Event-based fault detection of manufacturing cell: Data inconsistencies between academic assumptions and industry practice

Some problems with event-based faults in manufacturing systems cannot be handled by existing fault detection solutions, including finding faults in event-based data for systems for which limited information is known. A new fault detection solution that finds faults in event-based data using model generation is presented here. This solution assumes that some information is known about the system from its design information and data structure. An example application of this solution is presented for a Ford machining cell that has been experiencing a gantry waiting problem. In the course of this example application, five inconsistencies were found between relatively common academic assumptions made by this fault detection solution (as well as others) and the actual cell's set-up and data. These inconsistencies and possible means of addressing them are discussed. Some of these means to resolve the inconsistencies have been implemented, and preliminary results in generating models using the fault detection solution are presented.