Quality management in chemical processes through fuzzy analysis: a fuzzy C-means and predictive models approach

Abstract

Ensuring high levels of quality and efficiency is essential for compliance with ISO standards in chemical manufacturing. Traditional methods, such as Statistical Process Control (SPC) and Six Sigma, often lack adaptability and fail to offer interpretable insights. This study proposes a hybrid quality control model based on Explainable Artificial Intelligence (XAI), integrating fuzzy C-means clustering (FCM), machine learning (ML), and Fuzzy Inference Systems (FISs) to enhance defect prediction and interpretability in industrial environments. The approach uses fuzzy clusters to segment production batches, improving the understanding of process variability. A supervised ML model (XGBoost) is trained on historical data to predict defect probabilities, while an explainable FIS refines the final assessment using expert-defined rules. XAI techniques (SHAP and LIME) offer transparency and insight into the decision-making process. Experimental validation using a real-world white wine dataset, evaluated in terms of accuracy and interpretability, shows that the proposed model outperforms traditional approaches in both predictive performance and transparency. The results demonstrate the effectiveness of combining unsupervised clustering, predictive analytics, and fuzzy reasoning in an Industry 4.0 framework. This study provides a scalable and adaptable solution for real-time quality control in chemical manufacturing, improving decision support systems and enabling automated and explainable quality assessments.