Tool wear monitor method based on composite signal of multiple kinds of sensors

Abstract

The invention relates to a tool wear monitor method based on a composite signal of multiple kinds of sensors. An acoustic emission sensor and a power sensor are adopted for collecting signal information relevant to machine tool wear, and the defects of a single signal can be avoided through two signal collection methods. Two kinds of information are scientifically coupled through a cloud model algorithm, and the characteristic factor for reflecting tool wear in the signal can be extracted; a sparse bayes method is used for modeling, and then the tool wear is predicted; an SBL based recognitionmethod is adopted for carrying out modeling on data; a bayes matching pursuit algorithm is adopted for optimizing width parameters of an SBL model kernel function; and the tool wear is accurately predicted, and the efficiency and accuracy of tool wear monitor are improved.

Description

technical field [0001] The invention relates to a tool wear monitoring method based on multiple types of sensor composite signals, belonging to the field of tool wear detection. Background technique [0002] As one of the cores of smart factories, smart devices have the ability of self-identification, self-learning and self-maintenance of operating status as their important features. According to statistics, tool change and tool setting during processing account for about 20% of the equipment's running time. In addition, tool wear and damage have a significant impact on processing quality, processing efficiency, machine tool life and even the personal safety of operators. Therefore, accurate and efficient self-identification and automatic warning of tool running status are of great significance to improve the intelligence level of machine tools, which can effectively save costs and improve efficiency. [0003] Due to the complexity of the tool wear process in high-speed mi...

AI Technical Summary

Problems solved by technology

The prediction algorithm of artificial neural network, the model is too complex, needs a large number of experimental samples, and the calculation convergence is difficult. Support vector machine can realize the prediction of tool wear in small samples, but it is prone to over-learning phenomenon, the sparsity of the model is limited, and the probability information of the prediction result cannot be provided

Images