ABSTRACT This paper develops a reliable statistical tool for monitoring the mechanical properties of steel production under mild contaminations, heavy tails, and drifting baselines. Specifically, a dual‐robust adaptive joint CUSUM procedure is proposed and thoroughly investigated to monitor tensile strength when manufacturing reinforcing steel. The proposed procedure utilizes a robust median and median absolute deviation estimator for Phase I baselining and Phase II bounded‐influence monitoring based on Huber score functions, coupled with an adaptive reference update by rolling dispersion estimates and joint monitoring of mean and scale changes using a single cumulative statistic. A Monte Carlo simulation‐based algorithm is developed to achieve an accurate control of the in‐control average run length. Comprehensive computer experiments are conducted to investigate the monitoring performance under normally distributed, heavy‐tailed, contaminated, and drifting baselines, as well as under different mean and variance changes. Our results show that the proposed scheme preserves the near‐target in‐control performance under a clean dataset, significantly improves the robustness over the classical and the fixed‐robust CUSUM charts under non‐normal baselines, and optimally preserves the detection capability under Gaussian scenarios. The proposed method is applied to monitor the tensile strength of an integrated steel production process. The industrial experiments show a stable monitoring performance and a reduced sensitivity to transient variations as compared to the classical approach. Overall, the proposed framework provides a calibrated and ready‐to‐use solution to monitor mechanical properties in steel production.
Ali et al. (2026) studied this question.