Diagnosing Multiple Simultaneous Faults

Abstract

Most recent studies of human diagnostic reasoning, or “troubleshooting,” have concentrated on the human's ability to diagnose single faults in a system. Little attention has been paid to human diagnosis of multiple simultaneous faults. Multiple faults are usually functionally unrelated, but sometimes they interact with each other, resulting in potentially confusing symptoms. An experiment was conducted to test the relative difficulty of diagnosing multiple faults whose symptoms interacted to a lesser or greater degree. The experimental test-bed was a simulated binary adder (logic circuit) into which one or more faults could be inserted by the experimenter. Four levels of objective multiple fault difficulty were identified based on the type of evidence available and the type of reasoning required to successfully diagnose the fault. Subjective mental workload and the number of tests required for each diagnosis increased as objective multiple fault difficulty rose from level 1 to 4. The time taken for a diagnosis increased as difficulty rose from level 2 to 4, but level 1 was higher than expected, suggesting difficulties that were not captured in the initial classification. Further analysis has explored these finding through performance measures and verbal protocols. Overall, this research indicates that multiple faults differ in how readily they can be diagnosed. However performance and subjective mental workload depend upon a number of factors that are now just starting to be understood, including the ease and speed of extracting the sufficient evidence for diagnosis, prior exposure to the constituent faults, and the testing strategy used by troubleshooters.