Experimental Study on Model Predictive Control to Minimize Temperature Change of Vertical Plate With Varying Heat Generation

Precise temperature control of 0.001°C under noise-temperature change of 0.1°C is required in semiconductor manufacturing process. We made an experimental apparatus of a vertical steel plate placed in an atmosphere with a varying noise-heat-generation and a control-heater. The noise-heat-generation is that the heating-OFF and ON every 300 s, and it makes temperature change of 3°C at an object position in the plate without control. The control-heater is controlled with the model predictive control method of 5 s interval with two monitoring temperatures to minimize temperature change at the object position in the plate. In this work, we study the effect of the dynamic predictive model on the temperature change at the object position and examine how to make the best dynamic predictive model. Three methods to make the dynamic predictive model are examined: (1) dynamic step responses are obtained by experiment, (2) dynamic step responses are obtained by calculation with a network model of the object, and (3) both step response patterns are combined. When the step response patterns obtained by experiment and calculation are combined to use, the minimum temperature change at the object position is 0.06°C and 1/50 times smaller than that without control. Also, the effect of artificial error in the dynamic predictive model on temperature change at the object position is examined by numerical simulation.