INTELLIGENT STATISTICAL PROCESS CONTROL FOR ADVANCED FOOD PRODUCTION: CONTRIBUTIONS TO THE SUSTAINABLE DEVELOPMENT GOALS (SDGS)

Abstract

Objective: This study demonstrates how applied statistical methods combined with artificial intelligence can enhance production control in food technology by improving product quality, reducing variability, and supporting data-driven industrial decision-making. Theoretical Framework: The research is grounded in Statistical Process Control (SPC), Shewhart’s classical quality control principles, and Industry 4.0 concepts integrating machine learning into industrial monitoring. Support Vector Machines (SVM), together with Random Forest and Multilayer Perceptron (MLP) models, are investigated as intelligent extensions of traditional control-chart-based SPC. Method: Classical Statistical Process Control (SPC) techniques based on X̄ and R control charts were integrated with supervised machine learning. Real production data from an industrial filling process producing nominal 80 g “Choco Flips” packages were analyzed. Control limits were analytically derived and implemented in Python. A Support Vector Machine (SVM) classifier was trained on 300 samples using multiple kernel functions, with hyperparameter optimization performed via GridSearchCV and model evaluation based on cross-validation and ROC analysis. To assess robustness and comparative performance, additional supervised learning models, including Random Forest and Multilayer Perceptron (MLP), were also evaluated. Results and Discussion SPC analysis revealed intermittent special-cause variation in the X̄ chart, while the R chart indicated generally stable short-term variability. For predictive quality monitoring, the SVM with an RBF kernel achieved the highest and most consistent performance (accuracy ≈ 0.98, ROC AUC ≈ 0.92), confirmed through cross-validation and hyperparameter optimization. Random Forest and MLP models demonstrated comparable predictive accuracy, further validating the robustness of the proposed intelligent SPC framework. Research Implications: The integration of machine learning with SPC enables real-time anomaly detection, automated decision support, and reduced reliance on repetitive manual measurements, while maintaining reliable quality assurance. This real-time, intelligence-based SPC framework is particularly suitable for small and medium-sized food production enterprises seeking scalable and efficient quality-control solutions. Originality/Value: The study presents a practical and scalable intelligent SPC framework that integrates classical control charts with machine learning for predictive, real-time quality monitoring in regulated food production environments.