UTTERWORTH ~] E I N E M A N N

Composites 26 (1995) 669-673 1995 Elsevier Science Limited Printed in Great Britain. All rights reserved 0010-4361/95/$10.00

Application of Taguchi method for process enhancement of on-line consolidation technique

S.K. Mazumdar and S.V. Hoa* Concordia Center for Composites, Department of Mechanical Engineering, Concordia University, Montreal, Quebec, Canada HG3 1MB (Received January 1994; revised 1 December 1994)
During laser assisted processing of thermoplastic composites, the individual effects of laser power, tape speed and consolidation pressure on the interply bond strength cannot be easily separated unless a large number of experiments is carried out. Because the properties of an end product depend upon the selection of processing conditions, it is essential that the conditions under which a well consolidated part is obtained should be defined. To investigate the effects of individual process parameters from a minimum number of tests, the Taguchi method is applied. For this investigation, PEEK/carbon fibre (APC-2) thermoplastic composite rings were manufactured by a thermoplastic tape winding process at selected conditions, and the influence of dominant process parameters on interply bond strength was evaluated. The influence of three dominant factors, namely laser power, consolidation pressure and tape speed, on bond strength were investigated. The experimental design involved using L9 orthogonal arrays, and the percentage contribution of each factor to the quality of bonding was estimated by the ANOVA technique.
(Keywords: composites manufacturing; Taguchi method; PEEK/carbon composites; on-line consolidation)

INTRODUCTION In the manufacture of a composite component, several processing variables influence the performance of the product. The technique of defining and investigating all possible conditions in an experiment involving multiple factors is known as the design of experiments. In the old days it was believed that the scientifically correct way to conduct an experiment was to vary just one factor at a time, holding everything else fixed. Thus for a full factorial design, the number of possible designs N is:
N = L m

(1)

where L = number of levels for each factor and m = number of factors. Thus, if the quality of a given product depends on three factors, A, B, and C, and each factor is to be tested at three levels, the equation (1) indicates 33 (27) possible design configurations. It means the experimenter has to conduct 27 tests to get enough information about the process. As the number of factors or number of levels increases, the total number of tests also increases. For example, if an engineer wants to determine the effect of seven factors at two levels then the total number of

* To whom correspondence should be addressed

experiments becomes 27 (128), which will increase the cost of the experiments. Techniques such as fractional factorial design are used to simplify the design of experiments. Fractional factorial design investigates only a fraction of all the possible combinations. This approach saves considerable time and money but requires rigorous mathematical treatment, both in the design of the experiment and in the analysis of the results, to correctly understand the process. Each experimenter may design a different set of fractional factorial experiments. Therefore, a systematic approach to determine the effect of process parameters on the quality of end product is helpful. Herein lies Taguchi's contribution to the science of the design of experiments. He simplified and standardized the fractional factorial designs using a special set of orthogonal arrays. According to Taguchi's experimental design, only a minimum of 8 experiments instead of 128 experiments for seven factors at two levels are required to get enough information about the process. Detailed explanations of Taguchi methodology can be found in refs 1-4. For manufacturing processes where a large number of factors influence the final outcome, the Taguchi approach can be utilized to arrive at the best parameters for the optimum design configuration with the least number of analytical investigations. Therefore the Taguchi method has great potential in the area of low

COMPOSITES Volume 26 Number 9 1995

669

Taguchi method for process enhancement: S.K. Mazumdar and S. V Hoa

cost composites manufacturing and materials processing. In spite of this, use of statistical techniques in the area of composites manufacturing is lacking. Wilkins et al. 5 used the Taguchi method for property and process enhancement of a resin transfer moulding (RTM) process. Recently several researchers have addressed various aspects of thermoplastic tape laying and tape winding processes. Beyeler and Guceri 6, Grove 7. Nejhad s and Anderson and Colton 9 have developed models for the prediction of temperature history for a known amount of heat supply and known tape speed for a tape laying operation. Other researchers 1-~4have studied the experimental aspects of thermoplastic tape winding. Beyeler et al. 1 discussed the feasibility of laser processing by producing several rings using Ryton AC 40-60 prepreg tapes. Hauber and his co-worker 11,12used robots and hot nitrogen gas for the manufacture of circular cylinders. Werdermann et al. 13 designed and fabricated an on-line consolidation facility for the manufacture of circular rings and short tubes for thermoplastic composites. They used infra-red and hot nitrogen gas for heating the preimpregnated tow. Carpenter and Colton 14 used hot air as a heat source for the fabrication of circular rings. The major emphasis in these works ~q4 was on the design and development of an experimental set-up and the determination of quality of consolidation by optical microscopic study. None of the above researchers 1-14experimentally determined the temperature history, process induced deformations, crystallinity or quality of the bond. Mazumdar and Hoa 15-18performed qualitative analyses of the effects of dominant process parameters, such as the amount of heat supply, tape speed and consolidation pressure, on the temperature history, process induced deformation, crystallinity and quality of consolidation for laser and hot nitrogen gas aided processing. Agarwa119 studied the effect of laser power on crystallinity, void content, temperature distribution and interlaminar bond quality for a constant tape speed of 15.4 mm s-l. The above works lack any quantitative analysis on the effect of process parameters on bond quality. This paper emphasizes the quantitative analysis of the influence of dominant process parameters on bond strength. The study provides important information and guidelines for the thermoplastic tape winding process, and determines the percentage contribution of individual parameters to the bond quality. OBJECTIVES The present study is performed to fulfil the following two objectives: 1. to use a statistical method for determining the optimum condition; and 2. to estimate the contribution of individual process parameters, such as laser power, consolidation pressure and tape speed, to the quality of the bond. EXPERIMENTAL P R O C E D U R E APC-2 tape (ICI Fiberite) with a 6.35 mm width and 0.125 mm thickness was used to manufacture 146 mm diameter rings using a thermoplastic tape winding technique. A schematic diagram for manufacturing an APC2 ring is shown in Figure 1. Photographs of the set-up

B e a m delivery

Aircylinder
Mandrel

~_~ I Laser Head

\

system

laminate Laser beam .1 ~ Incoming

tape

Consolidated

Roller
Contact point

Cooler
~ P o w e Cont r roller
Figure 1 Schematicdiagram of the laser assisted thermoplastic tape winding process are shown in Figure 2. A 65 W CO2 laser having a wavelength of 10.6/an (MPB Technologies) was used for processing the composites. The experimental set-up used for manufacturing the rings is described in detail in ref. 20. In the tape winding process the three dominant processing parameters are laser power, tape speed and consolidation pressure. By changing the laser power and tape speed, the temperature history and thus quality of the part can be changed. For the present investigation, 10-ply rings were manufactured at selected processing parameters. Mazumdar 2 characterized the interfacial ply bonding of ring speci-

l~gmre2 Photographs from back and front of the experimentalset-up

670

COMPOSITES Volume 26 Number 9 1995

Taguchi method for process enhancement: S.K. Mazumdar and S.V. Hoa
mens using a short beam shear (SBS) test, a double cantilever curved beam (DCCB) test and a fracture surface study by scanning electron microscopy (SEM). They found that the SBS test is sensitive to interply bonding and can easily detect differences in the interply bond quality. Therefore, in the present case SBS tests were conducted for quality control purposes. IMPLEMENTATION OF TAGUCHI METHOD 60 W (A3), 151.4 kN m -l (B3), 13.0 mm s-1 (C2) For the present case only one experiment at each of the above conditions was performed to determine the main effects o f individual processing parameters. The optimum condition is identified by studying the main effects of each of the factors.

Analysis of results Designing the experiment

Experimental design involves defining all the possible conditions in an experiment involving multiple factors. An experimental design must satisfy two'objectives. In the first, the number o f trials is calculated and in the second, the conditions for each trial are specified. Taguchi developed several sets of orthogonal arrays (OAs) for designing experiments with various factors and levels. In the present case three factors at three levels are studied, as listed in Table 1. Three levels are selected when it is suspected that the influence of a factor on the result can be non-linear. F o r the present case, an L9 om, as shown in Figure 3, will be suitable for experimental design I. There are nine independent conditions in an L9. These conditions are described by the numbers in the rows. Experiments were performed at laser powers of 40, 50 and 60 W, consolidation pressures of 50.4, 100.8 and 151.3kN m -l, and tape speeds o f 6.42, 13.0 and 27.0 m m s -~. Here consolidation pressure is measured in terms o f load per unit width o f the laminate, with the assumption that there is linear contact between the consolidation roller and laminate 2. F o r clarity, the experimental conditions o f Figure 3 can be explained as follows: Experiment No. 1: 40 W (AI), 50.4 kN m -1 (B0, 6.42 mm sl (Ct) Experiment No. 2: 40 W (A0, 100.8 k N m -1 (B2), 13.0 mm s t (C2) Experiment No. 3: 40 W (A0, 151.3 k N m -1 (B3), 27.0 mm s 1 (C3) Experiment No. 4: 50 W (A2), 50.4 kN m -t (B0, 13.0 mm s-1 ((:2) Expertment No. 5: 50 W (A2), 100.8 kN m -t (B2), 27.0 mm s-t (C3) Experiment No. 6: 50 W (A2), 151.4 kN m -1 (B3), 6.42 m m s l (CO Experiment No. 7: 60 W (A3), 50.4 k N m -1 (Bl), 27.0 mm s"l ((73) Experiment No. 8: 60 W (A3), 100.8 k N m -l (B2), 6.42 m m s a (Ci) Experiment No. 9:
T ~ l e 1 Control factors and their levels for thermoplastic tape winding process Factor Laser power (W) Consolidation pressure (kN m 1) Tape speed (mm "l) Level 1 40 50.4 6.42 Level 2 50 100.8 13.0 ~ Level 3 60 151.3 27.0 Fig~e 3 An experimental lay-out using L9 array

As described above, tings were manufactured at specified conditions and the results of the SBS tests, in terms of a quality characteristic Y, were measured as shown below: Yl -- 25.66 MPa, Y4 35.09 MPa, Y7 -- 35.66 MPa,
TM

Y2 = 28.06 MPa Y5 = 29.79 MPa Y8 = 35.73 MPa

Y3 = 20.65 MPa Y6 = 29.15 MPa Y9 = 39.57 MPa

These results were recorded in the far tight column of the OA (Figure 3). Since there was only one test for each condition, the results were recorded in one column. For some trial conditions such as experiment numbers 4 and 7, multiple tests were performed. Experiments 4 and 7 were selected for multiple runs because they were close to the optimum conditions. The standard deviation for experiment number 4 with two test runs was found to be 0.564 and that for experiment number 7 with three test runs was found to be 1.773. Once the main effects are known, then new levels for control factors are selected to locate the best condition for higher performance: The results for the new levels are presented in ref. 20. T o speed up analysis, Taguchi has provided some key procedures which are used here. When these steps are strictly followed by different individuals performing the analysis, they are likely to arrive at the same conclusions.

Computation of average performance

To compute the average performance of factor A at level 1, i.e. At at 40 W laser power, we add the results for trials including factor A~, and then divide by the number of such trials. For A~, we look in the column for A and find that level 1 occurs in experiments numbers 1, 2 and 3. The average effect of At is therefore calculated by adding the results Y of these trials as follows:

ors Trials ~1

A Laser Power

B Pressure

C Tape Speed 1 2 3 2 3 1 3 1 2

Y SBS t e s t results (MPa)

25.66 28.00 20.65 35.09 29.79 29.15 35.66 35.73 39.57

COMPOSIJ'ESMolume26Number 9 1995

671

Taguchi method for process enhancement: S.K. Mazumdar and S.V. Hoa
Alav
=

(]I1

]I2 +

}'3)/3

= (25.66 + 28.00 + 20.65)/3

Quality characteristics
For a product obtained by thermoplastic tape winding, the bigger the value of bond strength is, the better the quality of the product. Thus from Figure 4 the A3 (60 W), Bl (50.4 kN m -1) and C2 (13.0 mm s-l) combination is likely to produce the best result and therefore represents the optimum condition for the present case. From Figure 4 it can be observed that the influence of tape speed on bond strength is non-linear. The exact trend of the curve is not known. More tests at 60 W laser power and at tape speeds in the range 13.0-27.0 mm s-l are needed to locate the optimum conditions. It is obvious from the results of Figure 4 that to further improve the quality of laser p~ocessed parts, one should study the effects of laser power above 60 W and tape speeds higher than 13.0 mm s-l. In the present case, tests are not conducted at a laser power higher than 60 W because of the limitations of the equipment. This information is very useful in deciding about further sets of experiments for obtaining higher performance.

= 24.77 The average effects of other factors are computed in a similar manner:
A2av : (Y4 + Y5 + Y6)/3 : 31.34

A3av = B~av = B2av = B3av = Clay = C2av = C3av =

(Y7 + (Y~ + (Y2 + (Y3 + (Yl + (Y2 + (Y3 +

Y8 + Y4 + }'5 + Y6 + Y6 + Y4 + Ys +

Y9)/3 = Y7)/3 = Y8)/3 = Y9)/3 = Y8)/3 = Y9)/3 = Y7)/3 =

36.99 32.14 31.17 29.79 30.18 34122 28.70

The above values are plotted in Figure 4 to show the main effects of each factor on the quality of the bond. It is clear from the figure that the increase in laser power increases the bond strength, whereas effect of consolidation pressure in the range 50.4-151.3 kN m -! is negligible. The influence of tape speed on bond strength is found to be non-linear, and the quality of interlaminar properties is found to be poor at lower and higher tape speeds. The reason for the increase in interply bond strength with the increase in laser power is that the higher laser power causes a higher temperature at the consolidation point. Higher processing temperatures would result in lower polymer viscosity and higher intermolecular diffusion, which will cause a greater degree of resin flow and molecular interdiffusion, and will result in better interply bond properties 18'2. The effect of variation in consolidation pressure in the range 50.4-151.3 kN m -1 on bond strength is negligible. The reason for this could be that the degree of intimate contact is not affected in this pressure range 2. Decreasing the consolidation pressure to 25 kN m ~ decreases the degree of intimate contact and thus affects the bond strength 2. Because there may be sufficient wetting between plies in the consolidation pressure range 50.4-151.3 kN m ~, variation in interply bond strength is found to be minimal. The effect of tape speed on the SBS test results is found to be non-linear. Mazumdar 2divided the tape speed into three categories; low speed range (thermoplastic degrades), medium speed range (good consolidation) and high speed range (poor wetting). Because of polymer degradation at lower speed and insufficient wetting at higher speed, interply bond strength is reduced 2.

Relative contributions of variables

The relative contribution of processing parameters is established by comparing their variances. The analysis of variance (ANOVA) technique is used for this purpose. Calculations of different terms of ANOVA are omitted here. For a detailed study on ANOVA refer to ref. 1. Results of ANOVA in terms of per cent contributions of each process parameter are expressed as: PA = 79.08%, PB = 2.72% Pc = 17.05%

40

35

/
/ 4

--o-- __~

o/

rn 20
I A1 I A2 I A3 I B1 t B2 I B3 I C1 I C2 I C3

Figere 4

Main effects of factors on SBS test results

where PA, PB and Pc are the per cent contribution by laser power, consolidation pressure and tape speed respectively. The results show that the laser power contributes 79.08% to the development of bond strength, and therefore laser power should be controlled properly. The contribution of consolidation pressure to the development of bond strength is only 2.72%, and therefore variations in consolidation pressure during processing would not significantly affect the quality of bond. The influence of tape speed during on-line consolidation is found to be 17.05%. In.most cases variation in the quality of a product is unavoidable because of variation in process parameters during manufacturing. This variation can be brought to a minimum by understanding the effects of process parameters. In the present case, laser power has the highest influence on the variation of the properties in a tapewound product. Therefore during thermoplastic tape winding, the variation in laser power should be minimized to achieve consistent quality of the end product. The present results are found to be very helpful in designing processing equipment for manufacturing complex composite components by thermoplastic tape winding21. In the fabrication of non-axisymmetric shapes, tape speed varies for a constant mandrel speed. For example, during manufacturing of rings of elliptical cross-sections with a semi-major axis of 7.8 cm and a semi-minor axis of 3.9 cm, it is found that the tape speed varies from 2 to 16.2 cm s1 during one revolution of the mandrel at a constant mandrel rotation of 10 rev min l (ref. 21). With the change in tape speed, laser power cannot be kept constant for uniform bonding between

672

COMPOSITES Volume 26 Number 9 1995

Taguchi method for process enhancement: S.K. Mazumdar and S.V. Hoa
layers. To solve this problem, either the laser power should be varied to compensate for changes in tape speed or the tape speed should be held constant to prevent the variation in laser power. Present results show that the laser power should be kept at a maximum level (60 W in this case) for higher bond quality. Reduction in laser power will decrease the bond strength. Therefore in the production of complex shapes such as elliptical rings by the tape winding technique, variation in laser power should be prevented and tape speed should remain constant at an optimum value. Mazumdar and Hoa 21 have designed and developed a thermoplastic tape winding machine for the fabrication of non-axisymmettic composite parts using a laser as the heating source. During the manufacture of non-axisymmetric parts, mandrel speed was varied using a computer controlled motor to achieve a constant tape feed rate. Calculation of mandrel speed for achieving constant tape speed was performed using analytical models21. to determine the per cent contribution of each process parameter on the quality of bond. The optimum condition for the tape winding process was thus estimated. It was found that bond strength increases with an increase in laser power. The effect of consolidation pressure on bond strength was found to be negligible. The influence of tape speed on the quality of bond was found to be non-linear, with the highest strength at a medium tape speed. It was estimated that laser power has a major contribution to the improvement of bond quality. The per cent contributions to the enhancement of bond quality of laser power, consolidation pressure and tape speed were found to be 79.08, 2.72 and 17.05%, respectively. REFERENCES
1 Roy, R.K. 'A Primer on the Taguchi Method', VNR Publishers, New York, 1990 2 Taguchi, G. 'Introduction to Quality Engineering', UNIPUB, Kraus International Publications, New York, 1986 3 Barker, T.B. 'Engineering Quality by Design: Interpreting the Taguchi Approach', Marcel Dekker, Inc., New York, 1990 4 Box, G.E.P. and Bisgaard, S. Quality Progress June 1987, 54 5 Wilkins, D.J., Karbhari, V.M. and Steenkamer, D.A. Proc. American Society for Composites, 1991, p. 843 6 Beyeler, E.P., and Guceri, S.I.J. Heat Transfer 1988, 110, 424 7 Grove, S.M. Composites 1988, 19, 367 8 Ghasemi Nejhad, M.N. 'Three dimensional thermal and residual stress analysis of in-situ thermoplastic composite filament winding' PhD Thesis, Department of Mechanical Engineering, University of Delaware, Newark, USA, 1992 9 Anderson, B.J., and Colton, J.S.J. Compos. Mater. 1990, 24, 150 l0 Beyeler, E., Philips, W. and Guceri, S.I.J. Thermoplastic Compos. Mater. 1988, 1, 107 11 Hauber, D.E. and Hill, L.A. 'Automated fiber placement of thermoplastic composites on complex shapes', Report EM90-658, Society of Manufacturing Engineers, Dearborn, MI, 1990 12 Hauber, D. 'Robotic filament winding with advanced thermoplastics', Report EM87-552, Society of Manufacturing Engineers, Dearborn, MI, 1987 13 Werdermann, C., Friedrich, K., Cirino, M. and Pipes, R.B. J. Thermoplastic Compos. Mater. 1989, 2, 293 14 Carpenter, C.E. and Colton, J.S. 'Proc. 38th Int. SAMPE Symposium', SAMPE, Anaheim, CA, May 1993, p. 205 15 Mazumdar, S.K. and Hoa, S.V. 'Proc. 38th Int. SAMPE Symposium', SAMPE, Anaheim, CA, May 1993, p. 189 16 Mazumdar, S.K. and Hoa, S.V. 'Proc. ICCM-9', University of Zaragoza, Spain and Woodhead Publishing, Cambridge, 1993 17 Mazumdar, S.K. and Hoa, S.V. Heat Mass Transfer Mater. Process. Manufacturing 1993, 26, 115 18 Mazumdar, S.K. and Hoa, S.V. ,L Thermoplastic Compos. Mater. in press 19 Agarwal, V. 'The role of molecular mobility in the consolidation and bonding of thermoplastic composite materials' PhD Dissertation, Materials Science Program, University of Delaware, Newark, USA, 1991 20 Mazumdar, S.K. 'Automated manufacturing of composite components by thermoplastic tape winding and filament winding' PhD Thesis, Department of Mechanical Engineering, Concordia University, Montreal, Canada, 1994 21 Mazumdar, S.K. and Hoa, S.V. Mater. Manufacturing Process. 1995, Ill(D, 47

Projection of optimum performance

From the above analysis, A3 Bt C2 is found to be the optimum condition for interply bond strength. The actual result for the above calculation is not known because it was not among the trial runs performed for the present case. Prediction of the result at the optimum condition is determined by the following relation1:
Yopt =

T/N +

(A3av -

T/N) +

(Blav -

T/N)
(2)

+ ( C 2 a v - T/N)

= average performance + contributions of A3, BL and C2 above average performance, where T -- grand total of all results; N = total number of results; and Yopt -- performance at optimum condition. In this case: T = 279.3, N = 9, A3av = 36.99, Blav = 32.14, C2av= 34.22 and therefore
ropt z

31.03 + ( 3 6 . 9 9 - 31.03) + (32.14 - 31.03)

+ (34.22 - 31.03) = 41.28 which is slightly above the value of 39.57 MPa obtained during trial run 9. CONCLUSIONS The Taguchi method was applied for the experimental design of a thermoplastic tape winding process. Using the Taguchi method, only nine experiments were needed

COMPOSITES Volume 26 Number 9 1995

673