This study focuses on the flame straightening process for correcting welding distortion in shipbuilding structures and experimentally analyzes the effects of key process parameters on deformation behavior. In addition, the applicability of an approxim...
This study focuses on the flame straightening process for correcting welding distortion in shipbuilding structures and experimentally analyzes the effects of key process parameters on deformation behavior. In addition, the applicability of an approximate model–based approach for predicting process conditions and deriving optimal process parameters is verified. Conventional flame straightening processes are largely dependent on the experience of skilled workers, leading to considerable uncertainty in the selection of process conditions. This limitation becomes more pronounced in thick plate structures, where increased material strength and thickness result in complex deformation behavior, frequently causing quality variations and rework. To address these issues, specimen-based experiments were conducted under conditions reflecting actual line heating practices in shipyards. Material type, plate thickness, gas flow rate, and heating speed were selected as design variables, and a full factorial Design of Experiments (DOE) was applied to construct a total of 24 experimental cases. For each case, deformation distributions were measured, and the deformation at the specimen end, where the maximum deformation occurred, was adopted as the reference response for analyzing the effects of the design variables. Experimental results and sensitivity analysis confirmed that plate thickness and heating speed are the dominant factors influencing deformation behavior, while gas flow rate and material type have relatively secondary effects. It was also observed that deformation at the heating start point (Y0) establishes the baseline shape of the overall deformation distribution, with deformation gradually increasing toward the end point (Y400). Based on these findings, approximate models using a Gaussian Process Regressor (GPR) and a Response Surface Model (RSM) were developed, and their prediction accuracies were compared. The results demonstrated that the GPR model more effectively captures local nonlinear deformation behavior, making it suitable for predicting deformation in thick plate flame straightening processes. Furthermore, considering the allowable deformation limits of the straightening process, a target deformation level was defined, and optimal process conditions were derived for each material and thickness category. Verification tests conducted under the derived optimal conditions showed that, although prediction deviations occurred in some cases due to physical sensitivity, the overall deformation distributions were stably controlled within the allowable limits. These results confirm the effectiveness of the proposed process condition derivation method. The approximate model–based approach presented in this study provides a quantitative process control framework beyond experience-based practices and demonstrates its potential applicability to future automation of flame straightening processes.