How to Diagnose a Platen Zone That Has an Unstable Temperature Only at a Specific PID Setting?

A heating zone on a platen is stable at its normal setpoint . When the setpoint is increased by 50 degrees , as a modification in the process recipe , the temperature starts to oscillate in a regular , rhythmic wave . Heater is OK. The villain is a PID control loop that has been adjusted for one particular thermal dynamic and has become unstable at some other operating point, usually because a derivative setting is now too aggressive for the new, slower reaction of the system.

The Setpoint-Dependent Nature of Instability
The PID (Proportional-Integral-Derivative) controller is utilised in composite curing, lamination or other thermal processes in heated platen systems to maintain a consistent temperature. Typically the controller parameters (Kp, Ki, Kd) are tuned at a realistic operating state, commonly a mid-range setpoint. However, if the setpoint is altered sufficiently (e.g. from 120°C to 180°C), the platen and the load behave thermally different. Major variations at higher set points:

The temperature difference between the platen and the surrounding air is greater hence more heat is lost to the surroundings.

Different thermal mass response as the thick platen and tooling absorb more energy before the temperature stabilises

Different temporal constants of heating and cooling of

A PID loop that was fully stable at a lower setpoint can become unstable at a higher setpoint. The oscillation is generally a regular sinusoidal wave about the setpoint with a duration of seconds to minutes. The amplitude may grow or stay the same. The heater and the temperature sensor are both working OK, the fault is completely in the control method.

Root Cause: Over-reaction of Derivative term to noise
The unstable temp specific PID setting platen fault is nearly invariably an extremely aggressive derivative setting (Kd). The derivative term of a PID controller acts on the slope of the error signal . ( i.e. the rate of change ) It measures the rate at which temperature is moving, to predict future error. If derivative action is appropriately calibrated, it improves stability and reduces overshoot.

But the derivative action is intrinsically vulnerable to high-frequency noise. Real temperature signals have modest, natural swings from electrical noise, sensor resolution constraints, and slight variations in heating. This noise is ignored if the derivative value is correctly specified. A derivative setting that is overly aggressive ( Kd too high ) tends to amplify the noise and causes the controller to make significant quick adjustments to the power output . These modifications then cause the temperature to overshoot, and the derivative term overreacts to this again, producing a persistent oscillation.

The oscillation tends to occur at a certain setpoint because:

At lower setpoints the slower reaction of the platen (due to decreased heat loss) will naturally damp the oscillations.

At higher setpoints, the system is more susceptible to the power fluctuations caused by the derivative because of faster heat loss and different thermal inertia.

The controller is overreacting fearfully to every little thermal ripple and needs a gentler, less enthusiastic touch.

Diagnostic Procedure: The derivative as the cause
The following step-by-step diagnostic method is performed with the press in a safe, operating condition. No hardware replacement is needed.

Step 1: Watch and write down the oscillation
The temperature trend of the affected zone is shown on the controller display or on the supervisory system of the plant. The following traits are noted:

Setpoint at which oscillation occurs (eg 180°C)

Time between peaks (oscillation period, e.g. 30 seconds)

Amplitude (peak-to-peak fluctuation, e.g., ± 5°C)

Pattern of power output (check the controller % output, if the output is oscillating in phase with the temperature, it is a control loop problem)

Step 2: Check The Health Of The Heater And Sensor
Hardware errors are excluded before tuning PID parameters as follows:

Check the thermocouple for the zone for loose connections or sporadic readings (as outlined in prior instructions). Noise can be injected by a loose thermocouple and amplified by the derivative term.

The contactor or solid state relay (SSR) of the heater is checked for correct switching. There is no arcing or sticking.

The controller is powered by a stable power source (no voltage fluctuations).

If the hardware is good, error is software error.

Step 3: Lower the Derivative Setting (Kd)
The value of the derivative is usually given in seconds (derivative time, Td), or as a dimensionless gain factor. Different manufacturers of controllers use different notations. The following empirical correction is made:

Record the current Kd value (e.g. 60 sec).

Reduce Kd by half (e.g. 30 seconds) It is a safe, conservative reduction and won't destabilise a well-tuned loop, but it will quiet an overaggressive one.

Watch the temperature response at the bad setpoint for at least 5-10 time constants (e.g. 10-20 minutes).

If the oscillation decays or greatly lowers in amplitude then the derivative term was the culprit. If the oscillation remains, Kd is further reduced (e.g. 15 seconds) or set to zero (derivative off). A loop with no derivative (PI only) will be stable by nature but can have a slower reaction and increased overshoot.

Step 4: If Reducing Kd Fixes the Problem, Tweak the Loop Again
For many operations a permanently reduced derivative value is acceptable. However, for best performance (quick reaction with little overshoot) the PID loop should be retuned especially at the troublesome setpoint.

Two choices exist:

Run the autotune function of the controller at the new higher setpoint. Most current PID controllers offer an autotune (or "auto-adapt") capability which automatically determines optimal Kp, Ki and Kd by exciting the system with a step change and evaluating the reaction. The autotune starts when the platen is stable and cold.

Manual retuning with a conservative stable set of settings that works over the whole predicted setpoint range . Manual tuning is tuning by changing the parameters in a systematic sequence (usually P first, then I, then D). For thick-platen systems with large thermal mass, derivative action is generally not needed; a well-tuned PI controller (Kd = 0) is stable and adequate.

Technical Accuracy: Derivative & Noise Amplification
Mathematically, the derivative action corresponds to the slope of the temperature signal across a limited time window. If the signal has high frequency noise (e.g. ±0.2°C oscillations every second) the derivative of that noise can be substantial (e.g. 10°C/second). When multiplied by a big Kd value this leads to substantial, fast changes in the controller output. The heater then reacts by rapidly switching on and off, causing greater temperature changes – a positive feedback loop.

This is the reason why unstable oscillations usually have a peculiar high frequency (smaller than the natural thermal time constant of the system). A correctly adjusted derivative ignores noise by:

A gain low enough that the noise-induced variations of the output are minor, or

•smoothing the temperature signal before differentiation with a derivative filter (low-pass filter). Most controllers have a setting called "derivative filter time constant".

Another non-invasive option is to turn on a derivative filter (if not already on) if the controller enables it.

Step 5: Check the new parameters for all setpoints
The derivative is then adjusted, or the loop is retuned, and the zone is evaluated at the initial setpoint (where it was stable) and at one or more intermediate setpoints. The answer should be steady everywhere. If the loop is sluggish (slow to reach setpoint) at lower setpoints a compromise is tolerated because stability is frequently preferred over speed in essential curing operations. The controller may also have numerous sets of PID parameters and automatically transition between them based on setpoint (gain scheduling). This is a sophisticated feature that some of the high-end temperature controllers have.

Practical Example: Stable at 120°C, Unstable at 180°C
A 300 mm thick steel platen is used for curing composite armour panels. The platen is stable at 120°C with PID parameters P=5.0, I=200 s, D=60 s. When the setpoint is increased to 180°C the temperature oscillates with an amplitude of ±6°C and a period of 40 seconds. The power output varies every 20 seconds from 0% to 100%. A check is made on the thermocouple and heater to see if they work.

Diagnosis: The derivative time (60 s) was set for the slower thermal dynamics at 120 °C. The system is more sensitive due to the more rapid heat loss at 180 °C and the derivative term overreacts.

Action: Kd is decreased from 60 s to 25 s. The oscillation stops and the temperature is stable at 180°C with a tiny overshoot (3°C) and no prolonged cycling. Then the autotune is repeated at 180°C, which gives new parameters: P=4.2, I=180 s, D=18 s. The revised parameters are stable at 120°C and 180°C.

Best Practices for PID Tuning: Preventing Repeats
Tune at the highest expected setpoint - If a process has numerous setpoints, tune the loop at the highest one. That's where it's most likely to become unstable.

Use conservative derivative values - For big, thermally sluggish systems (huge platens) derivative action is often not needed or set very low. In many industries PI control alone works just great.

Enable derivative filtering – If the controller is capable of doing so, an optional derivative filter (e.g. with a time constant of 2-5 seconds) lessens the sensitivity to noise.

Document PID parameters for each recipe - Some advanced controllers allow separate PID sets to be stored and remembered for each recipe. When the setpoint is changed, the relevant set is loaded automatically.

Perform periodic autotune The thermal properties of the platen may change over time due to fouling, sensor drift or variations in load (e.g. thicker moulds). Every year an autotune re-calibrates the loop.

Conclusion: Software is the Cure, Not Hardware Replacement
Temperature oscillation at a given setpoint is a typical control loop problem that is handled in software. The remedy is easy, change the PID parameters not the hardware. The most probable explanation for a platen zone becoming unstable only at a greater or lower temperature is the derivative term (Kd). Excessive derivative gain intensifies the natural thermal noise, forcing the controller to overreact and force the system into persistent oscillation. This may usually be addressed promptly by reducing the derivative (commonly by half) or turning it off altogether. For a more solid, long-term solution, the autotune procedure is rerun at the problematic set point or the loop is manually retuned using conservative values that remain stable over the whole working range.

The most sophisticated machine obeys a simple mathematical law that can be tuned. When dealing with hot platen control, the derivative term is a potent yet delicate instrument. The platen temperature starts to dance to an unwanted tune when it gets too aggressive. The fix is just lowering the "D" down and letting the system settle.