Integration of multi-stage experiments in continuous improvement projects: a design-for-capability approach

Abstract

Quality Engineering increasingly adopts data-driven methodologies for process optimization and capability assessment in advanced manufacturing. This study proposes an integrative framework combining a systematic literature review, Response Surface Methodology, and the Normal Boundary Intersection method to optimize the industrial cladding process of AISI 316 L stainless steel on AISI 1020 carbon steel. A key scientific contribution is the methodological efficiency achieved by consolidating three traditional experimental stages: measurement system capability assessment, manufacturing process capability evaluation, and subsequent modeling and optimization, into a single, unified Design of Experiments. This is accomplished by explicitly incorporating the Process Centralization Index (Ca) and the Measurement Centralization-Dispersion Index (Cgk) as direct objective functions within the multi-objective optimization. Four input variables (wire feed rate, arc voltage, welding speed, and contact tip to work distance) were experimentally investigated via a Central Composite Design. Quadratic RSM models were developed for Ca and Cgk. Sensitivity analysis identified wire feed rate and welding speed as the most influential parameters. NBI optimization generated a Pareto frontier, revealing the trade-off between measurement capability and process centering. The optimal solution (Va = 10.43 m·min⁻¹, T = 32.12 V, Vs = 50.78 cm·min⁻¹, N = 21.44 mm) yielded Cgk = 1.37 and Ca = 1.00, both classifying the process as capable. This efficient approach enables proactive design‑for‑capability, reducing experimental burden and supporting more robust, data‑driven decision‑making in continuous improvement projects for advanced manufacturing systems.