News

Temperature closed-loop control system of extrusion blow molding machine

Extrusion Blow Molding Temperature Closed-Loop Control: How to Stop Drifting Before It Kills Your Production

Temperature control in extrusion blow molding sounds simple. Set the heaters, wait for the thermocouple to read the target, and go. Except that approach has been causing wall thickness variation, inconsistent bottle weight, and mysterious quality shifts for decades. The polymer melt doesn't care what your thermocouple says. It cares about its actual temperature — and there's often a 10 to 20 degree gap between what the sensor reads and what the material experiences.

Closed-loop temperature control closes that gap. But setting it up right requires understanding where heat goes, where it's lost, and why standard PID tuning from the manufacturer's manual usually falls apart within a week of production.

Where Temperature Control Actually Fails in Blow Molding

Most machines have thermocouples embedded in the barrel and die. The controller reads those values and modulates heater bands to maintain setpoint. In theory, that works. In practice, the barrel temperature might sit at 190°C while the melt exiting the die is actually 178°C. That 12-degree difference is enough to change viscosity by 15 to 20%, which throws off parison weight, blow behavior, and final wall distribution.

The problem isn't the sensor or the heater. It's the thermal mass between them. The steel barrel absorbs heat, the polymer film coating the barrel insulates it, and the screw action generates shear heat that varies with throughput. The thermocouple measures the barrel wall temperature, not the melt temperature. And the melt temperature is what actually matters for product quality.

Die temperature is even worse. Thermocouples sit in drilled holes a few millimeters from the die face. The material flows past that point in seconds. The sensor lags behind actual melt temperature by 5 to 10 seconds during normal operation, and up to 30 seconds after a setpoint change. By the time the controller sees the temperature drop and cranks up the heater, the melt has already cooled and the parison is too heavy at the start of the next cycle.

Cooling water temperature in the mold is another blind spot. The machine controller usually reads the chiller outlet temperature, not the mold surface temperature. Water flow varies across circuits due to scale, pressure drops, and valve wear. One zone of the mold might be running 15°C warmer than the controller thinks, causing uneven crystallization and warpage that looks like a material problem but is actually a cooling problem.

Choosing Sensor Placement That Gives You Real Data

Thermocouple placement makes or breaks closed-loop control. A sensor in the wrong spot gives you stable-looking data that tells you nothing useful about the process.

For barrel temperature, move the thermocouple as close to the die adapter as possible. The last heating zone before the die is where most temperature variation matters for melt consistency. Some machines allow retrofitting a sensor directly into the die adapter — this gives you the closest reading to actual melt temperature without installing an expensive infrared pyrometer.

For die temperature, don't rely on the standard drilled-hole thermocouple. Add a contact probe that touches the die lip or the mandrel tip. These points experience the same thermal environment as the parison. A spring-loaded thermocouple pressed against the die face reads within 2 degrees of the actual melt temperature at the point of extrusion. The response time drops from 10 seconds to under 2 seconds, which is fast enough for meaningful closed-loop control.

For mold cooling, place temperature sensors in the return water lines of each cooling circuit, not just the supply. The temperature rise across the mold (supply minus return) tells you how much heat each zone is actually removing. If the return temperature is close to the supply temperature, that circuit isn't cooling effectively — maybe due to air lock, scale, or low flow. The controller can't compensate for what it doesn't measure.

Infrared sensors aimed at the parison or the mold surface give you a non-contact option for die and mold temperatures. Modern IR sensors are accurate to within ±1°C and respond in milliseconds. The downside is they need a clear line of sight, and dust or vapor from the mold can interfere. Use them as a secondary check rather than the primary control input — they're great for spotting hot spots but less reliable for steady-state control.

Tuning the Controller for Polymer Thermal Behavior

Standard PID tuning from the machine builder assumes a simple thermal system — heat in, temperature up, heat off, temperature down. Polymer extrusion doesn't work that way.

The screw generates shear heat that changes with RPM and back pressure. When throughput increases, the screw works harder, generating more internal heat. The barrel temperature might actually drop even though the heaters are at full power, because the polymer is carrying heat away faster than the heaters can replace it. A standard controller interprets this as "not enough heat" and pushes the heaters harder, causing overshoot when throughput drops.

Use a feed-forward control strategy alongside feedback. Feed the screw speed and throughput signal into the temperature controller as a feed-forward term. When RPM increases, the controller pre-emptively reduces heater power because it knows the screw will generate more heat. When RPM drops, it increases power before the temperature actually falls. This feed-forward plus feedback combination handles the dynamic thermal load much better than feedback alone.

Tune the integral term conservatively. Polymer thermal systems have long time constants — it takes 2 to 5 minutes for a barrel zone to fully respond to a heater change. If the integral gain is too high, the controller will accumulate error during that lag and then overcorrect massively when the temperature finally catches up. You get oscillation with a period of 5 to 10 minutes that ruins cycle-to-cycle consistency.

Start with proportional-only control. Set the gain so the system responds to a setpoint change within 3 minutes without overshooting by more than 3°C. Then add a small integral term — just enough to eliminate steady-state offset over 10 to 15 minutes. Leave derivative action off entirely. The noise in thermocouple signals combined with the long thermal time constant makes derivative term useless and often destabilizing.

Managing Die Temperature Independence from Barrel Zones

The die is the most critical temperature point in blow molding, and it's usually the worst controlled.

Barrel zones are massive thermal masses with big heaters. They're slow but stable. The die is a small metal block with tiny heaters trying to maintain temperature against a constant stream of relatively cool polymer. It's like trying to keep a cup of coffee hot by waving a hair dryer over it — the heat input is tiny compared to the heat loss.

Give the die its own dedicated controller, separate from the barrel zones. The die controller should run faster — scan every 0.5 seconds instead of every 2 to 5 seconds like the barrel. Use a higher proportional gain because the die has less thermal mass and responds quicker. But keep the integral term even smaller because the die temperature is more sensitive to throughput changes.

Install a thermal break between the die heater and the barrel. Without one, heat conducts up from the die into the last barrel zone, confusing both controllers. The barrel controller thinks the last zone is too hot and cuts power, while the die controller thinks the die is too cold and adds power. They fight each other, and the die temperature oscillates.

A ceramic or air gap insulator between the die and barrel adapter solves this. It decouples the two thermal zones so each controller works on its own system without interference. Most modern machines have this built in, but older equipment often doesn't. Retrofitting it is one of the highest-impact upgrades you can make for temperature stability.