Experimental Model Determination for Neurocontrol of a Thermal Conduction System

Recently a synthesis methodology for the infinite time optimal neurocontrollers for partial differential equations systems in the framework of adaptive-critic design has been developed. The adaptive-critic approach is applied to a thermal conduction system. The experimental setup representing a one-dimensional heat conduction problem developed by matching the experimental and simulation results using a tridiagonal-matrix algorithm is presented. The discrete domain forms of the state, the costate and the optimal control equations are derived using the distributed parameter model of the system. The synthesis is introduced of an adaptive-critic-based infinite time optimal neurocontroller for online temperature profile control. The representative states of the system for training the networks was generated using Fourier series-based smooth states profile algorithm. Finally, the action network is implemented for temperature profile control of the experimental setup.