Here is a brand new article co-written by Zsolt T. Kosztyán iASK researcher and it was released in the journal Results in Engineering.

Abstract

Control charts are vital for real-time production process monitoring tasks, as they help monitor and maintain consistent product quality and detect variations early. In statistical process control (SPC), the use of risk-based control charts (RBCCs) improves the traditional control chart logic by considering decision risks, which can be influenced by factors such as measurement errors or sampling procedures. The traditional RBCC methodology relies on one of the fundamental assumptions: that process data are statistically independent. However, the characteristics monitored by process management are not always statistically independent of one another. This raises significant concerns for practitioners, as many processes in real-world scenarios generate autocorrelated observations. In this study, the issue of autocorrelation in the commonly used RB average charts (𝑋̄ and the exponentially weighted moving average (EWMA)) is addressed by incorporating an autocorrelated structure into the RB design. The proposed approach extends the RB average charts to better account for the autocorrelation problem. To validate the adequacy of the proposed design, Monte Carlo simulations are carried out under different autocorrelation settings. The numerical results indicate that the presence of autocorrelation significantly distorts the performance of existing RB 𝑋̄ and EWMA designs in terms of the total cost of a decision. In contrast, the overall cost associated with the proposed RB charts is less affected as the autocorrelation factor increases. A real dataset is used to demonstrate the proposed design and its applicability in real-world scenarios.

Keywords: Autocorrelation, Control charts, Decision error, EWMA, Measurement uncertainty, Optimization, Phase I

The article can be read HERE with full text.