A robust and accurate automated registration method for turbine blade precision metrology

• Published in: International journal of advanced manufacturing technology 2018-08, Vol.97 (9-12), p.3711-3721
• Main Authors: He, Wantao, Li, Zhongwei, Guo, Yanyan, Cheng, Xu, Zhong, Kai, Shi, Yusheng
• Format: Article
• Language: English
• Subjects: Automation; B spline functions; CAD; CAE) and Design; Computer aided design; Computer simulation; Computer-Aided Engineering (CAD; Design optimization; Engineering; Holography; Industrial and Production Engineering; Inspection; Iterative algorithms; Mechanical Engineering; Media Management; Metrology; Noise reduction; Original Article; Outliers (statistics); Principal components analysis; Registration; Robustness; Signal to noise ratio; Three dimensional models; Tolerances; Turbine blades; Turbomachinery
• ISSN: 0268-3768; 1433-3015
• DOI: 10.1007/s00170-018-2173-9

Turbine blades with complex features are a critical part of turbomachinery, and the manufacture of geometric dimensions and tolerances is strictly controlled to ensure the efficiency and safety of the engines. Precision inspection under current industry practices is challenging and inefficient; however, inspection with an optical device is a promising technique. One key task involved is registration that aligns the measurements to the part model to achieve a fast and automatic inspection process. We present robust and accurate coarse and fine automated registration methods for turbine blade precision metrology. An iterative scan strategy is used to obtain sufficient point clouds to construct curves and surfaces using a B-spline method for registration. Then, principal axes of the reconstruction surface of the blade are calculated, and a principal component analysis (PCA)-based coarse registration method is used. The coarse alignment of the measurement data and the computer-aided design (CAD) model are optimized by fine registration using a common iterative closest point (ICP) algorithm. In this step, the signal-to-noise ratio is incorporated in the transformation of correspondence sets to reduce or remove the noise of outliers. These techniques have been implemented in a four-axis blade inspection system with a point-based conoscopic holography sensor. The results of measurement simulation experiments and inspection case studies indicate that the presented registration method is robust and accurate.