Development of a Two-Dimensional Bead Shape Prediction Algorithm for Fillet FCAW Considering Welding Speed and Torch Offset

Automated welding processes have become increasingly critical in the shipbuilding, automotive, and aerospace industries. Precise weld seam tracking is essential for reliable automated welding, as torch misalignment causes weld defects and significant variations in bead geometry. In this study, fillet welding experiments using flux-cored arc welding (FCAW) were performed under 15 conditions comprising five torch offset levels (−2, −1, 0, +1, +2 mm; positive values denote displacement toward the vertical plate) and three welding speeds (6.0, 7.5, and 9.0 mm/s). Twelve characteristic points were extracted from three bead cross-section zones—the vertical penetrated zone, horizontal penetrated zone, and bead surface zone—and predicted using a nonlinear exponential regression model. All regression coefficients were statistically significant at the 95% confidence level. Piecewise cubic Hermite interpolation (PCHIP) was subsequently applied to reconstruct smooth two-dimensional bead profiles from the predicted feature points. Quantitative validation of the primary leg-length parameters (LV and LH) yielded low prediction error (RMSE below 0.7 mm and MAE below 0.5 mm) across all tested conditions. The proposed algorithm runs in under 1 ms and is therefore well-suited for real-time integration into seam-tracking control systems.