Process variation analysis employing artificial neural networks

The artificial neural network (ANN) process modeling methodology is extended to model the variability of a synthesized process. The approach described expands on the evolutionary operation (EVOP) procedure developed by G.E.P. Box and also strives to minimize process variation. Useful features as well as disadvantages of the ANN average and noise modeling approach are summarized. The approach is judged to be capable of modeling multi-input systems and system variance conveniently and uses data sampled at different input settings to achieve experimental efficiency.<>