Adaptive radius filtering and PointCleanNet-based denoising method for repairing TBM milling cutter rings

Three-dimensional scanning technology is crucial for cutting tool restoration via precise point cloud data acquisition. However, environmental factors and equipment errors introduce noise, challenging restoration accuracy. The traditional noise reduction methods have the problems of being inconvenient to adjust parameters manually and being unable to effectively remove noise, especially when there are both outlier noise and mixed noise. This study proposes a hybrid approach integrating adaptive radius filtering with PointCleanNet, combining traditional algorithms and deep learning. An improved normal estimation is used for feature extraction. Experiments show it effectively denoises while preserving tool details, providing a superior data model for subsequent tool repair.