基于遗传算法优化神经网络刀具磨损监测研究

Research on Tool Wear State Monitoring Based on Genetic Algorithm Optimized Neural Network

通过对车刀磨损状态监测技术的研究,建立了基于遗传算法(Genetic Algorithm, GA)的BP(Back propagation)神经网络刀具磨损状态识别模型。选择振动信号和AE信号作为实验监测信号,对采集的振动信号和AE信号分别采用不同的方法进行分析,最终提取到与刀具磨损相关性强的特征作为原始特征。采用Relief-F算法对原始特征进行特征过滤得到最终特征样本。将测试样本输入训练后刀具磨损状态识别模型并查看识别结果。结果表明,识别模型的正确识别率达到96.296%,表明建立的GA-BP模型对车刀状态识别具有很好的分类效果。

Through the research on the monitoring technology of turning tool wearing state, a Back Propagation neural network tool wearing status identified model based on genetic algorithm is established. This paper chooses the vibrational signal and AE signal as the experimental monitoring signal. The vibration signal and AE signal are analyzed by different methods, and finally the feature of strong correlation with tool wearing is extracted as the original feature. The Relief-F algorithm is used to filter the original features to get the final feature samples. Input the test sample into the trained tool wearing status identified model, and check the recognition result. The correct recognition rate of the identified model is 96.296%, which indicates that the established GA-BP model has a good classification effect on the recognition of turning tool wearing state.