Fault Isolation by Comparison of Event Lists Using a Weighted Distance

A decision support system for operators monitoring complex systems is proposed. It consists in a fault isolation method based on pattern matching using binary information, in this case event lists. A training set composed of faults is used to create fault templates. Event lists generated by unknown faults are classified by comparing them with the fault templates using an original weighted distance. Each event is weighted using two independent criteria, one evaluating the frequency of appearance of the event in a given fault and the other evaluating the frequency of appearance of the same event in the other faults. The method was validated using data obtained from a realistic simulator of CERN Large Hadron Collider process and showed good performance.