Journal of Quality Engineering and Management

Abstract Purpose: Multivariate and multi-state processes refer to types of processes that involve a large number of variables at each production stage, which may be interrelated. The objective of this study is to propose a novel approach for selecting, reducing, and defining new control variables in complex manufacturing processes, enabling more effective and efficient quality control.
Methodology: This study employs an applied, descriptive research methodology. Machine learning techniques and dimensionality reduction methods, such as Principal Component Analysis (PCA), are utilized, along with regression and correlation analysis. To evaluate the proposed method, a case study was conducted using real production data from the slab manufacturing process at Mobarakeh Steel Company in Isfahan.
Findings: The slab production process consisted of three main stages: furnace, secondary metallurgy, and casting. In each stage, the proposed method was applied to reduce the number of control variables. For instance, in the furnace unit, nine initial variables were grouped into three clusters, and correlation and PCA were applied within each group. Key variables were extracted, and experts validated the results. The findings indicated that this approach effectively reduces the number of quality-related variables.
Originality/Value: The novelty of this research lies in integrating machine learning and dimensionality reduction techniques to optimize quality control in multistage, multivariate processes. This method provides an effective tool for quality engineers and process analysts, particularly when traditional methods prove ineffective.