An optimal reference iteration-based surface reconstruction framework for robotic grinding of additively repaired blade with local deformation

•A blade surface reconstruction framework is proposed based on the optimal reference iteration.•Fitted cross-section curves are aligned by sorting features-based parameter alignment method.•Aligned multi-section curves are discretized with curvature features-based discrete step model.•Blade reconstruction accuracy is 0.018 mm based on v-directional optimal reference principle.•The framework provides highly reliable reference surface for blade repair and remanufacturing. The additively repaired blade requires intelligent grinding process to restore the blade profile, and the essential prerequisite is to reconstruct a reliable reference surface at the repaired area with local deformation. In this paper, a novel surface reconstruction framework based on the optimal reference iteration in the v parameter direction is developed to overcome this challenging problem through three steps. In the framework, a parameter alignment algorithm based on the sorting features is proposed to align the fitted blade cross-section curves at first by considering the local deformation and curvature variation, while avoiding the alignment falling into local optimum. Then a curve discretization method based on the curvature features is presented to discretize the multi-section curves for preserving the high curvature variation features to the most extent. Based on these two steps, a curve fitting strategy based on the v-directional optimal reference iteration is suggested for surface reconstruction by virtue of the optimal reference principle and VMM (Variance-Minimization Matching) algorithm. Both the simulation and experimental results demonstrate the effectiveness of the proposed framework from the perspectives of the blade position errors and the profile accuracy after robotic grinding. The average point cloud error between the reconstructed model and the standard model is 0.018 mm, which is decreased by 51.7 % compared with the state-of-the-art method.