Research On Position Error Correction And Compensation Method Based On Machine Learning

In the field of industrial automation,dispensers are widely used in the manufacturing processes.Faced with the increasing accuracy and efficiency requirements of the market,computer vision is widely used in dispenser systems.Dispensers with high speed,high precision,and intelligence have become mainstream in the market.In order to achieve high precision,it is necessary to correct the error of the dispenser.The classic error correction method,such as the laser interferometer method,has the problems of long time,difficult operation,and high equipment cost,and the calibration process cannot be fully automated.Although the calibration method which using calibration board was commonly used at present achieves high efficiency and portability,it can only achieve local discrete calibration,and still has the problem of insufficient accuracy.Therefore,it is of great application significance to find a highefficiency,high-precision and highly intelligent automatic error correction method.This paper briefly introduces the error composition of the dispenser and analyzes the geometric error of the main position.An error correction scheme for obtaining error sample data by using calibration boards and industrial cameras and using radial basis neural network to approximate the error space to provide soft error compensation is proposed.In order to improve the positioning accuracy of the mark circle during the acquisition of error samples,an image processing algorithm dedicated to mark circle positioning was developed based on the open source visual library Open CV.In the processing of error samples,the Gaussian function was used as the basis function to build a radial basis neural network.A series of adjustments were made to the network's hyperparameters based on offline samples,and a hyperparameter combination suitable for error correction was obtained.Based on the above content,the error correction function module of the dispenser was developed and integrated into the dispenser software as a subroutine.Through experimental data and analysis,the error correction results are tested,and the accuracy and boundary processing are compared with the traditional bilinear correction,and the error correction accuracy is verified with a laser interferometer.Experimental test results show that the dispenser error correction algorithm proposed in this paper can effectively reduce the main geometric errors of the dispenser.Compared with the traditional bilinear correction algorithm,it has higher accuracy,better smoothness and better boundary processing.To meet the requirements of high precision and high intelligence of the dispenser.