Neural Process Enhanced Machining Error Evaluation for Coordinate Measuring Machines

Abstract

Coordinate Measuring Machines that provide ultraprecise results are widely used in precision manufacturing. However, the traditional point-by-point measuring approaches are time-consuming, limiting their applications in modern manufacturing, especially for large-scale surface measurements and online measurements. In this paper, we propose an efficient machining error evaluation approach for complex surfaces based on the neural process. Without sacrificing accuracy, the proposed method improves the efficiency of measurement by sparse sampling and reconstruction. Using the neural process, the reconstruction algorithm models the machining error accurately and robustly. Besides, by estimating the reconstruction uncertainty, we further improve the sampling efficiency by introducing adaptive sampling. Intensive experiments show that the proposed method can achieve accurate reconstruction with high efficiency compared to existing approaches.