Re-engineering legacy tech manual's troubleshooting procedures into smart model-based diagnostics

摘要

Technical Manuals used for field maintenance of US Army systems rely heavily on troubleshooting procedures which are presented in "flow chart" format. These flow charts guide the technician through test procedures to isolate the cause of an equipment malfunction. These procedures are static, that is they are highly structured around a pre-determined sequence of tests, do not become "smarter" over time with historical maintenance data and they only take into account those symptoms and faults which the original developer considered. They are often incomplete, sometimes wrong, and are very difficult to update and maintain. As the Army moves to computer-assisted methods of maintenance, the opportunity exists to significantly enhance the basic logic and knowledge representation underlying troubleshooting procedures. The enhancements include knowledge based reasoning about faults related to symptoms, the ability to dynamically relate faults to symptoms, the ability to use historical maintenance data to continuously improve maintenance capability, and the ability to house "expert" diagnostics information in a form that becomes usable and available to novice technicians. However, can tree-based troubleshooting logic from legacy systems be efficiently re-engineered to a knowledge-based system with the same benefits? This paper addresses the aspects of transitioning from flow-chart intensive knowledge representation to a format which provides the benefits described above.