Paper Title
PREDECTIVE MAINTENANCE USING IoT

Abstract
Unscheduled industrial equipment failures result in substantial production losses, elevated operational expenditures (OPEX) and increased safety risks. Traditional reactive and scheduled maintenance strategies often lack the predictive foresight required for optimal resource allocation. This research presents a scalable predictive maintenance framework that integrates multi-modal Internet of Things (IoT) data acquisition with a hybrid CNN-LSTM architecture for Remaining Useful Life (RUL) estimation and an ensemble learning classifier comprising Random Forest, XGBoost, LightGBM, Support Vector Machines and Gradient Boosting for earlystage fault identification. The proposed methodology leverages realtime vibration, thermal and current signatures captured via an IoT-enabled sensor network. These highfrequency data streams are processed through a specialized timeseries learning pipeline to detect subtle anomalies before catastrophic failure occurs.This framework facilitates the transition toward Industry 4.0 by enabling autonomous, data-driven maintenance decisionmaking.Experimental validation demonstrates an R² of 0.9935, MAE of 1.68 minutes, 93.4% classification accuracy and F1-score of 93.3%, confirming the framework's suitability for realtime industrial deployment. Keywords - Anomaly Detection, Condition Monitoring, CNN-LSTM, Deep Learning, Ensemble Learning, Industry 4.0, Internet of Things (IoT), Predictive Maintenance, Remaining Useful Life (RUL), Time-Series Analysis.