Integrated parameter and tolerance optimization of a centrifugal compressor based on a complex simulator

We apply integrated parameter and tolerance design (IPTD) to optimize the nominal values and tolerances of 10 geometric design variables of a centrifugal compressor based on a computational fluid dynamics simulator. The compressor quality is quantified by the conformance probabilities of three performance measures (PMs) and optimizing the simulator directly is time-consuming. Gaussian process (GP) emulators are constructed to predict the PMs. A conservative conformance probability estimator penalized by its standard deviation and an efficient method to compute it are developed to facilitate optimization. Eight tolerances are loosened and two are tightened. Design variable effects on quality are better understood.