ADDF observer-based adpative robust control of manufacturing processes
The purpose of this work is to develop and implement observer-based controllers for two nonlinear and time-varying manufacturing processes with the intent of demonstrating the level of performance provided by modern control techniques. Many manufacturing processes are complex in nature, and efforts to control them with simple linear techniques can perform poorly or fail. Recent developments in adaptive and robust nonlinear controller and observer theory make possible a systematic approach to robust and high performance control of such systems. ^ One suitable nonlinear controller and observer combination is described in this work. The controller, termed Adaptive Robust Control (ARC), is adaptive to maintain performance despite process variations and robust to significant disturbances. The Adaptive Divided Difference Filter (ADDF) is used for state estimation, and is designed to estimate nonlinear systems while remaining stable in the presence of significant model error and measurement noise. Numerous simulation examples are provided to demonstrate the general applicability and performance of the proposed controller/observer combination.^ Three experimental implementations of the ADDF-ARC approach are then presented. First the friction stir welding process (FSW) is described. The axial force of the FSW process is correlated with the weld quality, and varies significantly during the welding process in spite of constant process parameters. An ADDF-ARC controller for axial force of FSW is design and implemented, and experimental results are given. Next the control of machining contour error (CE) is considered. Suitable system models are developed and a controller is designed utilizing a nonlinear global task coordinate frame (GTCF) representation of CE. Simulated and experimental results are given. In the last case the CE control problem is extended to include the simultaneous control of average cutting force. A controller is designed, and simulation and experimental results are provided. In each case the proposed control approach significantly improves the process results.^
Yung C. Shin, Purdue University.