Intelligent IoT-Driven Advanced Predictive Maintenance System for Industrial Applications

• Published in: SN computer science Vol. 6; no. 2; p. 151
• Main Authors: Dhinakaran, D., Edwin Raja, S., Velselvi, R., Purushotham, N.
• Format: Journal Article
• Language: English
• Published: Singapore Springer Nature Singapore 01.02.2025; Springer Nature B.V
• Subjects: Accuracy; Adaptability; Air compressors; Algorithms; Blended learning; Collaboration; Computer Imaging; Computer Science; Computer Systems Organization and Communication Networks; Data Structures and Information Theory; Decision making; Decision support systems; Design; Effectiveness; Efficiency; Errors; Explainable artificial intelligence; Failure; Feature selection; Federated learning; Genetic algorithms; Heavy vehicles; Industrial applications; Information Systems and Communication Service; Machine learning; Manufacturing; Optimization; Original Research; Pattern Recognition and Graphics; Predictive maintenance; Privacy; Software Engineering/Programming and Operating Systems; Statistical methods; Vision; Explainable AI; Machine learning; Predictive maintenance; IoT; Supervised and reinforcement learning
• ISSN: 2662-995X; 2661-8907
• DOI: 10.1007/s42979-025-03695-x

Predictive Maintenance (PdM) aims to ensure the continuous operation of high-risk industrial systems. This challenge is especially critical in environments where equipment failure can cause major financial losses and disrupt operations. Traditional PdM approaches, while fairly effective, often fall short in accurately predicting failures due to their reliance on simple statistical methods and predefined rules. The proposed work introduces an Intelligent IoT-Driven Advanced Predictive Maintenance System (APdM) that addresses these limitations. By integrating advanced machine learning techniques, including supervised and reinforcement learning, with IoT technologies to enhance predictive accuracy as well as the operational efficiency. The system leverages Explainable AI (XAI) to ensure transparency in decision-making and employs federated learning for distributed model training, preserving data privacy across multiple edge devices. The proposed APdM system was evaluated using the Kaggle "Air Compressor Predictive Maintenance" Dataset, which provided extensive sensor data from heavy vehicle operations. Comparative analysis against four prevailing PdM approaches—Traditional PdM, Decision Support System-PdM, Genetic Algorithm-based PdM, and AI-based PdM—demonstrated the superior performance of our system. The results show that the APdM system, powered by intelligent IoT integration, diminishes the Mean Error Percentage (MEP) by up to 38.4% compared to traditional methods and decreases the Mean Absolute Error (MAE) by 36.7% relative to DSS-PdM. Additionally, the system achieves a 78% improvement in Symmetric Mean Absolute Percentage Error (SMAPE) over GA-PdM, underscoring its robust performance in real-world industrial scenarios. These outcomes validate the effectiveness of the proposed APdM system, marking a significant improvement in the field of predictive maintenance through the intelligent integration of IoT technologies.