抄録
Precise process temperature control of 0.001℃ under noise-temperature change of 0.1℃ is required in semiconductor manufacturing process. We studied control methods to minimize temperature change at an object position in a 2 dimensional vertical plate with a varying noise-heat-generation, natural convection cooling, and a control-heater. We numerically calculated 2 dimensional unsteady thermal conduction in the plate with feedback control, feed-forward control, and model predictive control of the control-heater. The temperature change at the object position can be decreased 1/80 times smaller than that without control-heat-generation by the feedback control with two monitoring temperatures. The temperature change at the object position can be decreased 1/1000 times (0.002℃) by the model predictive control of 5s interval with step response pattern and two monitoring temperatures. We found that the accuracy of the predictive model is very important for precise temperature control. Experiment was performed for the model predictive control with network model pattern, and the experimental result agreed with the calculation result.
