Evolutionary Process Optimization Methods for Multi-stage Manufacturing Systems with Applications to Semiconductor Yield Ramp-up

Over the years, different techniques have been
developed to address issues of continuous process
improvement during production. These techniques are
developed in different academic communities initiated by
different concerns. However, a common fate of these
techniques is the negligence by industrial practitioners. An
important reason was that the techniques’ unbearable
mathematical sophistication had diminished their
applicability during the age of expensive computing
resources. Another reason was that these techniques seemed
to over-shoot the needs of the relatively simple business
model and manufacturing processes at that time. In the first
year of this project, we re-examine these techniques,
including Evolution Operations (EVOP) and Optimum
Experimental Design (OED) and Ridge Analysis (RA)
developed by the applied statistics community,
Perturbation-based Real-Time Optimization (PRTO) by
mathematics and control communities and Ordinal
Optimization (OO) techniques by the discrete-event control
community. After reviewing these methods, we propose an
integrated, comprehensive evolutionary optimization method
that can be easily applied in practice.