基于神经网络的某鱼雷保障设备故障预测方法

摘要

In order to insure torpedo guarantee equipments in good status, a fault predication method of an equipment for torpedo guarantee based on radial basic function (RBF) neural network is put forward. Based on non-linear modeling capability of RBF neural network, a network diagnosing model is constructed by a set of measuring points in the equipment, the parameter estimations are given by learning the model, and the fault is judged by the actual output of model, predicate fault of the equipment under Matlab simulation environment, the predication result is basically concordant with the actual situation. The simulation result shows that the RBF neural network can resolve fault predication problem of the equipment, which possess exact and fast diagnostic ability, and provide scientific reference for preventive maintenance of complex equipment.%　　为确保鱼雷保障设备处于良好的工作状态，提出一种基于径向基函数(radial basic function，RBF)神经网络的某鱼雷保障设备故障预测法。利用RBF神经网络的非线性建模能力，在某鱼雷保障设备的关键监测点建立网络诊断模型，通过对该模型的训练学习，确定需要的参数估计，再根据该模型的输出值来判断故障，并在Matlab仿真环境下对该设备故障进行了预测，其预测结果与实际情况基本一致。仿真结果表明：RBF 神经网络作为预测网络能较好地解决该保障设备的故障预测问题，具有较准确和快速的诊断能力，可为复杂设备的预防性维修提供科学依据。