Neuronale Modellbildung und Temperaturregelung eines Vertikalofens für die Halbleiterfertigung

Due to rising costs and the call for innovations control technology in semiconductor equipment gains in importance. Compared to PID-control systems which have been most frequently deployed in manufacturing equipment so far, new intelligent control systems increase flexibility and yield. shorten ramp-up times as well as cycle times, and reduce therefore the production costs. The key word frequently used in this context is: model-based control. Since in many cases the physical model establishment of the controlled process is very complicated, a neural network serves in this work as a model which enables besides a fast model establishment also the adaptation to changing system conditions. The advantages of a predictive neural control system have been verified by simulating and controlling of an oxidation and LPCVD vertical furnace.