Rotational molding cycle time reduction through surface-enhanced molds, Part B: experimental study.

INTRODUCTION

The rotational molding industry has recognized that long production cycle times often hinder the overall process efficiency and further progress. This issue may be due to both manufacturing scheme design and processing constraints. This article focuses on the processing issues, in particular mold design for enhanced heat transfer. During rotational molding, the mold and the plastic powder are heated from room temperature to the melting temperature of the plastic, and then cooled back to room temperature. For convenience and cleanliness reasons, forced air convection systems are preferred for both the heating and cooling phases. As a result, it is imperative to find practical solutions to enhance convective heat transfer within the existing process, thereby reducing production cycle times.

Until recently, most efforts to shorten cycle times have been focused on optimizing process parameters, improving machine design, changing thermal characteristics of the shell material [1], introducing a small amount of internal pressure [2, 3], and applying internal cooling [4]. Limited attention has been given to mold surface modification as a technique to enhance heat transfer from oven or cooling bay to the mold [5]. This article explores and validates the potential for cycle time reduction in rotational molding through the application of extended surfaces (i.e. pins and fins) and surface roughness, the results of which have been presented in Part A of this study. Extended surfaces have already been shown to enhance heat transfer by increasing the heat transfer surface area. Roughness elements, on the other hand, are utilized in conjunction with turbulent flows, also producing significant increases in heat transfer rates. Even though these techniques have been employed in many applications such as heat exchangers, turbine blades, missiles, ship hulls, and reentry vehicles, they have seldom been used in the rotomolding process. Therefore, this study should provide a new option to the rotomolding industry.

In this article, the benefits of extended or rough surfaces to the exterior of a mold are investigated and demonstrated experimentally. Internal air temperature within the molds has been monitored throughout the experiments, providing a measure of the resulting cycle times. These recorded data are also compared to numerical predictions made using the RotoSim software package, a detailed description of the method being provided in Part A. Finally, a brief study of resulting part mechanical properties is presented, demonstrating the effect on such properties due to cycle time reduction.

[FIGURE 1 OMITTED]

EXPERIMENTAL APPARATUS AND PROCEDURES

Rotational Molding Machine

All experimental trials were carried out on a shuttle rotational molding machine (Ferry Express E-100). The specifications of this machine are:

* biaxial rotation,

* approximate oven size of 2.6 m x 2.6 m x 2.6 m,

* approximate cooling bay size of 4.3 m x 3.1 m x 2.8 m,

* oven fan capacity of 7.79 [m.sup.3]/s (16,500 cfm),

* cooling forced air fan capacity of 3.16 [m.sup.3]/s (6700 cfm), and

* cooling exhaust fan capacity of 3.30 [m.sup.3]/s (7000 cfm).

The machine is equipped with a computerized control panel that facilitates alterations to the different process parameter settings, and also to a compressed air pressure system. However, this machine did not provide options for changing the heating and cooling fan speeds and the choice for cooling methods i.e. only forced or natural air-cooling was available.

Molds

Three cubical aluminum molds were fabricated for rotomolding trials. The following dimensions were chosen for these molds:

* For the plain molds, the external dimensions were 300 mm x 300 mm x 300 mm with a wall thickness of 8 mm. The surface-enhanced molds have the same dimensions excluding roughness elements and pins, such that the internal mold dimensions are kept constant at 284 mm x 284 mm x 284 mm. All three molds have zero draft angle.

* For the roughness-enhanced mold, square-based pyramids with base dimensions of 4 mm x 4 mm, and a 2 mm height were used. Zero spacing is maintained between any two of these roughness elements (Fig. 1a).

* For the pin-enhanced mold, square-based pins with 24 mm height, 10 mm thickness, and 20 mm interpin spacing were chosen (see Fig. 1b).

* Each mold was made up of seven plates as illustrated in Fig. 2. Screws were used to assemble the plates to form each mold. A CNC milling machine was used to manufacture the roughness elements and pin array.

Many previous experimental investigations of rotational molding, especially at The Queen's University of Belfast, have used similar mold dimensions to those described here [2-4, 6]. The dimensions of the roughness elements and pins were selected based on theoretical study (presented in Part A) and preliminary experimental results presented elsewhere [5]. Because of the roughness elements and pins, the increases in mold mass are ~8.4 and 33.4%, respectively.

Polymer Material and Molding Conditions

The polymer used in all trials was Cotene 9042 polyethylene powder, which is available in black, natural, and standard colors. For the reported experiments, natural color polymer was used, which was specially formulated with a long-term UV stabilizer, and had a nominal melt flow index of 4 g/10 min and bulk density of 937 kg/[m.sup.3].

The experiments were carried out at two oven temperature settings (300 and 380[degrees]C) and three part wall thicknesses (3.2, 6.0, and 9.0 mm). In all trials, rotational speeds of 8.0 and 2.0 rpm were applied to the arm and plate, respectively. The peak internal air temperature was targeted at 200[degrees]C for each run, and a cycle was considered complete when the internal air temperature reduced to 80[degrees]C. A recirculating hot air oven system and forced air-cooling were employed for the heating and cooling cycles, respectively. Differences between the internal air temperature and mold surface temperature at the beginning of each trial were kept within 3[degrees]C. This check is very important to minimize the effect of mold preheating.

[FIGURE 2 OMITTED]

[FIGURE 3 OMITTED]

Rotolog--Data Acquisition Device

A Rotolog device [4] was used to monitor and record internal air temperature profiles during the trials. Using a wireless-based transmitter system, the device allowed remote and real time monitoring of various important temperatures. The Rotolog transmitter was housed in a portable stainless steel container, which was normally mounted near the mold. A high temperature shield to endure the harsh environment of the oven protected the electronics. This device could monitor up to four different temperatures simultaneously. Temperatures were measured using k-type thermocouples, and the data were then transmitted to a receiver located external to the rotational molding machine. The receiver interfaced with a computer and transferred the data for an on-screen display in graphical format. Rotolog devices are normally calibrated to an overall accuracy of [+ or -] 2[degrees]C.

RESULTS AND DISCUSSION

Cycle Time Reductions

The aim of the experimental program was to study the benefits provided by the surface-enhanced molds. The experiments were done using five sets of molding conditions i.e. part wall thicknesses of 3.2 and 6.0 mm for an oven temperature of 300[degrees]C, and part wall thicknesses of 3.2, 6.0, and 9.0 mm for an oven temperature of 380[degrees]C. Figure 3 shows the internal air temperature profiles for a part wall thickness of 3.2 mm at oven temperatures of 300 and 380[degrees]C, respectively. The cycle time reductions for the roughness-enhanced and pin-enhanced molds were 17 and 32% at an oven temperature of 300[degrees]C, and 16 and 26% at an oven temperature of 380[degrees]C.

For part wall thickness of 6.0 mm at oven temperatures of 300 and 380[degrees]C, the cycle time reductions were ~19 and 28%, and 20 and 25% for the roughness-enhanced and pin-enhanced molds, respectively. These results are presented in Fig. 4, demonstrating substantial cycle time reductions, and similar trends to those observed for the smaller part wall thickness. Because of a higher peak internal air temperature for the pin-enhanced mold (~4[degrees]C) as compared to the plain mold for an oven temperature of 380[degrees]C, a small drop in the cycle time reduction relative to the 300[degrees]C oven temperature setting was observed.

Figure 5 presents the internal air temperature profiles for an oven temperature of 380[degrees]C and a part wall thickness of 9.0 mm. The cycle time reductions for the roughness-enhanced and pin-enhanced molds were 16 and 28%, respectively. Again, these results demonstrate similar trends to those observed for the surface-enhanced molds described earlier. The average cycle time reductions achieved across all trials were 18% for roughness-enhanced mold, and 28% for the pin-enhanced mold. Significant reductions were made under all conditions, showing relative insensitivity to part wall thickness and oven temperature. Table 1 provides the summary of the experimental cycle time reductions due to the surface-enhanced molds.

[FIGURE 4 OMITTED]

[FIGURE 5 OMITTED]

Comparisons Between Predicted and Experimental Cycle Times

Comparisons between the predicted and experimental cycle times have been made to assess the credibility of the prediction methods described in Part A. Five sets of comparisons were made for each mold according to the combination of molding conditions specified. All temperature traces presented for these comparisons are of the internal mold air.

Plain Mold Comparisons. The comparisons between predicted and experimental internal air temperature traces for the plain mold are presented in Fig. 6. Figure 6a depicts the comparison between the predicted and experimental internal air temperature traces for an oven temperature of 300[degrees]C and a part wall thickness of 3.2 mm. The comparison during the heating stage is very good. It is interesting to note that the predicted temperature profile completed the full cycle earlier than the experimental trace. This is very likely due to the separation between the plastic and the mold wall, which occurs during the cooling stage. This phenomenon is due to the plastic part shrinking away from the inner mold surface during and after polymer crystallization. This event is referred to as "mold/part separation" in the following discussion. Separation between the plastic part and mold wall is not modeled by the RotoSim package. The air gap established between these two surfaces represents a significant resistance to heat transfer. As a consequence, the difference between the overall predicted and experimental cycle times was 16.9%.

For an oven temperature of 380[degrees]C and a part wall thickness of 3.2 mm, the comparison between the predicted and experimental internal air temperature traces is shown in Fig. 6b. Both the heating and cooling stages of the predicted and experimental profiles (without internal pressure) compare very well to each other. The difference in the overall predicted and experimental cycle times was 12.2%. Because of the "mold/part separation" effect, which is not modeled in the RotoSim package, the predicted profile completed the cycle faster than the experimental trace. To prove this hypothesis, an experimental temperature profile is included in Fig. 6b for the plain mold with the inclusion of internal pressure from the beginning of the cooling stage until the completion of the cycle. The internal pressure was ~17.2 kPa (2.5 psi), and was employed to eliminate (or at least minimize) separation between the part and mold wall. With the inclusion of internal pressure, the difference in the overall predicted and experimental cycle times was cut to 4.3%. This result suggests that the predicted heat transfer coefficients (i.e. oven to mold and mold to cooling fluid coefficients) are in excellent agreement with the experimental heat transfer coefficients, considering the limitations of the RotoSim package.

Figures 6c and d present the comparisons between the predicted and experimental internal air temperature traces for a part wall thickness of 6.0 mm and oven temperatures of 300 and 380[degrees]C, respectively. Differences between the predicted and experimental cycle times were 24.2 and 24.7% for oven temperatures of 300 and 380[degrees]C, respectively. The heating stage of the predicted profiles was shorter for both cases when compared to the experimental heating stage. Such differences are not unexpected due to several heat transfer coefficients being sourced from the literature (i.e. mold to internal air, internal air to fully melted layer, mold to powder, and powder to internal air). For a shorter predicted heating cycle (i.e. part wall thickness of 3.2 mm), such errors are not as evident because the proportion of part wall thickness to the mold wall thickness is just 40%. However, when the proportion of part wall thickness increase to 75% (i.e. the 6.0 mm part), errors induced through the use of incorrect heat transfer coefficients become more evident.

For an oven temperature of 380[degrees]C and a part wall thickness of 9.0 mm, the comparison between the predicted and experimental internal air temperature traces is depicted in Fig. 6e. The difference between the predicted and experimental cycle times is 31.0%. The gap between the predicted and experimental cycle times is further increased, which can be expected as the proportion of part to mold wall thickness increases to 125%. Similar to the two previous cases, the literature-based heat transfer coefficients and the "mold/part separation" effect contribute to a large difference in the overall comparison between the predicted and experimental internal air temperature traces.

[FIGURE 6 OMITTED]

Roughness-Enhanced Mold Comparisons. Figures 7a and b show comparisons between the predicted and experimental internal air temperature traces for a part wall thickness of 3.2 mm, and oven temperatures of 300 and 380[degrees]C, respectively. Differences between the predicted and experimental cycle times are 21.1 and 18.1%, respectively. As for the plain mold, cooling cycle times for the experimental profiles are expected to be longer than the predicted cycles. This is due to the "mold/part separation" effect explained previously. An experimental temperature profile of the roughness-enhanced mold with the inclusion of internal pressure is included in Fig. 7b, the pressure employed being ~17.2 kPa (2.5 psi). With the inclusion of internal pressure, the difference between the experimental cycle times as compared to the theoretical cycle time is reduced to 6.1%. This result suggests that the theoretical and experimental heat transfer coefficients (i.e. oven to mold and mold to cooling fluid coefficients) are in good agreement with each other, even though some qualitative differences exist during the heating stage.

Figures 7c and d provide comparisons between the predicted and experimental internal air temperature traces for a part wall thickness of 6.0 mm and oven temperatures of 300 and 380[degrees]C, respectively. Differences between the predicted and experimental cycle times are 28.6 and 27.5%, respectively. For an oven temperature of 380[degrees]C and a part wall thickness of 9.0 mm, the comparison between the predicted and experimental internal air temperature traces is shown in Fig. 7e. The difference in the experimental cycle time compared to the theoretical cycle time is 35.9%. Results obtained using the roughness-enhanced mold have shown very similar patterns as observed for the plain mold. The two main contributors to the observed errors are the "mold/part separation" effect and the literature-based heat transfer coefficients as explained previously.

[FIGURE 7 OMITTED]

Pin-Enhanced Mold Comparisons. For the pin-enhanced mold, two methods were used to predict the heat transfer coefficients to and from the mold. The first, method A, applied the free stream air velocity (i.e. 6 m/s) directly to the one-dimensional steady state analysis described elsewhere [5]. The second, method B, employed a computational fluid dynamics (CFD) package, Phoenics-VR, to estimate the average air speed around pins. The average air speeds obtained were then applied to the one-dimensional steady state analysis to approximate the heat transfer coefficient. The experimental internal air temperature profile is compared to both predicted temperature profiles in Fig. 8.

Figure 8a shows the comparisons between the predicted and experimental internal air temperature traces for an oven temperature of 300[degrees]C and a part wall thickness of 3.2 mm. The differences between the experimental cycle time as compared to cycle times predicted using methods A and B are 19.5 and 28.0%, respectively. The heating stage of method A compared quite well to that of the experimental cycle, while the heating stage of method B was moderately faster. Because of the "mold/part separation" effect, the predicted profiles have completed the molding cycle faster than the experimental trace. For an oven temperature of 380[degrees]C and part wall thickness of 3.2 mm, the comparisons between the predicted and experimental internal air temperature traces are shown in Fig. 8b. Differences between the overall experimental cycle time as compared to methods A and B are 20.2 and 27.7%, respectively.

[FIGURE 8 OMITTED]

An experimental temperature profile for the pin-enhanced mold with the inclusion of internal pressure is incorporated in Fig. 8b. With the inclusion of pressure, differences between the overall experimental cycle time as compared to methods A and B are reduced to 0.1 and 9.5%, respectively. While comparisons of the predicted cycle time for method A are very good, significant deviation from the experimental temperatures are noted during the heating stage of the process. This might be due to the inappropriate heat transfer coefficient values used for various interactions as explained in the previous cases. Increases in mold mass and oven temperature setting (i.e. shorter heating cycle) may make this difference more noticeable. Accordingly, the deviation is evident for the combination of pin-enhanced mold (which has 33.3% increases in mold mass as compared to 8.3% for the roughness-enhanced mold) and an oven temperature of 380[degrees]C (as opposed to 300[degrees]C). The pin-enhanced mold requires more energy storage before it begins to heat-up, resulting in the slower initial heating stage shown in the experimental internal air temperature traces.

Figures 8c and d present comparisons between the predicted and experimental internal air temperature traces for a part wall thickness of 6.0 mm and oven temperatures of 300 and 380[degrees]C, respectively. Differences between the experimental cycle time as compared to methods A and B are 30.9 and 38.0%, respectively for an oven temperature of 300[degrees]C, and 34.2 and 40.9%, respectively for an oven temperature of 380[degrees]C. For an oven temperature of 380[degrees]C and a part wall thickness of 9.0 mm, the comparisons between the predicted and experimental internal air temperature traces are shown in Fig. 8e. Resulting differences between experimental cycle times and methods A and B are 34.7 and 40.0%, respectively.

Again, these results reveal similar trends as noted for the plain and roughness-enhanced molds. While the mold/part separation effect and the literature-based heat transfer coefficients appear to be the reasons for the differences in the predicted and experimental cycle times, it is interesting to note that method A provides better predictions than method B. Further investigation should be conducted to understand how pins really behave in the rotational molding environment. Perhaps, the boundary conditions of the CFD analysis require further improvement to simulate the rotomolding process environment. In addition, there should be a better technique to estimate the average air speed between pins as compared to the method discussed in Part A.

Although prediction methods presented here fail to forecast the exact experimental cycle times, they have been proven successful for the prediction of relative cycle time reductions between the plain and surface-enhanced molds. For the same oven temperature setting and part wall thickness, the experimental cycle time reductions are on average 18 and 28%, while the predicted cycle time reductions are ~21 and 32%. These results indicate that the predicted and experimental relative cycle time reductions are in very good agreement. Therefore, the specified analysis methods can be employed to predict relative cycle time reductions, while further improvements should be made for the prediction of actual cycle times. This conclusion is demonstrated in Fig. 9. Figure 9a presents experimental air temperature traces of all three molds for an oven temperature of 300[degrees]C and a part wall thickness of 6.0 mm, while Fig. 9b presents the related predictions. The comparisons between the experimental and predicted cycle time reductions are presented in Table 1.

MECHANICAL PROPERTY TESTING

A brief mechanical property-testing program was completed to determine if there is any effect on the quality of the final parts produced using the surface-enhanced molds. While there are many tests available for checking the quality of the final plastic products, the authors believe that impact strength, tensile yield strength and part wall thickness distribution data should provide sufficient information for this initial study. The reader should bear in mind that the material used in all trials is the same, the only difference between the experiments being the mold used, i.e. the plain mold (current practice in the rotomolding industry) as opposed to the surface-enhanced molds. The main difference between the experiments is the thermal history experienced by the polymer, as the molds have produced different heat transfer rates. While roughness elements and pins have provided higher heat transfer coefficients than that of a plain surface, the required increases in mold mass due to roughness elements and pins will serve to lengthen the time required to store and remove heat.

[FIGURE 9 OMITTED]

Testing Procedures

Impact tests were carried out according to the Association of Rotational Molders (ARM) low-temperature impact test method [7]. This method covers the determination of the relative ranking of materials according to the energy required to crack or break flat, rigid plastic specimens under the specific condition of impact due to a free-falling dart. A dart weight is dropped at increasing heights until failure is achieved. The Bruceton method uses a multitude of test specimens. The drop height is determined by first picking an arbitrary drop height, decreasing it if the first sample fails, or increasing it if the sample survives the test. Height value and failure mode (i.e. none failure, ductile failure, or brittle failure) of each sample is recorded. Once all the samples have been tested, mean failure energy can be determined. Fifteen samples were taken from each mold of 3.2 mm part wall thickness for both oven temperatures of 300 and 380[degrees]C. All samples were tested at a temperature of -40 [+ or -] 2[degrees]C.

Tensile yield strength testing at 50 mm/min (according to ASTM D 638 standard) was also performed on samples produced in each of the three molds. This method is used to determine the tensile properties of plastics. Five samples were taken from each mold for an oven temperature of 380[degrees]C and a part wall thickness of 3.2 mm. The thickness of each sample was measured prior to the tensile test at three different points using a micrometer.

Determination of part wall thickness was another measurement conducted for each mold. Eighteen measurements were taken from each part produced in a 3.2 mm wall thickness for both oven temperatures of 300 and 380[degrees]C. The average, standard deviation, minimum and maximum values of the part wall thickness were determined.

Results and Discussion

Impact strength results taken from samples produced in the surface-enhanced molds were compared with those produced from the plain mold. Figure 10 presents the results for the falling weight impact strength tests performed on the 3.2 mm part wall thickness, for both oven temperatures of 300 and 380[degrees]C. These results suggest that the impact strength declines to some extent when using the surface-enhanced molds. However, the reduction in impact strength between the surface-enhanced molds and the plain mold was not consistent, and is of a moderate amount. While further investigation on the impact strength might be appropriate, the results suggest that a significant amount of the impact strength is still upheld through the use of surface-enhanced molds.

The average tensile yield strength values for the plain, roughness-enhanced, and pin-enhanced molds were 18.4, 18.3, and 18.8 MPa, respectively. These results show no measurable difference in tensile yield strength between the plain mold and the surface-enhanced molds. Hence, the use of surface-enhanced molds does not appear to have an effect on the tensile yield strength of the final product.

The results for the thickness distribution measurements are summarized in Tables 2 and 3. Table 2 presents data for an oven temperature of 300[degrees]C and a targeted part wall thickness of 3.2 mm. Data for an oven temperature of 380[degrees]C are revealed in Table 3. Both tables indicate that there is only a marginal difference in part wall thickness distribution amongst the three molds, and that no clear trends are evident. This suggests that the impact on the part wall thickness due to the use of surface-enhanced molds is minimal. To conclude, while initial mechanical property testing and thickness measurements show promising results, more data from different part wall thicknesses and part sizes are required to be conclusive.

CONCLUSIONS

The goals of this article were to quantify cycle time reductions gained through the use of surface-enhanced molds, and to validate the predicted cycle times against the actual rotomolding trials. The average experimental cycle time reductions were ~18 and 28% for the roughness-enhanced and pin-enhanced molds, respectively. The results have demonstrated that the developed heat transfer enhancement concepts work. For such relatively simple techniques, significant cycle time reductions can be achieved. The savings (obtained using an industrial machine) are very significant, inviting the rotomolding community to incorporate these techniques efficiently in an industrial setting.

While the prediction methods presented here are not able to predict the exact experimental cycle times, they have proven successful in predicting relative cycle time reductions between the plain and the surface-enhanced molds. The average predicted cycle time reductions were ~21 and 32% for the roughness-enhanced and pin-enhanced molds, respectively. These results show that the predicted and experimental cycle time reductions are in excellent agreement with each other. A small reduction in the impact strength results may or may not be a concern depending on the application of the final products. Although the initial mechanical property testing and thickness measurements provided show promising results, more data from different part wall thicknesses and part sizes are required to be conclusive.

ACKNOWLEDGMENT 

The authors would like to thank ICO Courtenay (NZ) Ltd. for their participation in the project.

REFERENCES

1. W. Yan, R.J.T. Lin, S. Bickerton, and D. Bhattacharyya, Mater. Sci. Forum, 437-438, 235 (2003).

2. R.J. Crawford, K. Passow, S. Percy, and M. Kearns, Rotation, IX, 30 (2000).

3. R.J. Crawford, M.C. Cramez, M.J. Oliveira, and A.G. Spence, Annu. Tech. Conf.-ANTEC, 1, 1250 (2002).

4. R.M. Khouri, "Reducing Cycle Times in Rotational Molding: A Theoretical and Experimental Analysis", Ph.D. Thesis in Mechanical and Manufacturing Engineering, Queen's University of Belfast, UK (2004).

5. M.Z. Abdullah, S. Bickerton, and D. Bhattacharyya, Polym. Eng. Sci., 45, 114 (2005).

6. S.P. O'Neill, "Analysis of Cooling in Rotational Molding", Ph.D. Thesis in Mechanical and Manufacturing Engineering, Queen's University of Belfast, UK (1999).

7. Association of Rotational Molders (ARM) International, Low Temperature Impact Test, Version 2.0, ARM International, Illinois 60523 USA (2000).

M.Z. Abdullah, S. Bickerton, D. Bhattacharyya

Center for Advanced Composite Materials, Department of Mechanical Engineering, University of Auckland, Auckland, New Zealand 

Correspondence to: M.Z. Abdullah; e-mail: [e-mail]

Contract grant sponsors: Association of Rotational Molders International, the United States of America; Foundation for Research Science and Technology, New Zealand.

TABLE 1. Summary of the experimental and predicted cycle time
reductions.

                                    % Cycle time reduction
                         3.2 (a)  6.0 (a)  3.2 (b)  6.0 (b)  9.0 (b)

Roughness-enhanced mold  17       19       16       20       16
  (experiment)
Pin-enhanced mold        32       28       26       25       28
  (experiment)
Roughness-enhanced mold  21       21       21       21       20
  (prediction)
Pin-enhanced mold        33       32       32       32       31
  (prediction)

The values 3.2, 6.0, 3.2, 6.0, and 9.0 indicate part wall thickness (in
mm).
(a) For an oven temperature of 300[degrees]C.
(b) For an oven temperature of 380[degrees]C.

           Falling Weight Impact Strength (J)
           Oven temperature:  Oven temperature:
Mold       300[degrees]C      380[degrees]C

Plain      53.4               56.1
Roughness  44.1               54.2
Pin        46.9               44.1

FIG. 10. Comparisons of falling weight impact strength between the
molds.

Note: Table made from bar graph.

TABLE 2. Summary of part wall thickness distribution for molded parts
with an oven temperature of 300[degrees]C and a targeted part wall
thickness of 3.2 mm.

                          Part wall thickness (mm)
Mold                Average  Std. dev.  Minimum  Maximum

Plain               3.24     0.26       2.79     3.67
Roughness-enhanced  3.33     0.15       3.15     3.65
Pin-enhanced        3.21     0.30       2.82     3.85

TABLE 3. Summary of part wall thickness distribution for molded parts
with an oven temperature of 380[degrees]C and a targeted part wall
thickness of 3.2 mm.

                          Part wall thickness (mm)
Mold                Average  Std. dev.  Minimum  Maximum

Plain               3.31     0.18       2.97     3.67
Roughness-enhanced  3.18     0.25       2.73     3.68
Pin-enhanced        3.25     0.19       3.04     3.64