Seite - 173 - in Adaptive and Intelligent Temperature Control of Microwave Heating Systems with Multiple Sources

5.3. ResultsofDifferentControlMethods reflects the robustness of the system identification algorithm and the MPCmethod. A phenomenon is worthy noted that in former results (figures 5.23 and 5.25), the highest temperature is mostly measured in the central part of the workpiece (T3). But in the second setup, the hot spot is possible to occur in any workpieces. For example, in the experiment showninfigure 5.30athehighest temperature isalwaysT4. Thisphe- nomenon denotes the same cooling effects of these five workpieces and also confirms that the active temperature distribution improve- ment isgeneratedbytheMPCmethods. NeuralNetworkbasedControl The SPSA based NNC method has also been tested using the same setup and temperature measurement scheme shown in figure 5.21. Before its real implementation, it was firstly tested in a number of simulations. These simulations use the similar idea that is the sys- tem identification part, which is that a NN was trained based on real experimental data to act as the plant, and then the NNC was applied tocontrol thiswell-trainedNN.Thestructureofsimulations isshown infigure 5.31. Simulated Plant (Well-trained NN) NN Controller NN Estimator Target Temperature Measured Temperature Control Input Estimated dynamics of the simulated plant Figure5.31. Structure used in the NNC simulations (the NN estimator was trained based on experimental data from the new CA3 with 8 heatingsources). 173