Using real-time impedance measurement to monitor and control rubber vulcanization.

Knowing the point at which sufficient vulcanization occurs during the production of rubber products would allow the producer to minimize cycle time while reducing product variation. The objective of this article is to familiarize the rubber industry with impedance monitoring as a means of measuring rubber vulcanization. This article will discuss the background and science of impedance curing monitoring, and report the findings of a laboratory study in which this technology was evaluated in a REP injection molding press. The laboratory study was a cooperative effort between Signature Control Systems and REP International.

Part I of this article appeared in the December 1999 issue of Rubber World.

Part II: Measurement of optimum cure and quantification of results

Introduction

Part I of the study clearly demonstrated that the system is generating real-time crosslink density curves, of characteristic shape corresponding to elastic torque curves, that respond as expected to variation in the press settings. Although these plots are compelling, it was the desire of the researchers to measure some physical characteristic of the cured material in order to verify the ability of the sensor to measure end-of-cure.

As a method that provided for rapid results, and could also be used with varying thicknesses, it was decided that a visual observation of the part cross-section would provide for a reasonable method to determine the approximate state of cure. This provides only an approximate method of measuring cure state. Clearly, this method can be criticized in its precision, but it provides for a simple empirical approach in quantifying the system operation.

Specifically, the length of an undercured plug of material in the center of the disk could be measured that would provide for a rough measure of the extent of cure in the part. Due to heat transfer through the material, it is obvious that the outer edges of the disk will cure before the center cures. Since the sensor was located on the periphery, at approximately 130 mm, the sensor would see end-of-cure in this area before the center had completed cure. This method of measuring cure state was used in the thicker pans, both 20 mm and 50 mm. Figure 9 provides a rough schematic of the arrangement.

[Figure 9 ILLUSTRATION OMITTED]

This method shows an approximate state of cure of the part as it exits the mold. It does not take into account the significant curing that may, occur during cooling of the part. This post-cure can be especially important for thick parts.

This method is somewhat inexact, in that it requires a degree of subjectivity in determining the precise edge of the uncured portion, and it has no direct correlation to mechanical properties. However, it does provide a method to correlate the impedance curves with a physical property of the part.

Specifically, a peak on the crosslink density curve (inverse capacitance curve) is an indicator of complete cure under the sensor window. Therefore, if a peak is observed, and the part is cut open, the area under the sensor (130-140 mm) should be cured. If the cure is stopped precisely at the time of reaching the torque peak, the uncured length would therefore be expected to be in the range of 110-130 mm. Further curing would simply decrease the uncured length. It should be noted that in a production application, the optimal sensor location would obviously be near the center of the part, so that the end-of-cure indication under the sensor would mark the completion of cure for the entire part.

Results

Individual portions of the test are discussed in detail below. As stated earlier, the laboratory analysis system provides for analysis and monitoring only. All end-of-cure indications were visually observed curve peaks. The press was then manually opened.

End-of-cure detection: 20 mm thickness

The first step in the procedure was the conduct of numerous reference cures, conducted at a mold temperature of 185 [degrees] C. Twenty-three cures were performed for various lengths of time, ranging from a minimum of two minutes up to a maximum of four minutes, 40 seconds. Each part was cut open as it came out of the mold and the uncured length was measured and recorded. An uncured length of approximately 120 mm (110-130 mm), the optimum uncured length, was found to correlate closely with the time that the 9 kHz inverse capacitance curve (crosslink density curve) flattened. Figure 10 shows 9 kHz inverse capacitance for two separate cures which had an uncured length of 120 mm.

[Figure 10 ILLUSTRATION OMITTED]

After completion of the reference cures, 32 cures were conducted at 185 [degrees] C, where the cure was stopped at the 9 kHz inverse capacitance flat, and the uncured length was measured. Nine of these cures were conducted in the reference condition. Thirteen of the cures were conducted while varying the injection speed from a minimum of 5 mm/s up to a maximum of 35 mm/s. Ten of the cures were conducted while varying the injection temperature settings from a minimum of 60-70 [degrees] C, up to a maximum of 100-110 [degrees] C.

Variation in these parameters was induced in order to insure that the correct cycle time to reach the 110-130 mm uncured length would also vary appreciably. In this manner, the use of the 9 kHz peak to indicate end-of-cure could be validated. The variation in cycle time for the 32 cures ranged from a minimum of 2 m 40s, up to a maximum of 3 m, 50s.

For these cures, the average uncured length was 127.5 mm, which is in the optimal window for end-of-cure prediction. Figure 11 shows the distribution of uncured lengths. The distribution shows that the ideal range of 110-130 mm was most frequently met at end-of-cure. However, some distribution outside the optimum range was also observed. Given the subjectivity in visually ascertaining the curve flat, and also in measurement of the uncured area, some distribution of this kind is probably unavoidable. The distribution shown in figure 11 is generally applicable, regardless of the specific kind of variation induced in the press.

[Figure 11 ILLUSTRATION OMITTED]

The final variation to be conducted for the 20 mm thickness was the important variable of mold temperature. The mold was cooled to 170 [degrees] C and the same end of cure criteria were used to open the press. The time to reach the 9 kHz flat extended out to approximately six minutes. However, the uncured length at the end of cure was typically in the range of only 20 mm, significantly different than the result at 185 [degrees] C. Similarly, when the mold was heated to 200 [degrees] C, the 9 kHz flat was observed at approximately two minutes. However, the uncured length averaged approximately 165 mm at this temperature.

The cause of this unexpected result is unknown. Several possibilities are under investigation. It is believed that optimal sensor placement (near the part center) would remove this effect, but this hypothesis has not yet been validated.

Testing of the end-of-cure prediction for the 20 mm part was generally successful. The system appeared to detect proper end-of-cure through a wide range of variables. Further testing required will include more accurate methods of correlating the impedance curves with mechanical properties. Additionally, further investigation of mold temperature variation for this part thickness is required.

End-of-cure detection: 50 mm thickness

Seven cures of a part with 50 mm thickness were conducted. The sensor did obtain a clean signal on the thick part. However, the 9 kHz signal reached a flat in only about five minutes when the part was cured at 185 [degrees] C. Inspection of the part then showed uncured lengths that were typically in the range of 180 mm. End-of-cure detection for these parts is further complicated by the fact that thick parts will continue to cure significantly from self-heating even after removal from the mold. Even cures as long as 10 minutes left the uncured length in the range of 140 mm.

It is likely that a model will be required to account for self-heating of large parts after they leave the mold. This method could be a model-based calculation that would look for an end-of-cure criteria, having factored in the part thickness and temperature to account for self-heating. Alternatively, it may also be possible that an end-of-cure indication in the impedance curves could be successfully correlated to a final cure state after self-heating is complete.

Finally, it should also be noted that frequency settings are typically varied for each part thickness. A full range of frequency sweeps was not completed for the thick part. It is also possible that a simple adjustment to the excitation frequency could produce a more satisfactory result.

Testing of the end-of-cure detection capability for the 50 mm part was generally inconclusive. Further testing required includes correlation of the impedance curves with final cure state, after self-heating is complete. Additionally, the material should be explored through a full range of frequencies.

End-of-cure prediction: 5 mm thickness

The clearest data demonstrating the effectiveness of the technology was obtained with parts of 5 mm thickness or less. Since the parts were too thin to examine the cross section, a different method was used to measure the end-of-cure. In these cures, it was observed that a slightly undercured part would result in a dimple of uncured material in the center of the part. The dimple size would increase if the part were cured less. The dimple size Could be easily measured. Additonally, the thin part tests were conducted by alternating between impedance-controlled cure times and fixed cycle times. In this way, it was possible to clearly discern the effectiveness of using the impedance control to determine end-of-cure, by comparing the final product with the part produced at a fixed cycle time.

The procedure for the 5 mm thickness was as follows. The mold was heated to a temperature of 190 [degrees] C. A series of parts was produced at various cycle times. It was observed that the part consistently had a dimple of approximately 20 mm diameter when cured for 50 seconds at the 190 [degrees] C temperature. This size was selected as the control size. Therefore, the objective was to have a dimple of this size on each part produced. By conducting a frequency sweep during the cures, it was further observed that the part exhibited a peak on the 2 kHz inverse capacitance that occurred at approximately 50 seconds. This peak was selected as the impedance control feature.

The temperature was then allowed to drift down to approximately 175 [degrees] C in an uncontrolled fashion. A series of ten cures was conducted as the temperature drifted down. The cures alternated between press opening on fixed 50 second cycle times and press opening based on the 2 kHz inverse capacitance peak.

Figure 12 shows the comparison in dimple size between the impedance control and the time control. Note the consistency in the part when press opening is determined by the impedance control, as opposed to the variability observed with the time-based cycle. Figure 13 shows a plot of the corrected cure times (determined by reaching the inverse capacitance peak) required to produce the consistent part.

[Figures 12-13 ILLUSTRATION OMITTED]

Testing of the system with 5 mm parts was very successful. The impedance control system was clearly capable of producing parts with minimum variability.

End-of-cure detection: 2 mm thickness

Testing with the 2 mm parts was the same procedure as with 5 mm parts. However, more tests were conducted to verify the system effectiveness.

The procedure for the 2 mm thickness was as follows. The mold was heated to a temperature of 185 [degrees] C. A series of parts was produced at various cycle times. It was observed that the part consistently had a dimple of approximately 20 mm when cured for 17 seconds at the 185 [degrees] C temperature. This size was selected as the control size. Therefore, the objective was to have a dimple of this size on each part produced. By conducting a frequency sweep during the cures, it was further observed that the part exhibited a peak on the 5 kHz inverse capacitance that occurred at approximately 17 seconds. This peak was selected as the impedance control feature.

The temperature was then allowed to drift down to approximately 170 [degrees] C in an uncontrolled fashion. Twenty-four cures were conducted as the temperature drifted down. The cures alternated between press opening on fixed 17 second cycle times and press opening based on the 5 kHz inverse capacitance peak.

Figure 14 shows the comparison in dimple size between the impedance control and the time control. Note the consistency in the part when press opening is determined by the impedance control, as opposed to the variability observed with the time-based cycle. Figure 15 shows a plot of the corrected cure times (determined by reaching the inverse capacitance peak) required to produce the consistent part.

[Figures 14-15 ILLUSTRATION OMITTED]

Parts are organized into four rows. The top row and the third row are parts produced based on the fixed cure time. The second and bottom row are parts produced based on the impedance control. Severe deformation is noted in the first and third rows (fixed cycle), as compared to the uniformity observed in the second and fourth row (impedance controlled).

Part S12 in the two final cures is the part produced based on the impedance control. The inverse capacitance curve did not reach a peak until 1 minute, 21 seconds into the cure. Part M12 was produced based on the fixed 17 second cure time. Deformation was noted in the M12 part.

Testing of the system with 2 mm parts was very successful. The impedance control system was clearly capable of producing parts with minimum variability.

Conclusions

The study was highly successful in demonstrating the ability of the impedance monitoring system to respond to factors that influence rubber vulcanization. Factors such as injection speed, injection temperature and mold temperature produced predictable changes in the crosslink density curves generated by the system. Additionally, the study demonstrated the ability of the impedance control system to produce consistent parts, even with significant variation in mold temperature.

More study is currently in the planning stages. Specific areas of interest include quantitative mechanical testing of production parts, inclusion of metal inserts, analysis of thick parts (including mold temperature influence), analysis of industrial parts and optimization of the sensor location in the mold. Further development is underway to establish and verify cure control criteria for a variety of materials and thicknesses.

The study clearly shows the potential to monitor and control rubber production processes with real-time monitoring of crosslink density. The real-time control system offers the potential to dramatically reduce cycle times, while simultaneously improving product quality.

References

[1.] Kranbuehl, eds. Runt and Fitzgerald. Dielectric Spectroscopy of Polymeric Materials, ACS, 1997.

[2.] Persson, A novel method of measuring cure - dielectric vulcametry, Plastics and Rubber Processing and Applications 7 (1987), 111-125.

[3.] Khastgir, A comparative study of step curing and continuous curing methods, Rubber World, January 1994.

[4.] McCrum, Read and Williams. Anelastic and dielectric effects in polymeric solids, Dover Books, 1967.

[5.] Von Hippel. Dielectric Materials and Applications. Cambridge, Technology Press of MIT, 1954.

[6.] McCrum, Read and Williams, 1967.