Research cooperation

94% tolerance compliance in temperature control

How aiXopt tunes a multi-zone temperature control across the whole operating range instead of for a single operating point.

Category: Research cooperation
Transferability: Approach transfers to thermal processes with several zones and changing operating conditions, such as furnaces, drying or tempering.

Challenge

Two temperature zones have to settle quickly on new setpoints without violating a tight overshoot tolerance. How strongly an actuator acts depends on the mass flow, which changes constantly during operation. A fixed controller setting is sluggish at low mass flow and overshoots clearly at high mass flow. On top of that, model deviations and external disturbances make every setting behave differently from case to case.

Solution

aiXopt's Small-Data AI learns the controller parameters not for one operating point but as a smooth characteristic curve across the whole mass flow range. Model deviations and disturbances are varied on purpose so that the setting becomes robust and does not only work in the ideal case. In validation with changing disturbances the robust characteristic curve kept the tolerance in 94% of the cases, while a non-robust setting violated it consistently.