INITIAL RESULTS OF AN ARTIFICIAL INTELLIGENCE-BASED PREDICTIVE MAINTENANCE SYSTEM FOR INDUSTRIAL EQUIPMENT

Abstract

Predictive maintenance (PdM) has emerged as a critical strategy for reducing unplanned downtime in industrial environments. This paper presents a preliminary investigation into an AI-powered predictive maintenance framework designed for complex industrial equipment. The proposed system integrates multi-sensor data acquisition, deep learning-based anomaly detection, and a real-time decision support module. Experimental evaluation on a benchmark industrial dataset demonstrates that the proposed LSTM-CNN hybrid model achieves 94.3% fault detection accuracy, outperforming conventional threshold-based and classical machine learning baselines by 11–17%. These preliminary findings confirm the viability of deploying AI-driven PdM systems within Industry 4.0 infrastructure and lay the groundwork for full-scale industrial validation.