Identification and Adjustment of Guide Rail Geometric Errors Based on BP Neural Network

• Published in: Measurement science review Vol. 17; no. 3; pp. 135 - 144
• Main Authors: He, Gaiyun, Huang, Can, Guo, Longzhen, Sun, Guangming, Zhang, Dawei
• Format: Journal Article
• Language: English
• Published: Bratislava De Gruyter Open 01.06.2017; De Gruyter Poland; Sciendo
• Subjects: BP neural network; Displacement; Error analysis; Feasibility; Finite element method; finite element model; Guide rail geometric error; Guide rails; Mapping; Mathematical analysis; Neural networks; stress; Stresses; Test procedures; the test plate
• ISSN: 1335-8871; 1335-8871
• DOI: 10.1515/msr-2017-0017

The relative positions between the four slide blocks vary with the movement of the table due to the geometric errors of the guide rail. Consequently, the additional load on the slide blocks is increased. A new method of error measurement and identification by using a self-designed stress test plate was presented. BP neural network model was used to establish the mapping between the stress of key measurement points on the test plate and the displacements of slide blocks. By measuring the stress, the relative displacements of slide blocks were obtained, from which the geometric errors of the guide rails were converted. Firstly, the finite element model was built to find the key measurement points of the test plate. Then the BP neural network was trained by using the samples extracted from the finite element model. The stress at the key measurement points were taken as the input and the relative displacements of the slide blocks were taken as the output. Finally, the geometric errors of the two guide rails were obtained according to the measured stress. The results show that the maximum difference between the measured geometric errors and the output of BP neural network was 5 μm. Therefore, the correctness and feasibility of the method were verified.