Extruder temperature control with adaptive reset.

Extruder barrel temperature control justifiably receives a great deal of attention in design and development in the extrusion industry due to its importance to the process. Extruders have heating and cooling systems on their barrels for two basic reasons. First, to provide for heat-up and to provide both heating and cooling to satisfy processing requirements. Secondly, to provide accurate control of the inner barrel surface, since it affects the stability and quality of both output and melt temperature.

Accurate barrel control presents a very difficult challenge due to the physical construction of the barrel and an extruder's processing characteristics. The thick barrel wall separates the control point, at the inner surface, from the heating or cooling source, at die outside. This not only results in a considerable temperature differential, but creates a significant thermal time lag that causes major control problems. The process itself also contributes to control difficulty since the load demand can vary with material changes and change from heating to cooling as screw speed increases.

Conventional PID (proportional, integral, derivative) controls have special features beyond proportional output to compensate for the difficulties of control. Integral action, or reset, shifts the output to compensate for load (figure 1), so that the control point agrees with setpoint. Derivative action, or rate, is meant to compensate for large changes. Both can only respond after an upset has occurred, however, so are inherently limited in their response (ref. 1). Recent improvements attempting to overcome these limitations include auto-tune and the application of fuzzy logic. Extruder manufacturers usually either supply machines with shallow thermocouples that appear to control well, but in fact do not control the inner surface, or supply proper deep thermocouples with sluggish control systems (ref. 2).

An alternate system that utilizes two thermocouples was developed and patented several years ago (ref. 3) that monitors the temperature at the heat/cool source as well as at the control point, providing an anticipatory feature that works as a sort of instantaneous derivative. Reset is still achieved through a conventional automatic function. The dual thermocouple system has proven to be superior to PID systems in stability (ref. 4), but with little or no advantage in upset recovery time.

Objective

One of the most problematic manifestations of temperature control limitations occurs in extrusion processes when the screw speed is changed, either as part of a planned sequence of string-up, change-over or splicing, or during an unplanned event such as a shut-down or emergency stop. Barrel temperature upsets occur due to the change in process load that results at the inner barrel surface. For example, a barrel zone that runs with high cooling due to high shear heat generation of the screw will drastically undershoot the setpoint if the screw is suddenly stopped, because the heat load is suddenly removed. It will then go into a heating mode in order to recover, entirely the opposite of the running mode. Upon restart, the situation must reverse, requiring not only considerable time to reach setpoint, but additional time to achieve stability.

This article discusses the results of a development program initiated with the purpose of creating a barrel temperature control system that could quickly return to a previous stable condition by somehow remembering where it was. If a particular zone was operating at a high cooling load at running speed, for instance, and for some reason was momentarily stopped, it would instantly apply that amount of cooling as soon as the process was restarted. Extensive testing showed that the most significant control parameter that could relate the heating or cooling load to screw speed was the reset value. The amount of reset, which can be thought of as load compensation (figure 1), is directly proportional to the amount of heating or cooling required. The shear heat generated by an extruder screw is largely a function of screw speed, so relating this control parameter to speed is a technically sound approach.

Development

A control scheme called adaptive reset was recently conceived, developed, tested and patented (ref. 5). Based on the proven dual thermocouple principle, the new system has added an internal look-up table in memory of reset values versus screw speeds. This table is automatically learned as the extruder is operated in a normal manner according to the program's rule sets and logic. These adaptive reset tables are stored as part of the controller's recipes for each material and screw design and then applied whenever screw speed changes. This action works as a sort of instantaneous integral. The system provides fast response, accuracy and stability, and for die first time, a positive method for accommodating large changes and interruptions.

Experimental

Extensive extrusion trials were conducted on both 63.5 mm and 114 mm extruders with various thermoplastic materials and screw designs, including combinations that required high cooling as well as high heating during the development and testing of the new control system.

This article presents the results of a series of trials on the 114 mm, 24:1 L/D water-cooled extruder with a DSB barrier screw and an EPIC II control system with data acquisition capabilities. All trials were run using a 0.9 melt index (MI) film grade LLDPE, and the exact sequence was repeated three times for the three types of barrel temperature controllers under study: auto-tuning PID, dual thermocouple and adaptive reset. The PID controllers were discrete units that were set to the manufacturers recommended settings and then auto-tuned. The latter two control schemes were both run in the EPIC II control system. During the dual thermocouple trial, adaptive reset action was disabled, but the system learned the reset values for the screw speeds that were applied during the final trial. During each trial the extruder was run at 20 rpm until the process and all barrel zone temperatures, which were set at 177[degrees]C, were stable. Screw speed was then changed to 60 rpm. After 30 minutes the extruder was suddenly stopped for 12 minutes, then restarted and brought back to 60 rpm. All barrel zone temperatures were monitored and recorded during the trials, as was melt temperature, the critical parameter that affects process.

Results

0.9 MI LLDPE is a fairly high viscosity material, demanding high torque and considerable cooling from the extruder when processed on a high output barrier screw. The reset values associated with screw speed for barrel zones 1 and 5 are shown in figure 2. These are the values that were learned and applied by the adaptive reset control system during die trials. The feed area, zone 1, shows a heating requirement at low speed, and then a cooling demand for speeds over 30 rpm. Zone 5 at the discharge shows an increasing cooling requirement for all speeds.

Figures 3 through 6 show the recorded deep barrel control thermocouple values over time during the three repeated trials with the three different control systems on the time scale from the first run. Figures 3 and 4 are of zone 1, figures 5 and 6 are of zone 5. Figures 3 and 5 show a screw speed change from 20 to 60 rpm and figures 4 and 6 show a screw speed change from 60 to 0 and back to 60 rpm. In all cases, the initial reaction to speed changes for all three controllers is identical since the deep thermocouples are reflecting the rapid change in process load - the addition or loss of polymer shear heating from the screw. The differences in the curves after that show the differences in the controllers.

In figure 3, going from 20 to 60 rpm, the PID controller undershoots the setpoint as it is recovering from the initial upset and then takes a considerable amount of time, about 25 minutes overall, to reach setpoint. The dual thermocouple scheme avoids undershoot but waits at the initial upset temperature longer, then goes through a reset, and reaches setpoint in about 18 minutes. Adaptive reset shows a dramatic improvement, with no lag or overshoot, coming right back to setpoint in only seven minutes.

Figure 4, with screw speeds going from 60 to zero and back to 60 rpm, shows the same characteristic patterns. The PID controller undershoots and oscillates, the dual thermocouple control hesitates and goes through an obvious reset, and the adaptive reset control quickly returns to setpoint in about a third of the time as the other two.

Figures 5 and 6 show similar response on zone 5, except that the PID controller, operating on a zone with a higher cooling load, appears to be much slower in reaction and never reaches setpoint. The other two systems behave similar to zone 1, with the superior performance of adaptive reset obvious. The affect on melt temperature is shown in figures 7 and 8 for dual thermocouple and adaptive reset controls, and shows the dramatic improvement that a quick recovery to setpoint can have on the process. In both cases, the new system brought the melt temperature back to a stable condition in about half the time.

Conclusions

Extruder barrel temperature control can greatly influence extrusion rate, melt temperature, quality and stability. Accurate and responsive control of barrel temperatures is therefore a critical factor for efficient and quality extrusion. Processes that include speed changes or frequent, even if unplanned, stop-start sequences need rapid recovery of barrel control.

The response of the PID controllers showed one of the weaknesses of auto-tune - that the automated tuning operation occurs with an empty, stopped extruder with no process heating or cooling load. Under operating conditions, the actual load is not only different, but can change from heating to cooling with a change in speed (figure 2). The same controller behaved differently on the two different zones, having excessive gain on zone I and overshooting, and not enough on zone 5, being very slow to get to setpoint.

The newly developed adaptive reset temperature control system is shown to give dramatic improvement in response to screw speed changes and start-stop events through a scheme of automatically learning and then applying reset values according to screw speed. It also exhibits the excellent stability inherent to die dual thermocouple control system. It offers extrusion processors significant improvements in quality and efficiency at no significant increase in capital cost.

References

(1). R.F. Bayless, SPE Antec technical papers, pg. 544, 1978. (2). H.E. Harris, SPE Antec technical papers, pg. 423, 1977. (3). L. Faillace, U.S. patent no. 4,262,737, 1981. (4). A. Hasson, C. Miaw, G. Balch, SPE Antec technical papers, pg. 979, 1986. (5). L Faillace, U.S. patent no. 5,149,193, 1992.