Study On Point Cloud Addition Iterative Closest Point Registration And Processing For3d Measurement Of Aero Engine Blades

With the fast development of China’s civil aviation industry, aviation engine bladesrepair technology is playing an increasingly important role in the aviation industry. As oneof the key parts of aero-engines, aero-engine blades, on the one hand, are easily damageddue to the harsh working environment, and on the other hand, require high precision oftheir shape. Therefore, three-dimensional measurement of the blades and welding repairtechnology are of great significance in the aviation field. In the3D measurement of theblades, the point clouds registration and processing algorithm are still immature, soresearch on highly accurate, fast3D point clouds data processing technology can helpincrease the competitive edge of the aviation industry.After comprehensively reviewing related works of3D point clouds processingalgorithm, the new methods of point clouds registration and3D processing algorithm areproposed. As aero-engine blades require accurate curved shape and high precision, a newpoint clouds registration algorithm based on ICP algorithm is proposed. Also proposed is animproved direct data reduction method, in which two existing methods were combined–enveloping box and curvature variation.A algorithm based on k-d tree realizes the denoise of unordered point clouds after thepoint clouds are got. But the number of the points are still very large, so enveloping boxand curvature variation algorithm are adapted to reduce the quantity of the point clouds. Inthe point clouds registration algorithms, the rough registration algorithm makes differentperspective point clouds to have similar position and can meet the requirement of ICPalgorithm about point clouds position. ICP algorithms in fact try to reach an optimalalignment for all available points, regardless whether some should not to be aligned at all.In the search of corresponding points, if there is no point to be found within the threshold,the addition to the point cloud will be applied to prevent errors occur. The new approach’sstandard deviation can reach to0.004mm, and it costs500s less time than the traditionalICP algorithm. Triangle filling method is adapted to deal with holes in point clouds, andthen process point clouds data into polygons, and use polygons to reconstruct the blade.For one type of aero-engine blade, its3-D model is reconstructed after processing andregistration. Experiment results show that the average distance and standard deviation meetthe requirements, and the new algorithms use less time, and their accuracy satisfy the actual requirements of industry.