Modeling and optimization of a hot forming process for friction stir-welded tailored blanks with an industrially suitable number of test specimens

Abstract Hot forming of friction-stir-welded tailored blanks (FSW-TWBs) in industrial ramp-up poses unique challenges: a high number of coupled joining and forming parameters, stochastic trial-and-error data, and strict requirements on crack prevention and dimensional tolerances. Unlike conventional studies that rely on designed experiments or focus on weld behavior in isolation, this work introduces a coupled process analysis (CPA)–based workflow that directly mines routinely collected ramp-up data to derive multivariate, predictive process models. We demonstrate its application to an automotive B-pillar TWB, building separate models for weld-seam crack probability and part geometry accuracy. Without a formal design of experiment, CPA automatically filters out irrelevant inputs, resolves multicollinearities, and provides robust cause–effect relationships. The resulting surrogate models support in-silico optimization, delivering crack-free components within tight forming tolerances and dramatically reducing additional experimental effort. The findings establish a practical route toward data-driven process control in early production phases and lay the groundwork for future integration of CPA models into closed-loop manufacturing systems.