Seite - 130 - in Adaptive and Intelligent Temperature Control of Microwave Heating Systems with Multiple Sources

4. ControlSystemDesign neglected. Because they locate below the setup and their heating effects are weaker than the other 4. In total 12 different heating sourcesareused. 2. All control combinations consisting of more than 2 activated sources are neglected, because 2 heating sources (maximum 4 kW power) is enough to keep the workpiece of middle sizes (1m2 ∼ 2m2) at the target temperaturerange(70◦C∼100◦C). 3. After the first two screenings, all left control combinations are fur- ther tested and the combinations that only heat the center region of theworkpieceareeliminated. Followingtheabovethreerules,30differentcontrolactionsarefinally selected to generate the action space. Combining the state and the ac- tionspacestogether, thereareentirely1830(61×30)differentQvalues to be estimated and saved by theQ(λ) learning controller. In theory, even the control period is 1 s, it need at least 1830 s to go through the overall state-actionspace. But inpractice, thehotspotsandcoldspots arenotmovingthroughthewholeworkpieceduringoneheatingpro- cess. It often happens that the states of the system stay in a limited number of states and never go to other states, which means the num- ber of practically possible state-action space is much smaller than the theoretical value. In this case, the control task is realistic to be solved inreal time(seetheresultsshowninchapter 4). Thecompletecontrol procedures of the hybrid multi-agent control system are shown as in figure 4.14. 130