Enhancing aero-engine blade inspection through fusion of contact probe and 3D scanner measurement data

Turbine blades are complex-shaped parts that play a critical role in the reliability and efficiency of aero engines. Their complex free-form shape must be manufactured with tight dimensional and geometric tolerances. Traditionally inspected with contact probes on a coordinate measuring machine (CMM), blade inspection could be enhanced by using 3D scanners. Non-contact measuring instruments, such as structured-light scanners or laser scanners, have gained popularity in recent years because they can measure objects quickly and with high resolution. Their characteristics make them effective for free-form surface inspection but, their general accuracy is not sufficient for turbine blade inspection. This work proposes a method for enhancing turbine blade inspection by data fusion of 3D scan and CMM measurements. The method uses sparse, high-accuracy CMM points to correct the systematic errors of the less accurate high-resolution 3D scan mesh, to produce a high-accuracy, high-density fusion result that can be effectively used for turbine blade inspection. The data fusion technique proposed is based on the Non-Rigid Iterative Closest Point (NR-ICP) algorithm. Widely used in computer graphics and animation, this algorithm is adapted in this study for data fusion applied to the field of metrology. The NR-ICP approach to data fusion has important benefits compared to traditional parameter estimation and residual approximation data fusion methods. The proposed method directly corrects the 3D scan mesh error by displacing its vertices without modifying the original connectivity and topology. The displacement of the vertices is regulated by a rigidity coefficient, that can correct initial alignment errors or local systematic errors. The fusion result is a higher-accuracy, high-resolution mesh representation of the blade that can be used in any CAD software without post-processing the data. The method is validated in a real-world experimental study on a metallic turbine blade, and the results demonstrate that the systematic errors of 3D scan data can be reduced by 64% when fused with CMM data. The method is also time-efficient: the proposed data fusion process enhances the low-resolution CMM data with high-resolution 3D scans without adding a significant amount of time for measurement acquisition and processing. The CMM inspection routine took about 3 hours to measure a low-density set of 1,464 points on the blade. The 3D scan high-density acquisition of 795,533 points and the fusion process to obtain the high-accuracy, high-resolution mesh representation of the blade only added 22 minutes to the inspection routine. To sample a high-density point cloud with around the same number of points using the CMM would take several days, which makes the proposed data fusion pipeline substantially faster than the CMM-only measurement.