Bayesian belief networks for effective troubleshooting

The maintenance of equipment, machinery and facilities is a vital part of the industrial process and requires millions of man-hours of technician time. A significant portion of this time is devoted to troubleshooting system malfunctions. We develop an automated system that uses Bayesian belief networks (BBNs) for effective troubleshooting. BBNs are ideal paradigms to represent the causality and uncertainty involved in troubleshooting problems. The automated system we develop generates a cost-effective sequence of testing operations. This optimal sequence generation algorithm is a unique blend of the graphical capabilities of the BBN and older constrained sequence generation algorithms. The test sequence takes into consideration the cost of testing a component and the probability of that component being faulty. An efficient graphical user interface is used to enable the user to develop the BBN and perform decision analysis. We use concepts of qualitative probability networks (QPNs), verbal mapping functions, and automated probability matrix generation to reduce the amount of input required