Learning of time templates from system observation

The use of time templates for discrete event fault monitoring in automated manufacturing systems gives a simple and versatile method of fault monitoring. However, ascertaining the correct knowledge of the time templates for the system is a significant issue. In this paper the authors present a simple online algorithm for the generation of templates by listening to observations corresponding to the correct operation of multiple concurrent, possibly identical, processes. Event pairs pertaining to individual concurrent process instances are segregated and the timing relationships between the events in each pair are determined. The paper also verifies several useful properties of the template sets thus determined