Search

This paper proposes an integrated, data-driven approach for quantifying and comparing pipe-welding operator quality performance for industrial construction projects using a Metropolis-Hastings algorithm-based fraction nonconforming estimation and an A/B testing algorithm. Existing quality-management and engineering-design data from a pipe fabrication company in Alberta, Canada are processed and analyzed to demonstrate the feasibility and applicability of the proposed approach. Through the use of a specialized Metropolis-Hastings algorithm, operator welding performances are quantified as Bayesian posterior distributions to account for uncertainty. Using an A/B testing algorithm, probabilistic differences between operator-quality performance distributions are computed to quantitatively compare their quality performance. Both visualized and quantitative decision metrics are provided for practitioners to (1) infer operator welding-quality performance for a particular weld type; (2) identify operators with exceptional quality performance; and (3) quantitatively measure differences in quality performance between operators. Implemented applications of the research findings are also discussed from the perspectives of production planning, employee training, and strategic recruiting. • An integrated, data-driven approach measures pipe welding operator quality performance for industrial construction • Quality management and engineering design data from a pipe fabrication company in Edmonton, Canada, are processed and analyzed • Operator welding performances are quantified as Bayesian posterior distributions to account for uncertainty. • Probabilistic differences between operator quality performance distributions are computed to make quantitative comparisons • Potential applications of findings are discussed in terms of production planning, employee training, and strategic recruiting [ABSTRACT FROM AUTHOR]