Jiang Zhinong

Multi-dimensional fault diagnosis method based on expert thinking

The invention discloses a multi-dimensional fault diagnosis method based on expert thinking. The running state information of each measuring point is monitored in real time and on line; the possible fault types of a device are judged according to acquired basic data and failure mechanism of the alarm measuring point, so that a responding fault matrix is established; a weight matrix is established corresponding to the fault matrix according to the acquired data and the failure mechanism of the each measuring point; and the fault matrix multiplies the weight matrix to solve a result of diagnosis and give a final diagnosis, so that the condition whether the device has faults is determined. The diagnosis method can remove the signal abnormal failure of the single measuring point at the single timing, can remove the false failure of the device as much as possible, contributes to distinguishing the fault types with the similar symptom at the same time to give the most practical fault diagnosis, solves the problem of a conflict of diagnosis based on the different measuring points and at different timing, and has a positive engineering guiding effect on field apparatus management and monitoring staff.

Development of the Task-Based Expert System for Machine Fault Diagnosis

The operating mechanism of expert systems widely used in fault diagnosis is to formulate a set of diagnostic rules, according to the mechanism and symptoms of faults, in order to instruct the fault diagnosis or directly give diagnostic results. In practice, due to differences existing in such aspects as production technology, drivers, etc., a certain fault may derive from different causes, which will lead to a lower diagnostic accuracy of expert systems. Besides, a variety of expert systems now available have a dual problem of low generality and low expandability, of which the former can lead to the repeated development of expert systems for different machines, while the latter restricts users from expanding the system. Aimed at these problems, a type of task-based software architecture of expert system is proposed in this paper, which permits a specific optimization based on a set of common rules, and allows users to add or modify rules on a man-machine dialog so as to keep on absorbing and improving the expert knowledge. Finally, the integration of the expert system with the condition monitoring system to implement the automatic and semi-automatic diagnosis is introduced.