MUMBAI, India, July 7 -- Intellectual Property India has published a patent application (202641076825 A) filed by Sns College Of Technology on June 22, 2026, for Apuwatch: Ai-Powered Auxiliary Power Unit Predictive Maintenance System.
Inventors include Dr. M. Siva Ramkumar; Farhana Sabreen R; Hareesh Kumar K; Kavin Charles J; Lakshana V; Dr. Husna Khouser G; and Mrs. B. Sajitha.
The application for the patent was published on July 03, 2026, under issue no. 27/2026.
Abstract: The present invention relates to an AI-powered end-to-end Auxiliary Power Unit Predictive Maintenance System designed to predict the Remaining Useful Life of APU engines from multi-sensor operational time-series data. The system incorporates a synthetic fault injection module that augments the NASA CMAPSS FD002 benchmark dataset with five distinct fault degradation patterns using Isolation Forest anomaly scoring, enabling robust model training across diverse real-world fault scenarios. An engine-based data splitting strategy eliminates temporal data leakage, and a feature engineering pipeline derives sixty-nine temporal and statistical features from fourteen active sensors comprising rolling mean, rolling standard deviation, rolling trend, operating-condition normalisation, and anomaly score. A LightGBM gradient boosting regression model trained on these features produces cycle-accurate Remaining Useful Life predictions with R-squared exceeding 0.99 on benchmark test data. The system provides a reusable inference pipeline accessible via CLI and REST API that processes unseen sensor CSV files and generates combined prediction and metrics reports. A web-based visualisation frontend designated APUWATCH delivers an aviation cockpit-styled dark interface with animated metric gauge cards, an interactive RUL time-series chart, dynamic maintenance status classification, and a semi-circular engine health index gauge. The system is deployed on Google Cloud Platform using Cloud Run for serverless serving, Cloud Storage for artefact and data management, Cloud SQL for prediction history and drift monitoring, and Cloud Scheduler for automated weekly retraining with MLflow model versioning. The invention bridges the gap between raw engine sensor data and actionable maintenance decisions, enhancing aviation safety, reducing unplanned downtime, and enabling condition-based maintenance in production environments.
Disclaimer: Curated by HT Syndication.
