Abstract: In the procedure of fault diagnosis by probabilistic causal network model，the solution of the diagnostic problem can not be directly gotten by the traditional reasoning algorithm，furthermore，multi-level，multi-node complicated causal network may lead to the problem of “combinatorial explosion” and the exponential increase in computational cost etc.The competition-based connectionist model reasoning algorithm based on probabilistic causal network is put forward and the algorithm is modified.The problem of multi-level，multi-node complicated causal network reasoning is solved by the algorithm presented，and the complexity of the diagnosis is reduced.Finally，the advantages and the engineering practicability are demonstrated by a practical example.

Key words: causal network, fault diagnosis, connectionist model

摘要： 在利用概率因果网络模型进行故障诊断过程中，传统的计算方法不能直接得到诊断问题的解，且当故障节点较多或网络层次较多时，存在着“组合爆炸”、计算量呈指数速度增加等问题。本文提出了以因果网络理论为基础的关联模式竞争求解算法，并对该算法进行了改进。改进后的算法解决了多层次、多节点的复杂因果网络模型推理难题，降低了诊断复杂度。最后通过一个实例验证了此方法的优越性及工程实用性。

关键词: 因果网络, 故障诊断, 关联模式