Diagnosis and Prognosis Background – ICSL

Interest in diagnosing and prognosticating faults in engineering systems is as old as engineering systems. Designers and users alike have an interest in preventing the occurrence of failure of a mechanism, a machine or any kind of device –i.e. an engineering system. Attempting to detect a fault before it becomes a failure intends to eschew the execution of corrective maintenance, i.e., we would like to do something before the system fails and needs repair by applying either preventive maintenance or predictive maintenance. Preventive maintenance, or PM, typically refers to performing regular, scheduled operations that keep the system running reliably. Predictive maintenance, or PdM, on the other hand, attempts to defer maintenance operations until the time they are required.


Taking into consideration the costs of maintenance and the safety and reliability of the different approaches to engineering system maintenance, as evidenced by the table above, it is clear that the Predictive Maintenance approach is usually preferable. The proposed research falls within the area of concern of Predictive Maintenance, and, more specifically, within such in-vogue efforts known as Condition Based Maintenance (CBM) and Prognostics and Health Management (PHM), which are techniques that utilize automated strategies for detecting a fault and provide ways of extending the uptime of damaged engineering systems. Within the arena of CBM and PHM we further find both Diagnostics and Prognostics Engineering.

Diagnosing an engineering system involves three activities. First a fault must be deemed to exist through fault detection techniques. Second, the fault is located through a process known as fault isolation. The concern of many industrial diagnostic systems spans over these two activities exclusively, so that their practice has become known as FDI. The third and final activity is fault identification, through which the severity of the fault is assessed either qualitatively or quantitatively.

Prognostic systems, on the other hand, are expected to provide predictive information about the Remaining Useful Life (RUL) or Time-to-Failure (TTF) of a deteriorating machine or machine component. Because a prognosis is thus inherently uncertain, in addition to this indication, a prognosis is typically expected to also offer some representation of the amount of uncertainty in the prediction.