Angular Error Compensation Technology For Comprehensive Performance Test Bench For Precision Reducer

As the core component of industrial robot,the performance of precision reducer directly affects the performance of robot.The key to improve the product quality of reducer lies in the high-precision testing of its comprehensive performance.The angular displacement of output and input shaft is the key basic quantity in the performance test of reducer.The angle measurement accuracy directly affects the test accuracy of reducer performance.According to the actual engineering requirements of angle measurement system calibration in precision reducer comprehensive performance tester,this thesis builds an automated calibration system that meets the structural characteristics and use needs of the instrument,and proposes a multi-data correlation angle measurement error compensation theoretical model.The application results show that the calibration process can be realized with high efficiency while ensuring the improvement of angle measurement accuracy,so that the operation process becomes simple and standardized.The angle measurement system is calibrated by the combination of autocollimator and polyhedral prism.The installation is designed and trial-produced to coaxially connect the polyhedral prism and the angle measuring system to realize the synchronous and gapless rotation of the polyhedral prism and the circular grating.Among them,the input side adopts the expansion structure,and the output side adopts the interference fit scheme.The process of motor control,autocollimator and circular grating signal acquisition,calibration data processing and other processes are modularized,and the automatic calibration program is written to simplify the calibration process and realize "one-button" measurement.The error compensation method is studied,the advantages and disadvantages of numerical compensation method and BP(Back Propagation)neural network compensation method are theoretically analyzed,and a multi data correlation error compensation method is proposed.By changing the relative position of polyhedral prism and circular grating in the circumferential direction,several groups of calibration data of the angle measurement error are obtained.The error harmonic method is used to preprocess measurement data,and different groups of data are associated and substituted into BP neural network for training to obtain the angle measurement error model of circular grating.The error compensation model is established by using error harmonic method,BP neural network method and harmonic neural network combination method,and the angle measurement system is compensated.The effectiveness of the compensation model is verified by experiments and the compensation effects of different models are compared.The experimental results show that the peak value of the residual error of the angle measurement system after compensation is within 2.20 arc seconds,and the standard deviation is 0.54 arc seconds,which is reduced to 37.5% and 36.0% before compensation.Numerical results prove that the proposed harmonic neural network combination method can effectively improve the accuracy of the angle measurement system,and make the angle measurement accuracy of the precision reducer detector better than ± 2 arc seconds.