MUMBAI, India, June 30 -- Intellectual Property India has published a patent application (202641073735 A) filed by Sr University on June 14, 2026, for Machine Learning Based Predictive Framework For Proactive Fault Management In Cyber-Physical Systems.
Inventors include Mr. Sathish Kumar Soora; and Dr. R. Vijaya Prakash.
The application for the patent was published on June 26, 2026, under issue no. 26/2026.
Abstract: The present invention discloses a machine learning based predictive framework for proactive fault management in complex cyber-physical systems. The Framework integrates five functional modules: (1) a multi-source data acquisition and integration module collecting heterogeneous time series data from sensors, controllers, and logs; (2) a preprocessing and feature engineering pipeline incorporating noise filtering, imbalance handling via SMOTE, and extraction of time domain, frequency domain, and health indicator features; (3) a multi-objective model development module combining supervised classifiers (Random Forest, XGBoost, SVM, LSTM, 1D CNN, Transformer) for fault detection and classification with unsupervised autoencoders and isolation forests for anomaly detection and LSTM based regression for Remaining Useful Life (RUL) estimation; (4) a real time predictive inference and decision engine operating on streaming sensor data through sliding windows, computing fault probabilities, fault type labels, RUL estimates, and anomaly severity indices, and triggering proactive maintenance alerts via a configurable decision logic module; and (5) a modular edge-fog-cloud deployment architecture integrated with industrial protocols (MQTT, OPC-UA, Modbus, REST API) and supported by a continuous learning loop for model adaptation under concept drift. Experimental evaluation demonstrates fault detection accuracy of 98.1%, fault classification accuracy exceeding 95%, RUL prediction errors of 2-6 hours, and anomaly detection accuracy of 96.8%, confirming significant improvements over conventional threshold based monitoring approaches.
Disclaimer: Curated by HT Syndication.
