A monocular-vision-based contouring error detection method for CNC machine tools

Abstract

Contouring error detection for machine tools can be used to effectively evaluate their dynamic performances. In this paper, a cost-effective monocular-vision-based contouring error detection method is proposed to precisely measure the two-dimensional error of an arbitrary trajectory under wide working range and higher feed rate conditions. First, a novel measurement fixture with high accuracy is designed and calibrated, in which 196 coded primitives are utilized to accurately characterize the motion trajectory of a five-axis machine tool. Then, to ensure the measurement accuracy and efficiency for high-speed contouring error, a telecentric imaging system is employed to capture the image sequence of the primitives, with a selected small field of view and low camera resolution. Moreover, to extend the working range of measurement system, a primitives' decoding and position estimation algorithm based on the pre-calibrated geometric constraint is proposed to measure an arbitrary contouring error in a wide range. Finally, the contouring error can be accurately assessed by data transformation and post-calculation. Experiments for the contouring error detection of a butterfly curve interpolation at 3000mm/min are performed in a five-axis machine tool. The results, compared with the cross-grid encoder, shows that the average detecting error is 4.2μm, which verifies the vision measurement accuracy and feasibility.