| High-performance components are widely used in aviation, aerospace, defense, electronics and other fields. NC machine tools are the main processing method for these parts with high precision and quality. However, there are many errors in the process of machining, which can decrease the accuracy of the parts. When rotary axes are used in combination with linear axes, the position and the orientation of the rotary axes have to be known accurately, because these parameters strongly influence the machined work piece. To improve the processability, error detection and compensation are important. The most common methods of error detection include laser interferometer, double ball bar, R-test, et al. However, complex installation and debugging, long measurement time, difficult error identification, et al are the disadvantages of above methods. Besides, every equipment has respective applicable scope, the same equipment is difficult to achieve various error of rapidly. Vision measurement with three dimensional measurement, convenience, and low cost provide a flexible means for error detection of machine tools. Therefore, this paper proposes a method for error detection based on binocular vision. The concrete contents include:(1) On the basis of the topology structure of the five-axis CNC machine, installation error of the rotary axis and the principle of vision measurement, the error detection and identification model is built to lay a foundation for error detection.(2) To ensure the precision of error detection, the image feature extraction methods with high efficiency and are studied. Focused on the highly corrupted images by noise, a noise suppression method is proposed based on global threshold and local threshold, which can improve the robustness of image segmentation and image preprocessing speed Effectively. Besides, a suppression method for strong light based on area ratio is proposed to avoid the low extraction accuracy caused by strong light. Last, a compensation method of the circle center based on harmonic conjugate constraint is proposed to solve the problem of the marker's deformation caused by projective transformation. That method can improve the accuracy of feature extraction.(3) Reference transformation model for image coordinate system and machine tool coordinate system is set up, which includes the transformation of image pixel coordinate system and visual coordinate system, the transformation of visual coordinate system and machine tool coordinate system. The internal and external parameters can be obtained through Zhang's calibration method, by which the transform of 2D pixel coordinate system and 3D space coordinate system can be achieved. The rotation and translation transformation of visual coordinate system and machine tool coordinate system can be set up by auxiliary process of machining and movement supporting method.(4) High-precision self-luminous target is designed and 5-free binocular vision measurement system is set up. Detection and identification for assembly error of rotary shaft of machine tool can be defined based on vision measurement. The measurement precision is evaluated using ball bar. The results show that the deviations of the actual axis line and the reference axis in X and Y direction are 0.0038mm and 0.0039mm, respectively. The angular errors are 0.0023°and 0.0021°(5) Angle positioning error of rotary shaft of machine tool is measured using vision measurement system, and precision test is conducted using laser interferometer. The difference of the two methods is 3.18". The interpolation X/Y circular trajectory error is measured, and precision test is conducted using plane grating. The difference of the two methods is 0.0041mm.The research result of this paper provides effective method to measure error of machine tool. It is both valuable in applications of visual technology expanding and machine tool measurement, which lays the foundation to explore new methods of error detection. |
PDF Full Text Request