An adaptive run-to-run optimizing controller for linear and nonlinear processes

Manufacturing processes encounter many hindrances to fulfilling operational equipment efficiencies. In the Introduction to this book we noted that two causes of trouble on the manufacturing floor are typically the batch-to-batch or “run-to-run” (R2R) drift of the actual output performances from the desired level, and also an inherent inability of the process to deliver consistent high quality. This last-stated problem has usually been tackled in an off-line manner by statistically designed optimization experiments. Designed experiments are very effective for developmental work, but are seldom used in an actual production situation. The first-stated problem (i.e., R2R control) of tool drift, or shift, has traditionally been managed by the tool operator tweaking various recipe setpoints of the process, such as temperature. The problem here lies in varying operator experience levels and attitudes to control. In addition, it is typical that only one input gets tweaked, whereas, in truth, a number of inputs affect the various process outputs in some multivariable, cross-correlated manner.