Roller Trajectory Extraction and Error Compensation Based on Projection Method and Gaussian Process Regression

To address the difficulty of directly applying finite element simulation trajectories to actual spinning machines, as well as the discrepancy between simulation data and equipment execution in the spinning process, this paper proposes a trajectory mapping method based on simulation trajectory extraction and data-driven error compensation. First, based on secondary development in ABAQUS, a discrete shell is introduced and combined with coordinate transformation to achieve accurate extraction of the roller center trajectory, and the simulated trajectory is converted into a discrete coordinate sequence. Subsequently, a roller trajectory acquisition and visualization system is developed, and machine motion data are collected and visualized for comparative analysis via the Modbus-RTU protocol. On this basis, to address the systematic deviation between simulated and actual execution trajectories, a trajectory error compensation method based on the projection method and Gaussian Process Regression is proposed. By modeling normal-direction errors and applying normal-direction compensation, smooth and stable optimization of the original simulation trajectory is achieved. Finally, experimental validation is conducted on a single-roller spinning machine, and the variation in trajectory deviation before and after compensation is comparatively analyzed. The results show that the proposed method effectively reduces trajectory execution errors, decreasing the average error from 0.503 mm to 0.229 mm, and significantly improves trajectory matching accuracy. This study provides an effective technical pathway for high-precision transformation of simulation trajectories to actual equipment in spinning processes and offers important support for the transition from experience-driven to model-driven spinning manufacturing.