目前的故障诊断系统中普遍存在缺乏通用性、对故障先验知识与系统动态数据利用不均衡、稳定性和可塑性难以两全等缺陷,本文针对这些问题提出了一种故障模型,运用符号学习与神经网络相结合构成的增量式混合型学习算法,成功地对故障示例集进行处理,在此基础上给出了一种通用故障诊断系统设计方法。实验表明,利用该方法实现的系统在进行故障诊断时可以取得较好的效果。 At present, there are many common shortcomings such as lack of commonality, unbalanced utilization of fault prior knowledge and system dynamic data, stability and plasticity, and so on. In this paper, a fault model is proposed to solve these problems. Using symbolic learning and Neural network to form an incremental hybrid learning algorithm, the successful failure of the sample set to deal with, and on this basis, a general fault diagnosis system design method is given. Experiments show that the system using this method can achieve better results in fault diagnosis.