| With the development of artificial intelligence (AI), there are the diagnosis techniques in a new phase , which is intelligent diagnosis phase. In this paper, the fault diagnosis problems of military electrical apparatus mechanism are investigated using expert system and neural networks. For the characteristics of the diagnosis object, the knowledge express method based on production rule and neural network are adopted as the knowledge express method. The acquisition of knowledge is regarded as a "bottleneck" problem in the construction of the expert system , a acquisition method using relational database is proposed to overcome the bottleneck problem in this paper. the knowledge management is independent of the expert system, the hierarchical structure of the object is considered in this method, a high-performance knowledge acquisition module is constructed, the query method of the relational data is applied into the knowledge searching at the same time. The reasoning process is accomplished more simply, quickly and easily as this method is adopted. In this paper, a expert system model based on the rules is also constructed, which has the better modularity and expandability. The experimental results demonstrate that this system is an effective fault diagnosis method.A two level BP neural network expert system is constructed in this paper, the network is trained using the sampled fault data, the hybrid intelligent diagnosis to some fault is realized sequentially. During the training process of the fault sample, some factors which have the impact on the network performance are analyzed, such as the number of the concealed level nodes and learning effectiveness etc, this training process can provide the scientific basis for the correctly chosen of the network parameter. In this paper, the diagnosis results show that all the fault sample can be convergent to the setting error value using the hybrid intelligent diagnosis system proposed , and the network has the generalization ability.
|
PDF Full Text Request