Cause associator network for fuzzily deduced conclusion in process control

The authors describe a neural network which serves a back-end tool of a process diagnostic system. The main idea is to have the network associate assignable cause(s) to the state of nonrandomness of a process. The back-end neural network relates a fuzzily deduced pattern to plausible cause(s) in a frequency trimming process. The overall framework of the diagnostic system for the process is described. Then, the pattern-cause associator neural network is outlined, and issues of initial training and how self-adjustability can be achieved are discussed. The results are especially pertinent to the assembly of a crystal resonator