Developing a Knowledge Base for Detection of Powertrain Failures by Reversibly Seeding Engine Faults.

摘要

Engine performance is traditionally measured in a dynamometer where engine speed, torque, and fuel consumption measurements can be made very accurately and environmental conditions are well controlled. Durability testing is also carried out in a dynamometer to assess reduction in engine output due to normal aging. However, the symptoms associated with incipient failures are not often studied since it requires either stressing engine components above their recommended limit or exchanging parts of known deviation with normal ones. This work describes a methodology for seeding faults in an engine by electronic means so that they can be reversibly turned on and off in a controlled fashion. The focus is on seeding faults that produce changes in engine output so that comparison between precise measurements done with laboratory instruments may be compared with estimates derived from on-board measurements. Thus, we have relied on a rather broad spectrum of measurement capabilities implemented in the dynamometer in order to acquire comprehensive information on the normal and abnormal behavior of the engine. A variety of engine parameters from the PCM, from add-on sensors and other instrumentation can be recorded and analyzed to detect statistically significant changes induced by the seeded faults. Thus, it is possible to build a knowledge base of measurable symptoms of abnormal behavior and study whether they could also be detected with practical on-board devices for implementing Condition Based Maintenance of powertrain systems.