Integrated correction of inherent errors for five-axis laser micro machining scan head based on machine vision (<italic>invited</italic>)

ObjectiveThe five-axis laser micromachining scan head (5A-LMSH) offers unique advantages for numerous precision manufacturing processes. However, existing methods for correcting the inherent five degree-of-freedom (DoF) pose errors of its terminal beam suffer from low integration, limiting both efficiency and controllability. In conventional correction approaches, the five DoFs are calibrated separately, requiring multiple workpieces to be processed with different correction patterns. Each workpiece must be manually placed, processed, transferred to an independent imaging system, and photographed, resulting in repeated cycles that significantly increase time costs. Moreover, the use of separate workpieces introduces alignment errors between different patterns, compromising the consistency and accuracy of the correction process. Additionally, traditional methods rely heavily on manual measurements using digital microscopes, which further limits precision. For example, the center of a cross-shaped mark is typically located by human operators, introducing subjective errors. Similarly, the line segments used for z DoF calibration are evaluated by visual observation of brightness and width, lacking quantitative criteria and repeatability.To address these limitations, this study proposes an integrated vision-based correction method. By designing a single correction pattern that encodes all five DoF errors and developing an automated visual algorithm for simultaneous error extraction, the proposed approach eliminates the need for multiple workpiece transfers and manual measurements. The method significantly enhances correction efficiency, accuracy, and automation, providing a practical solution for high-precision five-axis laser micromachining.MethodsThe experiments are conducted on a custom-built five-axis laser micromachining scan head (5A-LMSH) with a working range of 35 mm × 35 mm. A machine vision system equipped with a rotary stage is integrated into the platform. The machining plane can be rotated by 90° to align its normal with the optical axis of a fixed off-axis camera, achieving front-view imaging without additional optical elements. The vision system is calibrated using ZHANG’s method, achieving a resolution of 9.93 μm/pixel. An integrated 5×5 array correction pattern covering 28 mm × 28 mm is designed, where each sub-pattern contains functional units dedicated to extracting specific DoF errors (Fig.2). The correction process for all five DoF requires only a single manual placement of the workpiece. After laser processing, the rotary stage brings the workpiece to the vertical position for imaging. A visual algorithm is developed for error measurement. After image preprocessing (Fig.4(a)), the z DoF error is determined by analyzing the symmetry axis of line width variation sampled from parallel lines (Fig.4(b)). The x and y DoF errors are obtained by locating the center of the main cross in each sub-pattern (Fig.4(c)). The α and β DoF errors are calculated geometrically using the displacement between main and secondary crosses at two different plane heights (Fig.4(d)). The discrete error data are then fitted to continuous spatial distribution functions. Finally, a single-point correction function is constructed to enable software compensation of inherent errors across the full working range.Results and DiscussionsThe proposed method successfully measured and corrected the inherent five DoF errors. The visual system achieved a resolution of 9.93 μm/pixel. The normalized root mean square error (NRMSE) values for the fitted error models of x, y, z, α, and β DoF were 0.014041, 0.021361, 0.055071, 0.087736, and 0.098895, respectively, indicating good model accuracy (Fig.6). Post-correction experimental validation demonstrated significant error reduction. The mean absolute errors (MAE) for the five DoF were reduced to 31.4%, 20.1%, 14.4%, 37.4%, and 34.4% of their original pre-correction values (Figs.8 and 9). These results confirm that the integrated method effectively enhances the positioning and orientation accuracy of the terminal beam. The entire correction workflow is automated after the initial setup, eliminating the need for multiple workpiece changes and manual measurements associated with traditional separated DoF correction approaches.ConclusionsThis research presents an integrated software correction method for the inherent five DoF pose errors of a 5A-LMSH. The method integrates a novel hardware setup using a rotary stage with a specially designed calibration pattern and a fully automated visual algorithm. It enables the simultaneous acquisition and compensation of all five DoF errors from a single image, dramatically improving correction efficiency and automation. Experimental results verify the method's effectiveness, showing substantial reductions in all DoF errors. This integrated correction approach provides a valuable reference for enhancing the accuracy of galvanometer-based laser processing systems.