Monitoring Method of Tool Wear Condition in Turning Based on Metric Learning

Abstract

The tool wear state monitoring method for turning machining based on metric learning, firstly, the collected tool wear images are preprocessed, and then randomly divided into a training set and a test set in proportion, and the training set and test set are divided into a support set and a test sample set ; Establish a non-parametric neural network model structure, including feature extraction network, distance measurement function, and attention mechanism; initialize feature extraction network model weights, deviations, and learning rate parameters, and construct activation functions; use cosine distance functions as distance measurement functions; Attention mechanism Use the softmax layer structure; use forward propagation to calculate the current cross-entropy loss, and then calculate its gradient through backward propagation; optimize the weight and bias by running the gradient descent method, and finally use the learning parameters of the network model to predict and classify the test set Identify and output prediction accuracy. The invention has the characteristics of simple algorithm realization, error correction, high classification accuracy of tool wear state, short running time and strong real-time performance.

Description

technical field [0001] The invention relates to the field of machinery, in particular to a method for monitoring the wear state of a turning tool based on metric learning. Background technique [0002] According to statistics, the failures caused by tool problems accounted for 22.4% of the total failures. In actual production, workers judge the wear state of tools based on their own experience, such as the number and processing time of workpieces, and the surface quality of workpieces. and machining noise and chip formation. There will be some problems: the tool is changed before its service life, which will increase the cost of the prop; the tool is changed only after the tool is seriously worn, which will affect the machining accuracy of the workpiece, and even damage the workpiece and the machine tool. It is necessary to apply intelligent tool wear status monitoring technology. [0003] At present, the tool wear state monitoring technology is mainly divided into: tool w...

AI Technical Summary

Problems solved by technology

[0005] In response to the above problems, the object of the present invention is to provide a method for extracting tool wear features in images and better measuring the similarity between feature vectors by adding a feature extraction network and an attention mechanism on the basis of a mathematical model of metric learning Combining the forward propagation and backward propagation of the neural network, the improved model has the way of neural network thinking to solve the problems of high algorithm complexity, real-time identification and low accuracy of turning processing based on metric learning Tool wear status monitoring method

Images