PREDICTIVE MAINTENANCE IN INDUSTRIAL AUTOMATION: A SYSTEMATIC REVIEW OF IOT SENSOR TECHNOLOGIES AND AI ALGORITHMS

Abstract

Predictive maintenance has become a crucial strategy in industrial automation, utilizing AI-driven analytics, IoT sensor technologies, and advanced computing frameworks to enhance equipment reliability and operational efficiency. This systematic review, based on an in-depth analysis of 78 high-quality peer-reviewed studies, follows the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines to ensure a rigorous and transparent evaluation. The findings demonstrate that AI-based predictive maintenance models, particularly machine learning and deep learning techniques such as convolutional neural networks (CNNs) and long short-term memory (LSTM) networks, improve failure prediction accuracy by 30-60%, leading to 25-50% reductions in maintenance costs and increased equipment uptime. The role of IoT-enabled condition monitoring is evident in 49 studies, where real-time fault detection improved predictive accuracy by 15-35%, contributing to a 20-45% reduction in unnecessary maintenance activities. Furthermore, edge and cloud computing integration, analyzed in 51 studies, reveals that edge computing significantly reduces response time by 40-70%, while cloud computing enhances large-scale model training with a 60% increase in computational efficiency. The adoption of digital twin technology, supported by 42 studies, has demonstrated 25-50% higher predictive accuracy, reducing unplanned downtimes by 35-55%, although challenges related to high implementation costs and data integration persist. Sustainability has also emerged as a key focus, with 39 studies indicating that AI-driven predictive maintenance reduces energy consumption by 20-45%, leading to a 15-35% decrease in carbon emissions through optimized maintenance scheduling and energy-efficient AI solutions. Despite these advancements, challenges remain, as 31 studies highlight data quality issues, 19 studies raise cybersecurity concerns, and 14 studies discuss the interpretability limitations of deep learning models, which hinder trust and adoption. This review provides a comprehensive synthesis of AI-driven predictive maintenance, emphasizing its transformative potential in industrial automation while also underscoring the need for further research in model interpretability, cybersecurity, and cost-effective implementation to fully harness its capabilities for sustainable, intelligent, and highly efficient maintenance operations.