Tool Condition Monitoring for Deep Hole Boring Based on Deep Transfer Learning
Deep learning-based tool condition monitoring methods can accurately predict the breakage of the cutting tools in deephole boring process.However,due to factors such as assembly errors and differ-ences in processing machining environments,the tool condition monitoring model trained on one particular boring machine tool cannot be directly applied to similar boring machine tools.A tool condition monitoring method based on deep transfer learning is proposed to solve this problem.First,monitoring model is trained based on deep belief network using data obtained from the first machine tool.Multi-kernel maximum mean discrepancy is introduced at the top hidden layer of the monitoring model to measure the difference between the source domain and the target domain.Then unlabeled data obtained from the second machine tool is used to finetune the monitoring model.The example of tool condition monitoring shows that after deep transfer learning,the accuracy of the tool condition monitoring model for the target domain has been im-proved by 28.42％,which proves the effectiveness of the proposed method.
transfer learningdeep learningtool condition monitoring
万方数据
维普
