基于免疫神经网络的数控机床故障诊断研究

Research on CNC Machine Tool Fault Diagnosis Based on Immune Neural Network

针对BP神经网络在数控机床故障预测中出现的收敛速度慢和训练容易陷入局部极值问题,提出一种基于人工免疫算法优化BP神经网络(IMBP)的数据机床故障诊断算法。介绍了常见的数控机床故障类型和分类,阐述了人工免疫算法和BP神经网络以及人工免疫优化BP神经网络算法的工作流程。利用免疫算法的全局搜索性能先对神经网络权值和阈值进行全局优化,加快了BP算法训练过程的收敛速度,减少训练过程所需要的时间。通过仿真性能测试分析,结果表明:与BP、GABP和IMBP 3种算法对比,比BP神经网络算法的数控机床故障预测诊断提高了18.3%,比GABP神经网络算法提高了12.05%,提高了数控机床故障诊断精度。

Aiming at the problems of easy to fall into local extremism in training and slow convergence speed appeared during computer numerical control （CNC） faults forecasting in BP neural network, an optimizing BP neural network method （IMBP） of CNC machine tool fault diagnosis based on artificial immune algorithm was proposed.The common types and classification of CNC machine tool faults were introduced, and the workflow was described of the artificial immune algorithm, the BP neural network, the optimized BP neural network based on artificial immune algorithm.By using the global search capability of immune algorithm to global optimize the weights of neural network and threshold values, the training process and the convergence speed of BP algorithm were quicken up, and the time in need of training process was reduced.Through the analysis of performance and simulation test, compared with three al-gorithms of BP, GABP and IMBP, the results show that the fault diagnosis forecasting of CNC machine tool is improved by 18.3%as compared to BP, and 12.05%as compared to GABP, which improve the accuracy of fault diagnosis of CNC machine tool.