Page - 185 - in Adaptive and Intelligent Temperature Control of Microwave Heating Systems with Multiple Sources

formance of the distributed microwave heating system can be sig- nificantly improved by the advanced MIMO control system. The global temperature window (temperature window of the entire heat load) achieved by the RLC system (5◦C - 6.5◦C) is also com- parabletonormalcommercial industrialheatingequipments,such as thestandard in [Ind13](±2.5◦C). • The control concept that combines the real time thermal picture and the intelligent control system (Q(λ) controller) is unique. It is able to control the complete temperature distribution of the entire workpiece, and more straightforward than conventional methods basedonindividual temperaturevalues. Itseffectivenesshasbeen provedbyexperimental results. • The temperature control system and software developed in this dissertation have a great potential to be implemented in practical industrial applications, to promote the use of microwave heating to broader areas. They can also be transferred to other distributed microwavefeedingsystems,suchastheoctagonalmicrowavecav- ity (proposedin[LLHG14])andotherrelatedapplications. Based on the work done in this dissertation, the MIMO temperature control system could be further optimized in order to improve its re- liability and control performance. For example, more experiment pa- rameters could be monitored and taken into account in the modeling process, e.g. the air flow speed in the oven and the real-time tem- perature of the metal table. The involvement of such parameters will increasetheaccuracyof thecontrolmodel,aswellas thereusabilityof the controller. With a high reusability, control performance could be improved by reusing the former well-trained controller (such as the results of repeated trials shown in figures 5.34 and 5.37), especially for the reinforcement learning controller. In this case, a more power- fulreinforcementlearningcontrollercouldbebuiltbasedonacompli- cated function approximator (nonlinear function or neural network) and long time practical experiment training. Another research direc- tionwouldbethecombinationofmicrowaveandhotairheating. The hot air flow could be controlled using an independent PID controller toheat thecoldareaof the load,whichwillbeofgreathelptoachieve amorehomogeneoustemperaturedistributionbetweentheinnercore andsurfaceof theentire load. 185