Combining implicit and explicit knowledge to map and learn from infrequent incidents in support of reliability assurance

Statistical process control is commonly used to methodically measure parameters of significance in product or process performance. However in one-off production or at low production rates it may take a considerable time before a statistically significant sample is collected, if useful data samples can be collected at all. In other situations, for example in monitoring product performance, logging low incidence indications may be used to forecast potential product failure. Often the knowledge of experienced people must be relied upon to supplement recorded data. It is argued that this knowledge requires some context to facilitate its effective utilization, and that this context must be its relation to a physical object or process, leading to some form of information map. Ultimately, the outcome is a decision to take some system improvement action. Three case examples are presented. A model of a system as four generic subsystems" a physical subsystem, an information subsystem, a decision subsystem and a knowledge subsystem is used to characterize three cases. The design of suitable mapping systems is discussed.