Comparison of injection molding machine performance.

INTRODUCTION

The injection molding process is dependent on the repeatable production of parts to a required specification and quality. Some amount of variation is inherent but this can be minimized for any given set of polymer, mold, and processing conditions. Understanding and quantifying variation can improve process optimization and quality control. With the development of increasingly complex molding processes such as two-component, fluid assisted, and micro-scale molding, control and repeatability have become even more important in the successful production of high quality parts.

Several factors may affect the repeatability of the injection molding process, including processor control technology (ability to control screw position, velocity, etc.), melt temperature control, mold temperature control, material consistency, and batch variation. The direct influence of these factors is clear; the machine's drive technology and control system determines the accuracy of screw rotation, position, and injection velocity, therefore controlling the melt preparation, injection and packing phases of the cycle [1, 2]. Barrel and nozzle temperature controllers, through the use of closed-loop proportional integral derivative (PID), aim to ensure polymer is maintained at a consistent temperature from cycle to cycle. Mold temperature control influences part quality during cooling and solidification [3, 4] which can affect crystallization and part strength, and cause shrinkage or warpage. Small variation between batches of nominally identical material have been found to induce significant variations [5], especially when the molding process is operating in a narrow processing window.

Other factors can affect the consistency of a molding process. Servohydraulic or electric injection molding machines perform the same task with different operating characteristics [6]. Electric machines are often thought of as more precise than hydraulic systems and have been shown to be more energy efficient, although the increased flexibility of a hydraulic machine may lead to a more consistent product, and hydraulic machines currently cost significantly less than their electric counterparts [7]. The performance of hydraulic and electric molding machines has been investigated in detail here, with particular attention paid to energy consumption and efficiency.

The age of a machine may also affect its accuracy and repeatability; developments in control technology have led to increasingly sophisticated microprocessors being used in modern machines. Improvements in control technology have led to more accurate control of screw position, velocity, and rotation and therefore improvements in part reproducibility. Wear of mechanical parts can have a detrimental effect on process performance, for example on the screw or back-flow valve. This is more likely to cause a longer term effect on process variation, and can be monitored to determine when mechanical parts need replacing.

A period when all injection molding machines exhibit variation is during start-up [8]. This is due to gradual heating of various machine components (screw, mold, nozzle, hydraulic oil) during the first 10-100 cycles of a molding run to reach a "steady state" condition from cycle to cycle. Preheating the mold tool to operating temperatures can help to minimize start up effects but, due to energy input during plasticization, some transient thermal effects are inevitable. If the causes of start-up effects can be identified and quantified, such as size of machine and type of polymer being molded, setup times may be reduced. At the onset of long production runs (i.e., several days) start-up effects are often dismissed as negligible when the amount of scrap produced in the first cycles are compared to the total amount produced.

To monitor variation, techniques such as statistical process control (SPC) can be used on molded products [9]. This is, however, a time consuming and inefficient method of quality control with a large lag time between detection of a fault and corrective action. Monitoring systems are now often integrated into molding machines and many machine control processors are equipped with monitoring software allowing real-time plots of various process parameters such as injection pressure, switchover position, etc. This information improves understanding of machine performance and variation, but may not relate directly to part quality. For this reason, a number of monitoring methods have been developed to measure more directly some aspects of the molding process, such as melt temperature and pressure [10-12], cavity pressure [13], and strain in the tie bars [14, 15]. These methods provide a great deal of useful information regarding the process, but are generally unsuitable or too expensive to be considered in a production environment, and links with product quality are not well quantified.

Aims

The aims of the work were to quantify the performance of 4 injection molding machines, varying in age and technology, during molding of a common part from the same polymer using a common tool. Using instrumented molding machines and a sophisticated data acquisition system, particular emphasis was given to examining start-up dynamics, repeatability, and power consumption.

EXPERIMENTAL

Equipment

Four molding machines were used in the trials and are listed below. Detailed specifications are displayed in Table 1.

Machine A: Cincinnati ACT30, servoelectric molding machine, 30 metric ton clamp force, microprocessor control (1989).

Machine B: Fanuc Roboshot, servoelectric molding machine, 50 metric ton clamp force, full microprocessor control (2000).

Machine C: Battenfeld CDK750, servohydraulic molding machine, 75 metric ton clamp force, full microprocessor control (1997).

Machine D: Sandretto Series 7, proportional-hydraulic molding machine, 60 metric ton clamp force, microprocessor control (1987).

The 4 molding machines provided a diverse sample group for performance comparisons. Two of the machines were servoelectric and two were hydraulically driven, allowing a direct comparison between the two technologies. Two machines were relatively modern at the time of the experiments (1 and 4 years old), whereas the other two represented older machine technology (12 and 14 years old) with more significant wear.

In each case, an interchangeable "cassette" tool insert allowed a common mold tool to be used on each machine. This allowed for a direct comparison of each machine's ability to mold the same part. A tensile test bar mold was used to produce tensile test specimens to standard EN ISO 527-2 [16]. A general purpose injection molding grade of high density polyethylene (BP Chemicals HD5050EA) was used throughout the experiments and the molded part is shown in Fig. 1. Mold temperature control was achieved using water continuously pumped through cooling channels in each half of the mold using a 6 kW Churchill Conair water temperature control unit. Molded part weight was measured using a Mettler BB244 balance and dimensions (length, width, and thickness) were measured using a specially constructed jig in which spring-loaded linear voltage displacement transducers (LVDTs) sent signals automatically to a PC. Tensile strength of the molded parts was measured using an Instron Series 9 tensile test machine with a 2 kN load cell.

[FIGURE 1 OMITTED]

Each molding machine was highly instrumented and fully computer monitored. During molding trials the following in-process measurements were made at a frequency of 50 Hz, using in-house monitoring software and Microlink 2000 data acquisition hardware. Melt pressure in the nozzle reservoir was measured using a Dynisco PT422 melt pressure transducer, hydraulic pressure (in the case of the 2 hydraulic machines) was measured using a Dynisco IDA 354 transducer, melt temperature in the nozzle reservoir was measured using a Dynisco MTX infrared sensor and tool temperature was monitored using a J-type thermocouple. Positions of the sensors relative to the nozzle are shown in Fig. 2. Process energy measurements were made using a Hioki 3-phase unbalanced loads energy meter with Hall effect clamps on the machine's electrical supply. This monitored power consumption and power factor ([lambda]), an indication of process efficiency. For a purely resistive load, the current in the circuit is exactly in phase with the voltage, [lambda] = cos(0) = 1, the power being used is the multiple of the voltage and current. For a purely inductive load, the current in the circuit lags the voltage by exactly 90[degrees], [lambda] = cos(90) = 0, and while current is being drawn from the supply, no useful work is being done. For a circuit containing a mixture of inductive and resistive loads, the power factor lies somewhere between 1 and 0, the exact value being determined by the circuit makeup. Averaged values were taken throughout each molding run and calculated as power consumption per hour.

[FIGURE 2 OMITTED]

Details

Experiments were carried out to simulate production from start-up of the molding process. In each case molding was monitored immediately on start-up until production of 400 parts. Machine settings were kept constant throughout the experiments. Prior to molding, the mold tool was heated overnight to minimize the time taken for the process to stabilize. Mold temperature control on each machine was achieved by a dedicated controller that pumped water continuously at a set temperature through the cooling channels of both halves of the mold.

Each machine was set up to perform an identical molding in terms of shot size, injection velocity, hold pressure, and packing and cooling times. Due to the differences in size across the 4 machines it was difficult to ensure that all aspects of the molding process were identical in each case. For example, differences in screw diameter affected the amount of energy input into the melt during plasticization. In this case, screw rotation speed for each machine was set to transfer the same specific energy per mass of melt. Also, due to the variation in capacity of each machine it was necessary to run some machines at opposite ends of their processing range (e.g., machine A at 78% of maximum injection velocity compared to machine C at 38%) to achieve a common volumetric throughput during injection. Melt temperatures were set at 210[degrees]C on each machine and mold temperature set at 40[degrees]C. Table 2 shows the full list of molding conditions. Midrange values were selected for packing pressure and back pressure. The high density polyethylene (HDPE) studied here was a general purpose injection molding grade, which at these conditions could be molded easily and had a large processing window.

All process measurements described above were monitored during the first 4 seconds of each injection cycle. Each molded part was labeled and allowed to cool after ejection. After 24 hours the parts were weighed (including sprue and runner sections) and the dimensions were measured. The final 10 molded parts of each run were subsequently mechanically tested to determine maximum tensile strength.

RESULTS

Machine Comparison

Mean results taken over the final 100 parts of each experimental run (selected to represent steady state molding conditions) are summarized in Table 3. Nozzle temperature rise was defined as the difference between melt temperature immediately prior to injection, and peak temperature during injection. Initial screw position was defined as the position of the screw immediately prior to injection. Mean part weight ranged from 15.51 to 15.86 g across the 4 machines, a range of 2.2%. This highlights the difficulty in producing duplicate molded parts using different machines, even at identical set conditions. Ranges of mean part dimensions between the machines were small; 0.48% in length, 1.76% in width, and 5.12% in thickness. Machine D (1987 proportional-hydraulic) produced molded parts with the highest mean part weight and length. The existing tensile bar mold had been designed to the dimensions of the required standard [16] without allowing for shrinkage; therefore, some of the measured width and thickness dimensions were below tolerance for the standard ([+ or -]0.2 mm on the width and thickness of the specimens). Machine B produced the smallest mean parts with 3.6% shrinkage in specimen width and 7.25% in thickness. Average shrinkage across all 4 machines was 2.5% in width and 4.8% in thickness.

[FIGURE 3 OMITTED]

Melt consistency was quantified by melt pressure and temperature measurements made in the reservoir prior to the injection nozzle. A difference of 12.5% was observed between the peak melt pressure during injection across the two hydraulic machines C and D, whereas both servoelectric machines had a similar peak injection pressure. Figure 3 shows a comparison of typical primary injection pressure profiles from each of the 4 machines. A significant difference in injection profiles was observed although injection velocities in each case were set identically. Injection velocities were found to deviate from set values. Machine A achieved the closest injection velocity to the set value (faster by 1.1%), machines B and C were 8.2% and 9.4% slower, respectively, and machine D was 37.3% slower than the value set.

From Table 3, maximum melt temperatures during injection were higher than set values by at least 10[degrees]C on all machines. A temperature rise of up to 21.4[degrees]C (machine B) occurred during injection, this being caused primarily by adiabatic heating of the melt reservoir during a pressure increase of 860 bar in 0.5 seconds. Melt temperatures during primary injection are displayed for all 4 machines in Fig. 4. These correlate strongly to the pressure profiles included in Fig. 3, reflecting temperature rise due to an increase in melt pressure. Mean mold temperature was similar for each machine, approximately 2[degrees]C above the set temperature of 40[degrees]C. Initial screw position was different for each machine in order to achieve the same volume of shot size, although for hydraulic machines the initial screw position was farther back than set due to bounce of the screw at the end of plasticization.

Tensile strength of the parts was measured for the final 10 moldings from each molding machine (i.e., part numbers 391-400) and shown in Fig. 5. Maximum tensile load ranged from 0.87 to 0.93 kN, a difference of 7%. Error bars based on 99% confidence intervals for each machine data set determined the combined level of variation from the tensile testing process and the cycle-to-cycle molding process variation. Because the calculated confidence intervals from each molding machine did not overlap, differences in tensile strength between molding machines was determined to be statistically significant. Maximum tensile load was found to correlate most closely with melt temperature during injection, as shown in Fig. 5. Tensile strength was observed to be decrease linearly with melt temperature, possibly related to cooling rate inside the cavity affecting crystallization. A relatively large cross-section gate was employed on the mold tool used here to minimize the amount of orientation in the molded tensile bars. This large gate may influence the effect of cooling on both part weight and tensile strength by allowing melt to be forced into the cavity during the packing phase. Mold tool temperature appeared to have no measurable effect on tensile strength, for the narrow range of mold temperatures measured here.

[FIGURE 4 OMITTED]

[FIGURE 5 OMITTED]

Start-Up Dynamics

To examine machine performance during the transient period at the beginning of each run, the production of parts 1-50 was defined as "start-up." During this period, all 4 molding machines exhibited transient behavior as the screw, mold tool, hydraulics, and mechanics gradually heated up to an equilibrium operating temperature. Figure 6 illustrates the change in part weight during start-up from servohydraulic machine C, with mold tool temperature. Decreasing part weight was observed over the start-up period, which appeared to be negatively correlated to tool temperature. This relationship was observed for all 4 machines as shown in Fig. 7: change in mold temperature appeared to be the most significant factor in start-up even though molds were maintained at operating temperature overnight before each trial. Machine A, the smallest studied, exhibited least variation in tool temperature during start-up. It appears that start-up transients are affected by machine size, and heat-up of the mold tools fitted a logarithmic model y = A.ln(x) + C as shown in Fig. 7. Table 4 shows the gradient and intercept for each logarithmic fit of mold temperature rise. Gradient appears to correspond to the capacity of the machine, most likely due to the volume of metal in the mold tool itself. It should be noted, however, that the logarithmic curve fits used here are only applicable to transient start-up, as they increase indefinitely with part number and therefore are unable to model the process at equilibrium. The magnitude of rise in mold tool temperature during the start-up period did not appear to correlate with variation in part weight, which was comparable across the 4 machines. However, machine A exhibited the quickest stabilization in part weight, which was defined as the number of cycles until 3 consecutive parts fell within [+ or -]1 standard deviation of the mean weight. For machines A, B, C, and D this stabilization period was 25, 27, 31, and 47 cycles, respectively. Figure 8 shows peak melt temperature during injection for all 4 machines during start-up. A gradual increase in melt temperature was observed, but this stabilized in fewer cycles than mold temperature.

[FIGURE 6 OMITTED]

[FIGURE 7 OMITTED]

[FIGURE 8 OMITTED]

Process Variation

Process variation was quantified by statistical analysis of the final 100 molded parts in each run (part numbers 301-400) during steady state molding conditions. Coefficients of variation in part and process variables are listed for each machine in Table 5, and Fig. 9 compares variation in both part and process measurements. Variation in part weight on all machines was small (maximum of 0.071%), although part weight from machine D (worst case) had a coefficient of variation over 4 times higher than that of machine C (best case). These variations indicate that machine age was a significant factor in part quality variation, the two older machines having highest variation in both part weight and part length. Variation in process measurements such as melt pressure, temperature and screw position reflected the improved repeatability of servoelectric machines over hydraulic. This can be partly explained by the ability to control screw position, as shown in Fig. 10. During the course of the sampling period the two hydraulic machines, C and D, exhibited significantly higher variation in initial screw position (i.e., screw position immediately prior to injection). Examples of variation during primary injection are shown in Figs. 11 and 12 for machines D and B, respectively, which represent the worst and best case. Injection melt pressure is overlaid for 10 cycles (parts 391-400) and coefficient of variation is plotted for the measured period of 4 seconds. Machine D, shown in Fig. 11, was observed to vary significantly over 10 cycles with a coefficient of variation of up to 15%. Over the same sample period, machine B (Fig. 12) exhibited negligible variation with a coefficient of variation of less than 0.2%.

Energy Consumption

Average power usage from each machine is displayed in Table 3 during steady-state molding. Both hydraulic machines consumed significantly more power (average of 5.58 kW) than the servoelectric machines (average of 1.55 kW), by a factor of 3.6. The age of the machine also appeared to affect power consumption. In both hydraulic and electric machines, newer machines consumed less energy; by 16% for hydraulic and 7% for servoelectric, despite the newer electric machine having a higher maximum clamp force and capacity.

[FIGURE 9 OMITTED]

[FIGURE 10 OMITTED]

The increased amount of energy used to run hydraulic injection molding machines is largely due to the continuous inductive load required for the hydraulic motors. Inductive load also reduces the efficiency of hydraulic machines, as indicated in by the power factor ([lambda]) values displayed in Table 3. Machine B operated with a power factor of [lambda] = 0.98 (i.e., very efficient). The older servoelectric machine, A, was less efficient ([lambda] = 0.70) because of older technology, higher inductance servomotors.

Both hydraulic machines operated with power factor less than 0.5, which is typical for servo and proportional hydraulic machines. Power factor correction equipment may be fitted to servohydraulic machines to increase the electrical efficiency of the process [17].

DISCUSSION

From the results reported here, age of molding machine appears to be the single most important factor in determining part repeatability. This reflects developments in process control technology leading to improved screw position and melt temperature control. Mechanical wear may also have been a factor, as the two older machines studied here had been in operation (although not continuously) in a research environment for over 10 years. Part quality was more difficult to assess for the specimens molded here. Part dimensions were slightly below the stated tolerances for this standard because the original mold tool design neglected shrinkage. The two newer machines produced parts with the lowest mean weight, although a target weight had not been set. The only part quality variable to compare directly was tensile strength of the molded specimens. This was found to correlate to melt temperature during injection rather than to machine age or mode of operation. Part quality measurements did, however, highlight the difficulty in reproducing identical molded parts from different molding machines even when using the same mold cavity insert. Bhogesara et al. [18] also found that the same machine settings on different machines produced different results, introducing the concept of "machine personality".

[FIGURE 11 OMITTED]

[FIGURE 12 OMITTED]

Discrepancies between machine technology (hydraulic or electric) were observed. Most noticeable was repeatability of screw position and melt consistency during injection (as indicated by peak melt pressure and temperature). All-electric machines appeared to control the process with less variation. This is possibly due to the more rigid drive constituents of servoelectric machines while some inherent "flexibility" in the servo and proportional hydraulic machines exists, relating to oil compressibility. However, the variation in process repeatability measured on hydraulic machines did not cause large variation in part quality, as shown by the coefficients of variation displayed in Table 4. Coefficient of variation in screw position for machine C was a factor of 10.3 and 3.9 higher than that of machines A and B respectively, but had the lowest variation in part weight. Machine D exhibited the highest variation in part weight (a factor of 2.7 times higher than that of machine A) but this was not indicative of the poor screw position control (15.5 times higher variation than machine A). These findings agree with an earlier study [6], which concluded that electric machines perform with better mechanical repeatability but not necessarily part reproducibility.

Start-up transients were found to be common across all injection molding machines. Part weight decreased gradually over the production of the first 1-50 cycles in each case, most likely due to the gradual increase in temperature of machine components such as the mold tool. This transient effect appeared to be machine size dependent, smaller machines being disturbed least. This confirms gradual heating to be the cause: smaller machines have less volume of metal to heat and therefore take less time to reach equilibrium. There is no simple way of eliminating transients from start-up--such effects are inherent in polymer processing. It is arguable whether start-up variations are important. In a production run of several thousand parts, a transient variation in the first 50 may be considered negligible. However, for large moldings or expensive polymers, any scrap produced could be critical. Advanced mold temperature control techniques such as pulsed mold tool cooling [19, 20] may help to reduce start-up variation, as may the development of automated "expert" in-line control systems [21].

CONCLUSIONS

For the 4 injection molding machines studied here, age of machine was found to be the most significant factor in determining part quality and repeatability. Modern (less than 4 years old at the time of the experiments) servoelectric and servohydraulic machines performed comparably. Differences in performance were observed between machine technologies. Servoelectric machines exhibited better machine control in terms of screw positioning, but this was not necessarily transferred to part repeatability. Energy consumption of servoelectric machines was found to be on average 3.6 times less than those of hydraulic operation.

Start-up transients were detected for all machines during molding of the first 50 parts and part quality was related to mold temperature during this period. Size of mold tool was found to effect the time taken for the mold temperature to stabilize after start-up. Tensile strength of molded parts was found to decrease linearly with melt temperature during primary injection.

TABLE 1. Injection moulding machine specifications.

                              Machine A.            Machine B.
                              servo-electric        servo-electric

Age of machine at time of      12                     1
  experiments (years)
Technology                    Servo-electric        Servo-electric
Screw diameter (mm)            22                    28
Screw stroke (mm)              65                    95
Max. shot volume ([m.sup.3])    2.47 X [10.sup.-5]    5.85 X [10.sup.-5]
Max. injection velocity       180                   200
  (mm/s)
Max. vol. flow rate             6.84 X [10.sup.-5]    1.23 X [10.sup.-4]
  ([m.sup.3]/s)
Max. hydraulic pressure       --                    --
  (MPa)
Intensification ratio         --                    --
Max. screw speed (r.p.m.)     300                   300
Max. tool size (w X h) (mm)   260 X 260             320 X 320
Max. clamping force (tonnes)   30                    50

                              Machine C.            Machine D.
                              servo-hydraulic       prop. hydraulic

Age of machine at time of       4                    14
  experiments (years)
Technology                    Servo-hydraulic       Prop. hydraulic
Screw diameter (mm)            40                    40
Screw stroke (mm)             160                   100
Max. shot volume ([m.sup.3])    2.01 X [10.sup.-4]    1.26 X [10.sup.-4]
Max. injection velocity       106                   100
  (mm/s)
Max. vol. flow rate             1.33 X [10.sup.-4]    1.25 X [10.sup.-4]
  ([m.sup.3]/s)
Max. hydraulic pressure        11.43                 11.50
  (MPa)
Intensification ratio          13.78                 12.25
Max. screw speed (r.p.m.)     215                   900
Max. tool size (w X h) (mm)   420 X 420             301 X 301
Max. clamping force (tonnes)   75                    60

TABLE 2. Experimental moulding conditions for each machine.

Set melt temp ([degrees]C)            210
Set mould temp ([degrees]C)            40
Volumetric throughput ([m.sup.3]/s)     2.01 X [10.sup.-4]
Hold pressure (MPa)                    30.0
Back pressure (MPa)                    12.0
Screw speed (r.p.m.)                  Varied with machine
Shot volume ([m.sup.3])                 7.02 X [10.sup.-5]
Transfer position volume ([m.sup.3])    2.51 X [10.sup.-5]
Hold time (s)                          15
Cool time (s)                          25

TABLE 3. Mean product and process measurements from each machine during
part numbers 301-400.

                                       Machine A.      Machine B.
Mean product/process paremeters        servo-electric  servo-electric

Part weight (g)                         15.73           15.53
Part length (mm)                       177.83          177.72
Part width (mm)                          9.80            9.64
Part thickness (mm)                      3.90            3.71
Part max. tensile force (kN)             0.928           0.868
Peak. nozzle pressure (MPa)             87.91           86.15
Peak nozzle temperature ([degrees]C)   219.98          227.60
Nozzle temperature rise ([degrees]C)    11.00           21.40
(peak temp.-temp. prior to injection)
Mean tool temperature ([degrees]C)      42.40           41.30
Initial screw position (mm)             63.51           34.46
Power consumption (kW/hr)                1.58            1.51
Power factor                             0.70            0.98

                                       Machine C.       Machine D.
Mean product/process paremeters        servo-hydraulic  prop. hydraulic

Part weight (g)                         15.51            15.86
Part length (mm)                       177.80           178.58
Part width (mm)                          9.81             9.76
Part thickness (mm)                      3.83             3.80
Part max. tensile force (kN)             0.913            0.912
Peak. nozzle pressure (MPa)             79.45            89.35
Peak nozzle temperature ([degrees]C)   222.15           222.67
Nozzle temperature rise ([degrees]C)     8.80            18.61
(peak temp.-temp. prior to injection)
Mean tool temperature ([degrees]C)      41.89            41.32
Initial screw position (mm)             27.50            28.00
Power consumption (kW/hr)                5.08             6.08
Power factor                             0.42             0.47

TABLE 4. Coefficients of variation in part and process during part
numbers 301-400.

Coefficient of variation  Machine A.      Machine B.
(%)                       servo-electric  servo-electric

Part weight                   0.031           0.026
Part thickness                0.271           0.362
Part length                   0.011           0.008
Part width                    0.316           0.156
Max. nozzle pressure          0.161           0.076
Max. nozzle temperature       0.063           0.038
Nozzle temperature rise       1.140           0.662
Tool temperature              0.388           0.139
Screw position                0.055           0.146

Coefficient of variation  Machine C.       Machine D.
(%)                       servo-hydraulic  prop. hydraulic

Part weight                   0.016           0.071
Part thickness                0.121           0.191
Part length                   0.007           0.012
Part width                    0.075           0.060
Max. nozzle pressure          0.708           1.832
Max. nozzle temperature       0.042           0.105
Nozzle temperature rise       1.650           1.778
Tool temperature              0.039           0.128
Screw position                0.568           0.851

TABLE 5. Fitting parameters for logarithmic model [y = a * ln(x) + c] to
describe mold tool temperature during machine start-up.

                       Machine  Machine  Machine  Machine
                          A        B        C        D

Maximum clamp force
  (tonnes)             30       50       75       60
Constant A. gradient
  ([degrees]C/cycle)    0.071    0.515    0.827    0.813
Constant C, intercept
  ([degrees]C)         42.08    38.88    38.04    37.10

Contract grant sponsors: IRC in Polymer Science and Technology; DTI MMP13 Materials Measurement Programme; Battenfeld; Mitsui Fanuc.

REFERENCES

1. P.D. Coates, Mater. World, Jan., 10 (1994).

2. J.W. Bozzelli and J. Cardinal, Proc. SPC PD3 Regional Tech. Conf., 35, B1 (1996).

3. W.C. Bushko and V.K Stokes, SPE ANTEC, 506 (1994).

4. I.J. Catic, Polym. Eng. Sci., 19, 893 (1979).

5. A.L. Kelly, M. Woodhead, R.M Rose, and P.D. Coates, Plast. Rubb. Compos., 29, 23 (2000).

6. W. Michaeli and T. Robers, Polimeri, 14, 299 (1993).

7. H. Thoma, S. Malterdingen, and B. Stillard, Kunstst. Ger. Plast., 82, 7 (1992).

8. C.C. Davis and J.C. Hudson, SPE ANTEC, 474 (1991).

9. D.B. Hunkar, SPE ANTEC, 275 (1987).

10. P.D. Coates and R.G. Speight, Proc. Inst. Mech. Eng. Part B: J. Eng. Manuf., 209, 357 (1995).

11. R.G. Speight, P.D. Coates, J.B. Hull, and C. Peters, Proc. Inst. Mech. Eng. Part E: J. Process. Mech. Eng., 115 (1997).

12. A.J. Dawson, A. Key, M. Kamala, R.M. Rose, and P.D. Coates, Polymer Process Engineering, P.D. Coates, ed. Institute of Materials, London (1997).

13. B. Souder, J. Davis, and J. Bak, SPE ANTEC, 814 (1996).

14. E. Moritzer and M. Neumann, SPE ANTEC, 616 (1999).

15. J. Ulik, SPE ANTEC, 3659 (1997).

16. British Standard BS EN ISO 527-2 (1996).

17. L. Ambs and M.M. Frerker, J. Energy Eng., 95, 55 (1998).

18. A.R. Bhogesara, R.E. Nunn, and S.P. McCarthy, SPE ANTEC, 576 (1995).

19. A.L. Kelly, M. Woodhead, P.D. Coates, P.S. Allan, and R. Evans, SPE ANTEC, Paper No. 617 (2002).

20. P.S. Allan, B.A. McCalla, Y. Mubarak, and D. Mulligan, SPE ANTEC, Paper No. 59 (2002).

21. R.G. Speight and A.R. Thomas, SPE ANTEC, Paper No. 995 (2000).

A.L. Kelly, M. Woodhead, P.D. Coates

Interdisciplinary Research Centre (IRC) in Polymer Engineering, School of Engineering, Design and Technology, University of Bradford, Bradford BD7 1DP, UK

Correspondence to: A.L. Kelly; e-mail: [e-mail]

First presented at SPE Annual Technical Conference (ANTEC), 2001, Paper #238.