A field based study of troubleshooting in computer-controlled manufacturing systems

Methods are described for collecting and analyzing data derived from the observation of troubleshooting in computer-controlled manufacturing systems. The data were collected by direct observation of troubleshooting episodes and by obtaining retrospective reports from the subjects immediately following the episodes. To analyze the data, the following attributes of each activity within each episode were characterized: the information source consulted, the beliefs held, the action performed, and the reasons given. Initial findings concerning information acquisition, information utilization, and sources of difficulty in troubleshooting are presented. The authors found that troubleshooters consult a multiplicity of information sources, of which the displays to process control computers is but one of many sources referred to frequently. They also found that troubleshooters use many different kinds of information to make decisions, including information about the ease of performing various tests and historical information about occurrences at the equipment. Based on reported beliefs of possible cause during each troubleshooting activity, it was found that the troubleshooters think about the system in multiple ways, including both physical and temporal levels of abstraction. While it was discovered that difficulty can occur at any point in the troubleshooting process, much of the difficulty with diagnosing faults can be attributed to fixation on an incorrect belief or to a lack of refinement of beliefs.