文中第一部分研究了利用Kohonen 网络进行发动机故障诊断的特点与算法。并以JT9D发动机为例对算法的有效性进行了检验,23 个故障样本的确诊率达到87% 。文中还以一个典型例子进行了详细分析。文中第二部分提出了利用Kohonen 网络进行故障诊断结果排序的方法,该方法有助于给出简明的诊断结论。文中第三部分提出了利用Kohonen 网络提取故障样本群的代表性样本的方法。该方法对于经验故障方程的建立十分有用。研究结果表明,自组织映射模型程序简单、应用方便,诊断结果不受训练样本的病态影响,特别是能够以非常简单而直观的方式揭示样本群的统计特性,是一种适用于发动机分类故障诊断的较好的、具有特色的神经网络模型 The first part of this paper studies the characteristics and algorithms of engine fault diagnosis using Kohonen network. Taking JT9D engine as an example, the validity of the algorithm was tested, and the diagnosis rate of 23 fault samples was 87%. The article also a detailed analysis of a typical example. The second part of the paper presents a method to sort the fault diagnosis results using Kohonen network, which can help to provide a concise diagnosis conclusion. The third part of the paper proposes a method of using Kohonen network to extract the representative samples of fault samples. This method is very useful for establishing the empirical failure equation. The results show that the self-organizing map model is simple and easy to use, and the diagnosis results are not affected by the pathological changes of the training samples. In particular, the statistical properties of the sample groups can be revealed in a very simple and intuitive way. The better, featured neural network model