EXPLAINABLE MACHINE LEARNING FOR QUALITY PREDICTION AND PROCESS OPTIMIZATION IN ADVANCED MANUFACTURING

Abstract

Advanced manufacturing systems are undergoing rapid transformation through the adoption of Industry 4.0 technologies, including Artificial Intelligence (AI), Machine Learning (ML), Industrial Internet of Things (IIoT), Cyber-Physical Systems (CPS), cloud computing, digital twins, and big data analytics. These technologies generate enormous volumes of manufacturing data that can be utilized to improve product quality, optimize production processes, reduce operational costs, and support intelligent decision-making. Machine learning algorithms have demonstrated exceptional capabilities in predicting manufacturing quality, detecting process anomalies, forecasting equipment failures, and optimizing production parameters. However, many highperformance machine learning models, particularly deep learning architectures and ensemble learning techniques, operate as black-box systems that provide highly accurate predictions without offering transparent explanations for their decisions. The lack of interpretability limits industrial acceptance because manufacturing engineers require understandable and trustworthy explanations before implementing critical process modifications. Consequently, Explainable Machine Learning (XML) has emerged as an essential research area that combines predictive accuracy with model transparency to support reliable decisionmaking in advanced manufacturing environments. This research proposes an Explainable Machine Learning framework for intelligent quality prediction and process optimization in advanced manufacturing systems. The proposed framework integrates Industrial Internet of Things (IIoT)-enabled sensor data, Enterprise Resource Planning (ERP) information, Manufacturing Execution System (MES) records, and machine learning algorithms with Explainable Artificial Intelligence (XAI) techniques to provide accurate, transparent, and interpretable manufacturing predictions. Real-time production data are collected from multiple manufacturing resources, including production equipment, quality inspection systems, environmental monitoring devices, machine vision systems, vibration sensors, temperature sensors, energy meters, and process control systems. The integrated manufacturing dataset undergoes preprocessing, feature engineering, normalization, anomaly detection, and feature selection before being supplied to machine learning models such as Random Forest, Support Vector Machine (SVM), Gradient Boosting, Extreme Gradient Boosting (XGBoost), Artificial Neural Networks (ANN), and Long Short-Term Memory (LSTM) networks for quality prediction and process analysis.