基于神经网络的机械磨损故障光谱定位诊断法

Diagnosis of Wear-Induced Breakdown of Machine by Spectrometric Analysis Based on Artificial Neural Network

在分析常用光谱定位诊断方法的基础上提出了基于神经网络的光谱定位诊断法;将机械摩擦副材质的元素含量作为神经网络输入,将材质所对应的部件作为神经网络输出,建立了相应的神经网络训练样本;通过整理训练样本和训练神经网络,利用神经网络超强的非线性映射能力和容错性实现了磨损故障部位诊断;通过算例分析验证了所提出的诊断方法的可行性和准确性.结果表明,所建立的方法简洁有效,并具有很高的诊断精度.

The spectrometric method to diagnose the wear of frictional parts based on artificial neural network (ANN) was established on the basis of analyzing commonly used spectrometric localization diagnosis methods. Thus the training samples were established using the elemental compositions of the frictional pair materials as the inputs of ANN and the corresponding frictional parts as the outputs of ANN. The diagnosis of the wear failure locations was realized by coordinating the training samples, training the ANN and making use of the powerful non-linear mapping ability and the error-tolerating ability of the ANN. The precision and the feasibility of the established diagnosis method were validated by the analysis of some examples. It was found that the established diagnosis method was applicable to diagnose the wear status of frictional parts with convenience and good precision. Cr4Mo4V, 2Cr3WMoV-1, 1Cr18Ni9Ti, H62 and QA10 are taken for example in the research and analysis.