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

ListofFigures 2.17. Comparisonofmeasurementdelaysusingthe fiberoptic sensor (FOS)andthe infrared camera (IRC). . . . . . . . . . . . . . . . . . . . . . . 34 2.18. Temperaturecomparisonbetween FOSand IRC (after compensation). . . . . . . . . . . . . . . . 36 3.1. Sketchof microwave heating setup. . . . . . . . . . 38 3.2. Diagram ofblack-boxmodeling. . . . . . . . . . . . 39 3.3. Illustrationof the linearsystemidentification process. . . . . . . . . . . . . . . . . . . . . . . . . . . 58 3.4. Neural network structure . . . . . . . . . . . . . . . 66 3.5. Illustrationof differentactivationfunctions. . . . . . 68 3.6. Diagram ofrecurrentneuralnetwork . . . . . . . . . 70 3.7. Principleofsupervised learning . . . . . . . . . . . . 71 3.8. Neural network approaches used in thisdissertation. 76 3.9. Notationsused to denote the input andthe outputof different nodes. . . . . . . . . . . . . . . . 78 4.1. PrincipleofMPC control algorithms. . . . . . . . . . 86 4.2. Temperaturecontrol systemof HEPHAISTOS using MPC. . . . . . . . . . . . . . . . . . . . . . . . 87 4.3. Practically implemented GAbased nonlinear MPCsystem. . . . . . . . . . . . . . . . . . . . . . . . 101 4.4. Twocontrol structuresofNN controller. . . . . . . . 102 4.5. Weightsupdate in the standard SPSAalgorithm. . . 105 4.6. Semi-directcontrolstructure. Thebluedashed linerepresents two possible learningapproaches. . 107 4.7. Procedures in thesemi-direct NNcontrol system (part 1). . . . . . . . . . . . . . . . . . . . . . 108 4.7. Procedures in thesemi-direct NNcontrol system (part 2). . . . . . . . . . . . . . . . . . . . . . 109 4.8. Reinforcement learning controller. . . . . . . . . . . 112 4.9. Actor-critic control system[GBLB12]. . . . . . . . . 121 4.10. HybridTD learningcontrol system. . . . . . . . . . 123 4.11. Target temperaturecurveduringtheheatingprocess . 124 4.12. Procedures in theWatkins’Q(λ) learningcontrol . . 126 4.13. Areadiscretization. . . . . . . . . . . . . . . . . . . . 129 4.14. Procedures in thehybrid multi-agentQ(λ) learning control system. . . . . . . . . . . . . . . . . 131 x . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .