High-fidelity morphology modelling with sparse point cloud of complex surface based on VMD feature extraction and predictive reconstruction

High-fidelity morphology modelling of complex surfaces is the basis for accurate characterization of surface quality and quasi-realistic performance analysis in areas such as digital process design and digital twin. High-fidelity morphology modelling is usually based on high-resolution data, which is inefficient to acquire and reconstruct, while conventional methods using sparse point cloud have poor modelling accuracy. To address this problem, a high-fidelity morphology modelling method with sparse point cloud of complex surface based on variational mode decomposition (VMD) feature extraction and predictive reconstruction is proposed. The algorithm is based onVMDfor mode decomposition of surface morphology feature data of surface parts, firstly, a feedback mode separation method is used to determine the number of decomposed modes; secondly, a morphology feature extraction based on high-resolution point cloud is proposed, uses morphology feature functions to fit and extract surface features, and finally, a high-fidelity morphology model reconstruction with measured sparse point cloud is used, and an innovative high-fidelity prediction with shape and morphology feature reconstruction method is proposed to reconstruct the surface morphology of sparsely sampled surface parts into high-resolution surface morphology of parts. The experimental results of the turbine blade indicate that the sampled data volume of the sparse point cloud is only 6.25% of the high-resolution point cloud, and the average morphology error of the reconstruction algorithm is about 2 μm, with the average error rate and the standard deviation error rate of 0.08% and 0.21%, respectively. The proposed morphology modelling method has high accuracy and low complexity, which can achieve efficient and high-fidelity reconstruction with sparse point cloud of complex surfaces.