Surface Reconstruction And Digital Estimation Of The Springback Of Auto-body Panels

Springback is the inevitable physical phenomena of Auto-Body panel forming process, establish a simple and efficient method for estimating springback has great significance for guiding to trim mold, improving product quality and shorting the production cycle. Compared to springback, some defects like small protrusions and depressions on the panel's surface are different scales of deformation, and the technique of post-processing also different. The effect of local defects must be eliminated when estimating springback, to avoid make more effect to trim mold due to the error of springback estimation.Surface reconstruction is the premise and foundation of springback digitization estimation which bases on reverse engineering, where the feature preserving is the key point during reconstruction, and estimating the normal vectors of point cloud models is the key problem. Using pre-processing to eliminate outliers and significant noise points of point cloud models, on this basis, it adopt weighted principal component analysis(WPCA) to estimate normal vectors. According to the principle of making propagation process as much as possible close to the curve along the tangent direction, concluding the priority measure rules of normal propagation, and make the normal vectors consistent. Using anisotropic projection and feature region growth sampling technologies to keep both sides of the feature region will not influence each other when estimating normal vectors. The test results show that surface reconstruction results can preserve the geometrical features of sharp edges perfectly.For the problem of combining the method of manual specify and automatic align the benchmark of springback, it should specify the benchmark of springback firstly,and give it a large enough weight, then add to ICP solving system by the way of soft constraint, constitute the weighted iterative closest point algorithm, and finally align the model under fusion benchmark constraint conditions. On the one hand it canconsider the effect of springback benchmark, on the other hand it can make the error of benchmark position uniform distribution, reducing the error of benchmark position influence on springback estimation as much as possible. Compared to classical ICP algorithm, it has the same solving framework and lower complexity.