Wednesday, March 31, 2021, 13:00
Automatic impedance matching is a key aspect of modern plasma technology and a challenging task because of the dynamic nature of plasma processes. There are various different hardware topologies and control algorithms addressing this problem. The here introduced state of the art algorithm suffers from sensitivity to the pre-calibrated data that is used for the control. In this thesis we illustrate a novel approach in which we use
numerical optimization techniques to estimate the calibration error from data gathered during a plasma process without manipulating the hardware. The impedances within the matchbox are modelled to allow a Non Linear Programming (NLP) formulation that estimates the model parameters. The parameter estimation is simulated and experimentally tested on a plasma chamber setup with corrupted calibration data.
Furthermore, we propose an approach to improve the matching performance by iteratively learning new controls over repeated trials using an Optimal Control Problem (OCP) formulation. For repeated processes with the same initial conditions we aim to learn controls from the information of past trials by adding the goal of a constantly decreasing absolute reflection coefficient during the matching process.
Online via Zoom
Meeting-ID: 627 9173 7415