| Summary: | 碩士 === 國立高雄第一科技大學 === 系統資訊與控制研究所 === 96 === The traditional vehicle failure detection is basically done by the skilled technician in view of some abnormal phenomena or noises observed and heavily relied on his accumulated experiences. However, well experienced and skilled technician is often not available due to the difficulty to train one of them. Therefore, developing an automatic vehicles fault monitoring and diagnosis system has become necessary. As to vehicle fault monitoring and diagnosis, some phenomenon or signal changes are the basis for fault source reasoning, and they are also the only information that can be visually observed or measured. However, those individual information observed usually can not be used to reason out the fault source directly. Usually, the fault source that may produce those fault phenomenon also depends on the system operation state. In other words, phenomenon observation or measurement alone usually cannot directly deduce fault source, but often can only deduce some of the fault symptom. It is quite possible that combining multiple fault symptom can then the fault source be deduced. Only when the source has been deduced can then the fault be fixed. Therefore, how the signal observations, the fault symptoms, and the fault sources are related to one another is the kernel knowledge for diagnosis, and is also the main issue to be addressed in this research.
In this research a "vehicle fault symptom / source definition framework" was proposed to consolidate those diagnosis knowledge into a systematic framework, through which various fault symptom or fault source detection rule can be built into the knowledge base for diagnosis. To implement such a vehicle diagnosis system, an embedded system was designed and tested. The experimentation showed that the proposed framework is basically feasible.
|
|---|
