Model identification of nonlinear sputter processes

Abstract

A nonlinear control-oriented model for sputter processes based on artificial neural networks and ordinary differential equations is developed. The process is analyzed by use of first-principle models to approximate the process structure. Hence, sputter processes can be described by a static nonlinearity, which results from the plasma processes, and linear and nonlinear dynamics that represent the actuator systems. The experimental identification with a validation of the proposed process structure, the identification of the parameters and the validation of the identified model are discussed. The experiments demonstrate that the identified model predicts the process behavior accurately and can be used for model-based control design.