MODEL-BASED DIAGNOSIS OF SPECIAL CAUSES IN STATISTICAL PROCESS-CONTROL

Industry has recognized that effective use of automated diagnostic sof tware can greatly enhance process quality and productivity. Simultaneo usly, significant advances have been made in the technologies of proce ss modelling, using techniques such as neural networks, regression met hods, and various analytical approaches. Here we will present a simple method to perform model-based diagnosis. The method is simple to impl ement, intuitively appealing, and requires information that should be standardly available. The method requires as input current process dat a, set-point information, and a predictive process model, and outputs a table of diagnostic scores which indicate the likelihood of a partic ular factor being the cause of an observed special cause on a statisti cal process control chart.