Design of an intelligent putty spraying experimental teaching platform for industry-education integration

The existing putty spraying process continues to rely heavily on manual operation, resulting in challenges such as surface flatness inconsistencies, low construction efficiency, and considerable material waste. These issues have become impediments to industrial upgrading. The transition to automated solutions necessitates fundamental breakthroughs in key enabling technologies, including high-precision three-dimensional (3D) perception, intelligent decision-making, and precise motion control. To address these technical challenges and concurrently support the talent development objectives of emerging engineering education in intelligent manufacturing, this research presents an industrial robot–based intelligent putty spraying system that leverages core algorithm innovations to advance automated spraying technology.

The developed system is built on a robust hardware platform integrating a multisensor array for 3D laser reconstruction and a seven-axis industrial robot for execution. The core methodology is structured in a multilayered architectural approach. In the perception layer, a high-speed point cloud stitching process is employed to create a holistic digital twin of the workpiece. An improved normal distribution transform–iterative closest point (NDT-ICP) hybrid point cloud registration algorithm was proposed, specifically engineered to enhance alignment accuracy for large-scale scans, successfully reducing the average registration error to 0.08 mm. In the analysis layer, a novel point cloud template matching method was developed, augmented by curvature feature enhancement. This algorithm demonstrates strong capability in identifying subtle geometric variations, achieving a recognition accuracy exceeding 96% for submillimeter-level defects, including dents and bulges. This process is further refined through improved guided point cloud filtering and adaptive contour matching processing to precisely isolate defective regions. In the decision and execution layer, a dynamic path planning algorithm was designed. This algorithm strategically fuses a bio-inspired neural network (BINN) for real-time obstacle and coverage mapping with non-uniform rational B-spline (NURBS) optimization for trajectory smoothing. This synergy enables the real-time generation of efficient, collision-free, and fully covered spraying paths tailored for complex free-form surfaces, effectively minimizing redundant motions and path repetition.

Experimental validation confirmed the superior performance of the proposed system and its constituent algorithms. The enhanced NDT-ICP registration algorithm exhibited higher precision and robustness compared with conventional curvature-based region-growing segmentation methods. All defect identification errors were confined below the 0.05 mm threshold, with the most notable improvement observed in bulge defect recognition, where accuracy increased by up to 70%. The BINN + NURBS dynamic path planning algorithm consistently generated the most efficient paths, achieving the lowest recorded metrics in path repetition rate and overall redundancy among all benchmarked algorithms. This optimal pathing directly translates to reduced cycle times and mitigates material waste caused by overspraying. In a practical application scenario involving the coating of an aluminum alloy sidewall panel, the system demonstrated a nine-fold increase in spraying efficiency compared with skilled manual labor, alongside a 17% reduction in material consumption. The final coating quality consistently met the stringent requirements of the Q/CR546.1 industry standard, validating the system’s practical reliability and effectiveness.

The intelligent putty spraying system developed in this project represents a significant advancement in automating a traditionally manual-dependent process, providing a reliable technical solution for industrial applications requiring sophisticated path planning and intelligent decision-making capabilities. As a representative case of intelligent manufacturing, the project has been incorporated into the experimental teaching curriculum, effectively strengthening students’ innovative and practical abilities in addressing complex engineering challenges.