Research On Classification Of Tool Wear Based On Texture Of Workpiece Surface

Tool is one of the core components of machine tool processing,and its working state directly affects the quality of processed products.When the tool is severely worn,it will damage the quality of the workpiece,or even lead to edge collapse,which will cause the machine to stop.Therefore,it is important to monitor the tool status in real time to improve the production efficiency and processing quality.Tool wear detection method based on workpiece surface texture is convenient,flexible and non-contact measurement.With the rapid development of machine vision,image-based detection method shows more advantages in accuracy and speed.Under this background,this paper obtains the workpiece surface image through wear experiment,constructs the tool wear classification model based on feature extraction and deep learning,and improves the classification accuracy of the model through direction correction,feature fusion and model structure improvement.A support vector machine model(GA-SVM)based on genetic algorithm parameter optimization is proposed.Aiming at the problem that some texture features are sensitive to direction,a direction correction method combined with downsampling is proposed.The corrected gray level co-occurrence matrix feature and the number of connected domains feature are extracted,and GA-SVM is established to compare and analyze various feature combinations.Finally,the classification accuracy of the test set of the proposed method is over 87%,which is higher than that of the original method.The results show that the proposed SVM classification model combined with direction correction and feature fusion can significantly improve the classification accuracy of rotating data sets,and it can still achieve good classification effect in the application scenarios of small sample size(70samples)and high noise.A convolutional neural network model combined with attention mechanism(SEDCNN)is proposed.The optimal super parameter combination of the model is obtained through comparative experiments.The original data set is directly input into the model after data enhancement and expansion for learning and classification.The final test set classification accuracy reaches 98.7%,which is better than the traditional convolutional neural network model and traditional SVM model.It is verified that deep learning method is effective in tool wear classification tasks under big-sample and low-noise scenarios.The two models proposed in this paper have different advantages in different production scenarios.Generally speaking,they have achieved good classification results in tool wear classification task based on workpiece surface image,which provides a new idea for rotating texture processing and has practical value in industrial field.