| CNC machine tools are industrial tools for machinery manufacturing.As a key indicator of its manufacturing level,the periodic inspection of machine tool accuracy in unload conditions is critical for machine tool appreciation.The accuracy of the machine tool is divided into static accuracy and dynamic accuracy.To reflect the actual machining state,only the verification of the static accuracy is not enough,and the dynamic accuracy needs to be measured.For high-quality and high-efficiency machining of complex parts with variable curvature,the lack of machine tool execution capability can easily lead to contouring error of a running trajectory,which becomes the main error source of dynamic accuracy.Thus,it is of great significance to study the corresponding test method to improve the accuracy of CNC machine tools.Existing contouring error detection equipment has limitations in measurement accuracy,speed,range and path form,etc.To explore the application of vision in machine tool contouring error measurement,and expand the performance indicators(e.g.,measurement speed and measurement range)of the vision method.A vision-based method for contouring error detection is proposed in this paper.From three aspects of the accurate representation of the distortion behavior of the depth of field(DOF),the high-quality acquisition of the movement images of the machine tool,and the efficient processing of the marker images,high-accuracy vision-based measurement methods are studied.Besides,the application of vision system is expanded by studying the high temporal-spatial method.Finally,four sets of vision testing systems are built to apply and verify the above critical technical issues.Specific research contents are as follows:(1)The multi-dimensional high dynamic motion trajectory measurement of CNC machine tools puts higher requirements on the DOF of the visual lens.In close-range photography,the distortion of the path image is large,which seriously affects the detection accuracy of the contouring error.To this end,a camera calibration method considering the distortion partition of the DOF is carried out.Firstly,aiming at the local convergence of the distortion,linear conformation-based model for separately solving and correcting distortion is established.Secondly,a corner control based method for point linking and fragments grouping is proposed.And the DOF distortion characteristics of two experimental conditions(i.e.,small object distance with large format,large object distance with small format)are revealed.On this basis,an equal-radius partition method for 2D object plane distortion and 3D DOF distortion is proposed.Simultaneously,to meet the requirements of the non-touchable adjustment of the lens(zoom ring and focus ring),a DOF distortion model which independent of the focusing state is constructed.Thereafter,the distortion partition method and the DOF distortion model are combined,and embedded in the process of camera calibration.Finally,the accurate characterization of the DOF distortion behavior and the high-precision camera calibration are realized.The DOF distortion experimental system is built to carry out both the distortion and camera calibration experiments.The comparison results show that the accuracy of the partition correction and calibration method is more than 2 times higher than that of the unpartitioned method.(2)For the contouring error test environment,the illumination is time-varying and unevenly distributed,making it difficult to obtain high-quality images of machine tool movement.In addition,the collected image sequence contains a large number of features,and the exact matching brings challenges to the correspondence determination of the same-named features in multiple images.Thus,the high-quality image acquisition method for machine tool movement,as well as the fast and accurate processing methods for marker images are studied.In terms of image acquisition,the transfer model of light energy between object and image space is firstly analyzed.Then,four kinds of circular markers with different optical characteristics are selected,and the imaging rules under their respective suitable illuminations are revealed.Additionally,the imaging quality of the markers is compared.On this basis,a machine tool movement enhancement method based on lithography marker and uniform lighting is proposed.To match different applications,three sets of measurement fixtures and five cooperative targets are formed.Among them,the large-scale measurement fixture equipped with strobe lighting suppresses the motion blur of the marker,which meets the measurement requirements of dynamic and wide-range contouring error.Image acquisition experiments show that the markers are imaged with high intensity uniformity(standard deviation is 0.54 pixels),high brightness(232.3),high contrast(81.27%)and high sharp in edge(gray gradient of edge is 55.5),which verifies the high quality imaging of the enhanced markers,and provides a prerequisite for high precision positioning.In the subsequent image processing,firstly,an image processing method based on shift cycle is proposed.Then,for its efficiency defects in encoding and decoding,both the method based on maximum-zero-start and the method based on finding the flag bit are proposed to accurately decode the efficient encoding markers.1200 images are collected from the left and right cameras,respectively.Each image has an average of 36 coded markers,and a total of 86400 coded markers are decoded and positioned,those three methods get 100%correctness.In addition,the decoding efficiency of the latter two methods is obviously better than that of the shift cycle method,which verifies the efficiency and accuracy of the image processing methods.(3)Aiming at the problem that the measurement speed and measurement range of the vision method is difficult to be improved synchronously,the constraint relationship between imaging parameters under constant bandwidth is analyzed,and a vision high-temporal measurement method based on error distribution is proposed.In terms of the measurement speed,a test speed improvement method by synchronously reducing the camera resolution and FOV is presented.While,in terms of wide-range measurement,for the invisibility of the reference primitive in images caused by small FOV,a marker is selected as the reference primitive to represent the whole motion trajectory,and a wide-range contour measurement method based on prior geometric constraints is studied.Moreover,the idea is applied to the theory of monocular telecentric vision,binocular 3D vision and monocular 3D vision,and a series of wide-range contour calculation models are constructed,which indirectly improve the spatial resolution of vision methods.Finally,a reference conversion method based on the orthogonal auxiliary motion of the machine tool is established,and the test data in machine tool coordinate system are obtained.After comparing with the nominal trajectory,the contouring error is detected.With the proposed method,the frame rate can be increased from 33 fps to 590 fps,and the measurement range is increased from 85 mm × 85 mm to 231 mm × 231 mm.Based on the above critical technical issues,four sets of vision-based systems are built for testing contouring error of CNC machine tools.And contouring error measurement experiments of small-scale equiangular spiral,wide-range butterfly curve,wide-range equiangular spiral and wide range space trajectory are carried out correspondingly.The experimental results show that the vision detection error is less than 1/3 of the contouring error to be detected,which indicates the validity and accuracy of the vision method.The research work provides an important reference for the dynamic accuracy improvement of machine tools. |
PDF Full Text Request