Hybrid spectral/intelligent modeling and decoupling control for a nonlinear distributed parameter system in IC packaging

An intelligence-based modeling method is proposed for the curing process, a nonlinear parabolic distributed parameter systems (DPSs). The proposed modeling method combines model reduction techniques and intelligence based identification methods of nonlinear ODE (ordinary differential equation) systems. The developed neural network model is of state space form and fits the general model-based controller formulations, so that the control techniques used for ODE models can be directly applied. Moreover an approximate decoupling-control is developed to replace PID controllers originally equipped for the cure process for a better cure performance.