A neural network based approach for surveillance and diagnosis of statistical parameters in IC manufacturing process

Presents a new approach for monitoring and diagnosing potential faults in the IC manufacturing process. A backpropagation neural network based diagnosing model is employed to synthesize the complicated mapping from process measurements to the unmeasurable process disturbances. This model is trained to detect significant shifts of the disturbances. Due to the inverse mapping diagnosis becomes very efficient and is quite promising for real time applications. Several mathematical issues involved in this approach and an illustrative example are discussed.