Page - 177 - in Adaptive and Intelligent Temperature Control of Microwave Heating Systems with Multiple Sources

5.3. ResultsofDifferentControlMethods In both MPC methods, only the parameters in the system estimator have to be updated. In other words, as long as the system estimator is able to estimate and predict the dynamics of the plant perfectly, the corresponding control solution is also perfect or guaranteed. But in the NNC system, both the controller and the system estimator have to be updated and trained during the heating process. Since the con- troller is a NN trained by the unsupervised learning scheme, it takes muchlongertimetoreachthesamelevelofcontrolperformanceasthe MPC methods. Nevertheless, a well trained NNC can achieve even better control performance than the MPC methods, without any ex- tra requirements of the setup or the material. There is also another big advantage that the NNC method outperforms the MPC methods, which is that the NNC method is model-free. It can be implemented tocontrolboth linearandnonlinearplantsandnoadditionalphysical insight is needed. Therefore the NNC method could be more easily transferredtoanyothercontrolapplications. 5.3.2. PerformanceofIntelligentControlMethods Compared with the adaptive control methods, the implementation of the intelligent control method is more complicated. Additional Lab- VIEW operations are required to obtain the coordinates of the max- imum and the minimum temperatures such as illustrated in figure 5.35. These coordinates are then compared with predefined region boards (such as in figure 4.13) to determine which region the high- estandthe lowest temperaturesare. Figure5.35. LabVIEW program (steps) to obtain the coordinates of the maxi- mumtemperature. 177