AI Based predictive Maintenance System for Industrial Equipment

Unexpected equipment failures can interrupt production, increase maintenance expenses, and create safety concerns in manufacturing environments. This paper describes an AI-based predictive maintenance system, called PrimeAI, that uses live IoT sensor information to identify signs of equipment failure before the failure actually happens. The system combines a Long Short-Term Memory (LSTM) network, which learns changes in sensor readings over time, with an XGBoost model that looks for abnormal patterns in engineered sensor features. Because failure cases are much less common than normal operating cases, SMOTE oversampling and class-weighted training are used to reduce the effect of class imbalance. The system was evaluated using streaming sensor information collected from 50 industrial machines located across three manufacturing facilities. On unseen test data, PrimeAI achieved 94.2% precision and 91.7% recall. The model provided an average warning of 14.3 days before a documented failure. During deployment, unplanned downtime decreased by 68%, while maintenance costs decreased by 42% when compared with the baseline maintenance approach. These results indicate that combining IoT monitoring with machine learning can give maintenance teams useful advance information and support better planning.