Abstract: Effective identification of tool wear and wear type is an important means to monitor machining status.Aiming at the lack of manual participation in the identification of conventional tool wear types,an intelligent recognition method based on convolutional neural network tool wear types was proposed.By analyzing the tool wear type and wear process,the network structure for identifying the wear state of the tool was designed.The network model was pre-trained by convolution auto-encoder,and the parameters were fine-tuned by backpropagation algorithm combined with adaptive moment estimation algorithm.The experiment proved that the average recognition rate of the model for tool wear type reached 96.25%.A method for automatic detection of tool wear value based on tool wear type identification was proposed.The experimental results showed that the error rate of the test method was less than 10%,and the average error rate was 5.93%,which could meet the practical application requirements.

Key words: tool wear type, convolutional neural networks, model pre-training, image recognition

摘要： 刀具磨损尺寸与刀具磨损类型的有效识别是监测加工状态的重要手段。针对常规刀具磨损类型的识别需要人工参与的不足,提出一种基于卷积神经网络刀具磨损类型的智能识别方法。通过对刀具磨损类型和磨损过程的分析,设计了识别刀具磨损状态的网络结构,采用卷积自动编码器对网络模型进行预训练,并通过BP算法结合Adam算法对模型参数进行微调,建立了有效的刀具磨损类型识别模型。实验证明该模型对刀具磨损类型的平均识别率达到96.25%。最后,提出了基于刀具磨损类型识别的刀具磨损值自动检测方法,实验结果表明该检测方法的误差率在10%以内,平均误差率为5.93%,能够满足实际应用需求。

关键词: 刀具磨损类型, 卷积神经网络, 模型预训练, 图像识别