Optimal measurement points decision and welding process diagnosis considering welding distortion

Abstract

Over several decades, considerable efforts have been made by shipbuilding industry to improve the quality of their products. Ship and offshore structures are constructed out of millions of intermediate parts. Typically, welding joins the intermediate parts, and the welding process inevitably causes distortions. The distortion of each welding process is small. However, stacked up distortions through the welding processes causes a large dimensional difference between ideal products and actual products. The misalignment of blocks necessitates additional steps to calibrate and correct the dimensions. Therefore, the shipbuilding industry has focused on dimensional accuracy control to reduce necessity for additional corrective steps. Geometrical variation measurement is an essential process in the accuracy control. Multiple location of points should be measured to compare the dimensions of real product with a designed product. As the number of measurement points increases, not only the accuracy of measured geometrical variation of product would become higher, but also the amount of measured data as well as the measuring time, would greatly increase. Thus, the optimization is required to determine the optimal number of measurement points as well as their locations. This study proposed optimal measurement location that considered welding distortion. Variation simulation was conducted to distinguish the effect of each welding distortion propagation. The critical welding process could be diagnosed by using the relation of propagated welding distortions. The Fisher information matrix was generated using data from the propagated welding distortion, and the optimal measurement points that maximized its determinant were identified. The candidate measurement points are determined by the Nyquist sampling theorem. Deformed shapes were reconstructed from the data of optimal measurement points.