verify fusion bonding of the two film sheets welded together. Samples were embedded in epoxy resin, and were prepared following standard metallographic polishing techniques. Some interesting features from the welding process could be identified such as which side compressive force was exerted upon. (Fig. 16)

WELD

Figure 16 MICROGRAPH OF THE WELDED REGION In figure 16 a TPU section is shown at 50x magnification.

The upper right edge in the picture is the top sheet that was in contact with the horn face and thus the indentation is a result of the compressive force exerted by the actuator stack. Some necking on the lower right end of the weld is also seen here. The lower left edge follows the shape of the thin edge fixture used to weld this sample. By inspection of the micrograph, a uniform weld appears to be obtained. The weld width is roughly the same as the film thickness.

Figure 17 SEM MICROGRAPHS OF A TPU WELD. TOP: 1500X, BOTTOM: 3500X. SCALE BAR IS 10 µM IN BOTH

CASES. WELD LINE FADES INTO A MORE HOMOGENEOUS WELD. (ARROWS)

8 Copyright © 2004 by ASME

FINDING THE OPTIMUM

The highest value of η serves as the criterion to choose the optimum. From this statement and by inspection of the response tables, two distinct optimum designs were selected:

• Optimum I (Matrix I): Tw1, Th1, P3.

• Optimum II (Matrices II and III): Tw2, Th1, P3.

These optima are the linear combination of the preferred settings observed in the response plots.

CONFIRMATION RUN A confirmation run is necessary to authenticate the optimum designs previously found. It consisted of four replications for each design. The burst results showed an increase in burst strength and the S/N ratio was improved.

• Optimum I (Matrix I): Tw1, Th1, P3.

Highest burst pressure Pbmax: 1.933 bar

Mean burst pressure Pbmean: 1.840 bar

Mean burst pressure from matrix I PImean: 1.651 bar

Mean signal to noise ratio η: 28.51 dB Mean η from matrix I: 27.38 dB

• Optimum II (Matrices II and III): Tw2, Th1, P3.

Highest burst pressure Pbmax: 1.933 bar

Mean burst pressure Pbmean: 1.839 bar

Mean burst pressure from matrices II and III PII/III

mean: 1,646 bar Mean signal to noise ratio η: 28.51 dB Mean η from matrices II and III: 27.35 dB

As can be seen, significant improvement from the original

configuration has been achieved. A noticeable increase in average pressure is observed. Note that the mean values of η for the two optima are similar. The behavior of the burst samples can be seen in the following two figures. Mode II failure is consistently repeated and in some cases, long yielding times were attained. From Figures 18 and 19, one observes that the optimum II exhibits longer burst times and therefore might be the preferred design. The somewhat noticeable difference in the pressure build-up slope can be attributed to the fact that the equipment was operated manually and lacked any automated pressure control device.

The application of the Robust Design methodology to this previously unknown process did achieve satisfactory results. These results can be measured first by the actual performance of the welds, and then by the successful application of the Taguchi technique to improve the welds. It is a preliminary study that could benefit from further development. The two optima had similar values of η but optimum II had longer burst times.

Figure 18 BURST RESULTS FROM CONFIRMATION RUN 1

Figure 19 BURST RESULTS FROM CONFIRMATION RUN 2

CONCLUSIONS The ability to weld thin film polyurethane using an ultrasonic sealing process has been explored in this investigation. In conventional thin film polymer ultrasonic welding, the ratio of weld width to film thickness is many times larger than unity. In this study, a circular sealed perimeter with a nominal weld width to film thickness ratio of unity has been achieved. An initial parameter study was conducted to find the factor settings used with the Taguchi robust design method. These initial designs were used as inputs to the robust design study of thin film polyurethane sealing.

The results of the robust design study indicated a particular trend, corroborated by the data analysis: that a better weld is obtained with a larger welding energy input. This situation is normally achieved by increasing the weld time or the ultrasonic wave amplitude. After completing this study input from ultrasonic sealing manufacturer has suggested that for this application, higher amplitude setting should be used. This would deliver a stronger weld, as the plastic layers would reach

9 Copyright © 2004 by ASME

a complete and more uniform fusion bonding. This should be investigated in future studies.

Burst testing served the purpose of providing a quantitative assessment of the welded joints. Seals that were obtained with higher energy settings did perform better. From two different identified failure modes, mode II provided the best results characterized by a strong weld. These features were also observed in the confirmation runs where both optima did perform under failure mode II, achieving better results than the initial experiments. It is of importance to emphasize that the robust design process was implemented without any previous knowledge of optimal process parameter settings. In the end satisfactory results were obtained and a more thorough understanding of the process was gained.

Qualitative evaluation of the welds by means of optical and scanning electron microscopy did not provide in-depth information due to the amorphous nature of the polymer used. However, these methods served to illustrate at a microscopic level the surface features of the weld. Optical microstructure characterization methods cannot be employed due to the absence of such constitution. Other more advanced material methods such as differential scanning calorimetry are best suited to analyze the materials subjected to the ultrasonic welding process and can offer more detailed information.

The robust design methodology proved a straightforward methodology suitable for manufacturing environments where the ultrasonic welding procedure is normally used. The small number of experiments required to characterize and optimize the process is a desirable characteristic for industry.

ACKNOWLEDGMENTS Support for this research has been provided by: the State of Indiana 21st Century Research and Technology Fund.

REFERENCES

[1] Ramakrishna S., Mayer J., Wintermantel E. and Leong K.W., “Biomedical applications of polymer-composite materials: a review”, Comp. Sci. and Tech. Vol. 61, Iss. 9, pp. 1189-1224, 2001

[2] Greenwald, A.S., Boden, S.D., Goldberg, V.M.; Khan, Y., Laurencin, C.T. and Rosier, R.N., “Bone-Graft Substitutes: Facts, Fictions, and Applications”, J. Bone Jt. Surg. Am., Suppl. 2, Part 2, Vol. 83, Iss. 11, pp. 98-103, 2001

[3] Benatar, A., Eswaran, R.V., Nayar, A.K., “Ultrasonic Welding of Thermoplastics in the Near-Field”, Polymer Engineering Science, December 1989, Vol. 29, No. 23, pp.1689-1698.

[4] Michaeli, W., Korte, W., “Quality Assurance in ultrasonic welding on the basis of statistical process models – prediction of weld strength”. Welding in the World, Vol. 35, Issue 5, September- October 1995, pp. 348-351.

[5] Van Wijk, H., Luiten, G.A., Van Engen, P.G., Nonhof, C.F., “Process Optimization of Ultrasonic Welding”, Polymer Engineering and Science, May 1996, Vol. 36, No. 9, pp. 1165-1176.

[6] Liu, S.J., Lin, W.F., Chang, B.C., Wu, G.M., “Optimizing the Joint Strength of Ultrasonically Welded

Thermoplastics”, Advances in Polymer Technology, Vol. 18, No.2, pp 125-135, 1999.

[7] Volkov, S.S., “Technological special features of ultrasound welding of polymer films”, Welding International, Vol. 15, Part 3, pp 243-248, 2001.

[8] Herrmann, T., Lynch, B., "Ultrasonic sealing of flexible pouches through contaminated sealing surfaces”, SPE ANTEC Conference Proceedings, 2003, 61st, Vol. 3, pp 2659-2635.

[9] Ageorges, C. and Ye, L., Fusion bonding of polymer composites (from basic mechanisms to process optimization), Springer, New York, 2002.

[10] American Society for Testing and Materials, F1140-00 Standard Test Method for Internal Pressurization Failure Resistance of Unrestrained Packages for Medical Applications, Book of Standards Vol. 15.09, West Conshohoken, Pennsylvania, 2001

[11] Peace, G.S., Taguchi Methods ® - A Hands-On Approach to Quality Engineering, Addison-Wesley Publishing Company, Reading, Massachusetts, 1993.

[12] Phadke, M.S., Quality Engineering Using Robust Design, Prentice Hall, Englewood Cliffs, New Jersey, 1989.

[13] Condra, L. W., Reliability Improvement with Design of Experiments, 2nd Edition, Marcel Dekker, New York, New York, 2001

[14] Wu. Y., Wu, A., Taguchi Methods for Robust Design, American Society of Mechanical Engineers, New York, New York, 2000.

10 Copyright © 2004 by ASME