Application of connectionist learning methods to manufacturing process monitoring

摘要: It is demonstrated that connectionist learning networks can monitor manufacturing processes to determine causal relationships with an accuracy competitive of conventional statistical techniques. Moreover, the network operates online, in realtime, and substantial savings computational complexity as compared CIM Two approaches are compared. One employs standard procedures find correlations between sensor measurements quality. The data from production line collected over a period time, made offline at infrequent intervals using analyses such linear regression. second approach estimate incrementally, collected, online real-time. estimates updated incrementally procedures. Simulation results presented for fluorescent bulb line. >