In-situ Detection And Prediction Method Of Micro-grinding Tools Wear Based On Machine Vision

At present,in the micro-grinding process of hard and brittle materials,electroplated diamond micro-grinding tools still have problems such as rapid wear,unclear wear characteristics,and low detection efficiency,which affect the processing quality and geometric accuracy of small parts and microstructures.This paper aims to improve the traditional micro-grinding tools wear detection method,build a microgrinding tools wear in-situ detection system based on machine vision technology,a method for in-situ acquisition of physical quantities characterizing micro-grinding tools wear,a micro-grinding tools wear state identification method and a microgrinding tools wear prediction model driven by physical feature data,i n order to realize the in-situ detection of the wear stage of micro-grinding tools and the prediction of the wear life of micro-grinding tools.The main work of the paper is as follows:(1)The wear form and wear characteristics of micro-grinding tools are analyzed,an in-situ visual inspection scheme suitable for micro-grinding tools is designed,and the feasibility analysis of the detection scheme is car ried out.The hardware design and software function design of micro-grinding tools visual inspection scheme are respectively carried out,and a visual inspection platform for micro-grinding tools monocular fixed magnification is built.Through the calibrat ion and performance analysis of the software system in micro-grinding tools visual inspection system,the in-situ detection of micro-grinding tools wear is realized.(2)According to the inspection requirements of micro-grinding tools,formulate the image processing inspection process.The edge extraction of the micro-grinding tool is realized by grayscale transformation,adaptive median filter,threshold segmentation based on region growing method and sub-pixel edge detection algorithm.The connected domain of the micro-grinding tool contour is obtained by the connected domain mark,and the diameter of the micro-grinding tool grinding head is measured by using the line scan code.Through the error analysis of the measurement results of different detection methods,the reliability of the measurement results after image processing is verified.(3)The wear measurement method of the micro-grinding tool was analyzed,and the grinding head diameter measurement method and grinding head area measurement method of the micro-grinding tool were compared according to the wear trend of microgrinding tools.A method for characterizing the wear characteristics of micro-grinding tools with the loss of the cross-sectional area of the grinding head as the index.This method is used to analyze the influence of different process parameters on the wear of the micro-grinding tool,and the wear stages are divided according to the wear trend of the micro-grinding tool.The loss range of the grinding head cross-sectional area in different wear stages is analyzed,and the K-means clustering method is used to cluster the grinding head cross-sectional area loss to determine the wear state of microgrinding tools under different process conditions.(4)A data-driven prediction method is used to establish a micro-grinding tool wear prediction model based on genetic algorithm optimized BP(GABP)neural network.The spindle speed,feed rate,micro-groove depth,grinding length and initial grinding head area are used as inputs.The amount o f micro-grinding tool grinding head cross-sectional area loss is predicted.This model is compared with the prediction effect of the traditional micro-grinding tool wear prediction model based on Gaussian process regression.The results show that the avera ge error accuracy of the micro-grinding tool wear prediction model based on GABP neural network under the same working conditions and variable working conditions reaches 5%.It can not only predict the loss of grinding head section area,but also predict t he wear life of micro-abrasives under different process parameters and different grinding lengths.