Abstract
Many production processes involve curved sample surfaces, such as welding or additive manufacturing. These pose new challenges to characterization methods for quality inspection, which are usually optimized for flat extended sample geometries. In this paper, we present a laser ultrasound (LUS) method that can be used to efficiently detect defects (e.g., voids), without extensive scanning effort and without a prior knowledge of the defect location, in finite samples with curved surfaces. The developed method starts with generalized simulations of the LUS wave patterns in samples with varying radii of curvature and width as well as varying excitation size and mechanism (thermoelastic or ablative). Based on the wave pattern analysis, it is possible to predict how every point in the weld can be reached with only few excitation spots. In a second step, we assume a grid of finite size defects at locations at which such voids are most likely formed and perform a thorough simulation analysis that is based on B-Scans to find a few pairs of excitation-detection points most favorable for finding defects anywhere in the weld seam. These results are then compared to the wave pattern analysis, discussing similarities and deviations from the predictions. In a final step, the simulations are compared to experimental results, verifying the almost threefold increase in the detectability of defects by choosing the predicted optimal excitation-detection positions. It is expected that this method will significantly improve the reliability and time efficiency of detecting internal defects in samples with curved surfaces in potential industrial applications.