An Algorithm of Constructing Fault Tree Based on Ant Colony Algorithm
Present an algorithm to construct Fault Tree based on Ant Colony Algorithm. The idea conies from the phenomena of ants cluster food into a heap. An ant selects one favorite food from food heap at a time. After many times, the food heap is divided and a food chain is formed. For fault diagnosis, the system malfunction can be considered a food heap, and all attributes of the system be considered as a food chain that is described as a path of nodes, each node represents an attribute. If all paths constructed by ants are the same, that indicates all attributes related with the system malfunction are found out. If not, the further classifying should be done. Then ants again select food attributes in the new clustering. This process of clustering and selecting is continued till the paths found out by all ants are the same. A fault tree is formed step by step in the process. The attribute values are fuzzified to avoid over-fitting to training data. The Algorithm essentially is a rule-mining approach. A case to construct the fault tree of a relay is studied, the result shows that the algorithm is effective.
fault tree analysis ant colony algorithm data mining fault diagnosis relay
ZHOU Chunlai LI Zhigang
Information Engineering School of Communication University of China, China Institute of Electrical Apparatus of Heibei University of Technology, China
国际会议
The First International Conference on Maintenance Engineering(首届维修工程国际学术会议)
成都
英文
901-908
2006-10-15（万方平台首次上网日期，不代表论文的发表时间）