An Intelligent Industrial IoT Framework for AI-Based Predictive Maintenance and Smart Manufacturing

Abstract

Predictive maintenance has emerged as a key enabler of smart manufacturing by minimizing equipment failures, reducing operational costs, and improving production efficiency. This study proposes and validates an Artificial Intelligence (AI)-driven Industrial Internet of Things (IIoT) framework for predictive maintenance and real-time manufacturing optimization. The proposed framework integrates a hybrid edge–cloud architecture with advanced machine learning models to enable low-latency analytics, intelligent decision-making, and continuous process optimization. Performance evaluation was conducted using two publicly available benchmark datasets: the NASA Turbofan Engine Degradation dataset for Remaining Useful Life (RUL) prediction and the SECOM manufacturing dataset for fault classification. For RUL estimation, Long Short-Term Memory (LSTM) and GPU-accelerated, hyperparameter-optimized Extreme Gradient Boosting (XGBoost) models were developed and compared, with the LSTM model achieving the best predictive performance (R² = 0.79). To address the severe class imbalance in the SECOM dataset, a robust preprocessing pipeline incorporating Synthetic Minority Over-sampling Technique (SMOTE) and Principal Component Analysis (PCA) was implemented, significantly improving failure detection compared with conventional baseline models. Furthermore, system-level simulations demonstrated the practical benefits of the proposed framework, achieving a 93.28% reduction in decision latency through edge computing and a 75.51% reduction in operational downtime. These findings demonstrate that AI-enabled IIoT architectures can substantially enhance predictive maintenance, operational resilience, manufacturing efficiency, and resource utilization, supporting the realization of intelligent and sustainable Industry 4.0 ecosystems.