Fault Diagnosis of a wheel Loader by Artificial Neural Networks and Fuzzy Logic

This paper employs both the technologies of neural network and fuzzy theory so that they can make up the deficiencies of each other. A hierarchy model, which employs a sub-ANN neural network deals with the tasks first then applies fuzzy treatment later, has been set up to fulfil the requirements of quick response, strong self-learning and clear expression hierarchical diagnostic. Memberships in the fuzzy part would update themselves automatically and the sub-ANN network part enables rapid and accurate learning. Diagnosis datum originate from various state parameter sensors, for instance, pressure sensors, temperature sensors, fluid level sensors, and gear box vibration sensors, etc. These parameters would be periodically updated and sent to a remote diagnosis center via Internet where they will be processed. Numeric simulation and practical applications verified the efficacy of this model. Inherent limitations in early proposed fault diagnosis systems have been alleviated which are based on the data base of knowledge regulations, such as lack of agility, unsustainable for unfamiliar circumstances, etc. Both the calculation workload and the size of the knowledge data base have been reduced