Neural network based yield prediction of microwave filters

In this paper, a neural network has been used to improve the computational speed of the yield prediction procedure for microwave filters, consisting of both tunable and non-tunable elements. A feed forward neural network is trained using the back propagation algorithm to predict the starting points for the optimizer used in the yield prediction algorithm. This technique has been used to study the yield of several different filter structures, producing the same electrical response. It has been found that the computational speed improvement depends on the yield of the filter and as the yield approaches 100 % the computational savings are of the order of 30 %.