Point Cloud Based Model-free Path Planning Method for grinding Large Complex Forging Parts

Abstract

Abstract Large and complex forgings, serving as crucial load-supporting components in the fields of energy, ships, transportation, etc., demand high dimensional accuracy and surface quality during post-processing. The automation of grinding for such large and complex forgings presents a common and pressing challenge within the forging industry. One of the main obstacles lies in the substantial thermal deformation and various random forging defects, making the automatic generation of grinding paths a difficult task. In this paper, we propose an algorithm for identifying random defects in large and complex forgings. By combining the RANdom Sample Consensus (RANSAC) algorithm with the Modified Iterative Closest Point (M-ICP) algorithm, we register the standard component point cloud with the forging point cloud, thereby obtaining the point cloud representing the random defects that require grinding. Subsequently, we classify the random defect point cloud based on defect area size and establish a strategy for generating grinding paths. Utilizing the positional coordinate information within the random defect point cloud, we directly generate robot grinding paths without relying on a CAD model. Finally, we conduct robot grinding experiments on large and complex forgings. The experimental results demonstrate that the model-free generation method for grinding paths accurately identifies the characteristics of random forging defects, efficiently plans robot grinding paths, and significantly improves grinding efficiency and quality. This approach offers an intelligent solution for the post-processing of large and complex forgings.