基于BP神经网络的火炮反后坐装置故障诊断

火炮反后坐装置的故障诊断,采用3层BP神经网络.通过基于标准梯度下降或基于标准数值优化的改进算法,同时考虑算法本身性能,及问题复杂度、样本集大小、网络规模、网络误差目标和所要解决的问题类型.并以解析算法的部分计算结果作为神经网络学习样本集,由相关方程式算出样本库内的故障现象及与故障原因相对应的训练样本,完成故障诊断.

Fault Diagnosis of Gun Anti-Recoil Mechanism Based on BP Neural Networks

The fault diagnosis of gun anti-recoil mechanism adopted three-level BP neural network. Based on standard gradient descent algorithm or improved algorithm of standard numerical optimization, algorithm performance, problem complexity, the size of sample set, network size, network error goal and problem types were considered. Some calculation results of resolve algorithm was taken as the learning sample set of BP neural network; and the faults in sample base and the training samples of fault reason were calculated by corresponding equations and the fault diagnosis was realized.