In-situ Measurement Method And System Construction Of Surface Roughness In Clean Cutting Based On Machine Vision

Surface roughness is related to the friction,assembly and fatigue resistance of parts.Comprehensive,efficient and accurate measurement of surface roughness is important in the process of experiment and actual production.Most of the existing roughness measurement methods are offline measurement,and the evaluation standard is mostly used for two-dimensional plane,while the research on the surface roughness in-situ measurement of three-dimensional space is less.Machine vision technology is a comprehensive technology based on optical technology,computer technology and information technology.The combination of machine vision technology and surface detection technology provides a new research method for the detection and recognition of cutting workpiece surface quality.Therefore,aiming at the clean cutting mode of high speed dry cutting,this paper studied the measurement method of surface roughness based on machine vision technology,and develop the corresponding surface roughness in-situ measurement system to realize the in-situ and nondestructive measurement of surface roughness in high speed dry cutting.In this paper,the typical hard to machine material superalloy GH4169 was used as the machining material.Different cutting parameters were selected to carry out the experiment of clean cutting-high speed dry milling,and the surface morphology of the milled material was observed and analyzed.The laser scanning microscope was used to collect the machined surface image,the surface roughness Sa was measured and the surface images were selected to establish the image sample set to provide supports for the subsequent study of surface roughness measurement method.The preprocessing methods of machined surface images including image gray conversion,image filtering and image enhancement were analyzed.Extracting gray components from HSI color space,median filtering and the image enhancement method of square transformation were selected as preprocessing methods based on LabVIEW machine and vision module.The gray value in the surface image was converted to the height value to construct the three-dimensional image of the machined surface,which further proved that the gray value of surface image can be used to measure the surface roughness.Based on the gray level co-occurrence matrix of milling surface image,the feature values of texture distribution were calculated,and the relationships between roughness and six feature values of dissimilarity,contrast,homogeneity,correlation,entropy and energy were analyzed.It was found that the relationships between roughness and dissimilarity,contrast,entropy and energy are obvious.The dissimilarity,contrast,entropy and energy were determined as the features of feature vector.In order to solve the problem of surface debris interference,the binary processing was carried out on the machined surface image to extract the surface debris information,and the area of debris in the image was calculated according to the binary particle area.This paper analyzed the numerical relationship between the area of debris and the change rate of feature value,obtained the fitting formula between the numerical relationship,put forward the optimization method of the feature value of the surface image with debris,solved the problem of the change of the feature caused by the interference of debris,and selected the surface image with debris for feature value optimization experiment and found the proposed feature value optimization method has good effect.Two parallel surface roughness measurement methods based on image texture feature were proposed.One method of surface roughness measurement method was based on LabVIEW classifier.In LabVIEW development platform,the nearest neighbor method was used to classify 150 machined surface images,and the samples were trained according to the image feature vector,the feature classifier including the feature vector and roughness value of the sample image was generated,and the roughness value of the tested image was classified by the feature classifier to realize the measurement of roughness.Another method was based on BP neural network,and the four features of image dissimilarity,contrast,entropy and energy were taken as the input layer,and the surface roughness was taken as the output layer in MATLAB to predict surface roughness.The same 150 machined surface images were selected for training,and the roughness was predicted according to the characteristic values of the samples to be tested by mixed program of LabVIEW and MATLAB.In addition,30 machined surface images were selected to test the effect of the two methods.It was found that the error of the surface roughness measurement method based on LabVIEW classifier was 9.52%,and the error of the surface roughness measurement method based on BP neural network was 9.81%.This paper designed the in-situ measurement system of surface roughness which consist of hardware part and software part.The image acquisition device and the moving parts device constituted the hardware part.The image acquisition device was composed of industrial camera,micro lens and coaxial light source,and the moving parts device was compose of screw,sliding table,stepper motor and controller to realize the rapid acquisition of workpiece surface image.Software part of surface roughness measurement system was mainly developed based on the LabVIEW,and combining with MATLAB.The software was divided into image acquisition module,image preprocessing and analysis module,roughness measurement module and operation panel.The roughness measurement system was tested on DMG DMU-70V five axis CNC machining center,which meet the needs of in-situ roughness measurement and has high detection speed.