竞争学习模糊神经网络及在导弹故障诊断中的应用

尽管基于解析冗余的故障诊断方法有许多突出的优点而越来越多地得到的研究和应用，但它依赖于系统的模型，当系统存在非线性或不确定性时，存在难以建模的困难，模糊神经网络可以通过学习建立系统的模型，且模型参数有明确的物理意义，初始参数易于选择，成为解决这一问题的优选途径，作者通过把模糊神经网络的学习转化为竞争聚类和线性优化问题，基于竞争聚类和最小二乘原理，提出了一种模糊神经网络学习算法，并在某伺服机构上进行

FUZZY NEURAL NETWORK BASED ON COMPETITIVE LEARNING AND ITS APPLICATION TO FAULT DIAGNOSIS OF MISSILE

The fault Diagnosis based on analytical redundancy depends on system's model,while it is difficult to get the system's model when the system is non linear or uncertain time varying,there are some difficulties to put this method into application of non linear system.The fuzzy neural network can build the system's model by learning,and the model's parameter established by this method has clear mean,it provide an effective way to solve the above problem.The key of the Application of the fuzzy neural network is the determination of its parameters.Different from the other researchers,the authors convert the learning process to cluster and linear optimum,based on competitive learning and least square error criterion,suggest a learning algorithm for the fuzzy neural network system,the experiment in certain servo mechanism get very good result.