MUMBAI, India, Nov. 2 -- Intellectual Property India has published a patent application (202641124151 A) filed by K. Mekala Devi; T. S. Valarmathi; R. Ashwini; and S. A. Engineering College on October 16, 2026, for Multi Dimensional Fault Localization In Nanosatellites Using Inertial And Pressure Sensor Fusion.
Inventors include K. Mekala Devi; T. S. Valarmathi; and R. Ashwini.
The application for the patent was published on October 30, 2026, under issue no. 44/2026.
Abstract: Nanosatellites, typically weighing between 1-10 kg, have emerged as cost-effective platforms for_ scientific research, Earth observation, and technology demonstration missions. However, their operation in the harsh space environment—characterized by extreme thermal variations, radiation exposure, vacuum conditions, and microgravity—makes them susceptible to multiple failure modes including sensor drift, actuator degradation, power system anomalies, and communication link disruptions. Timely fault detection, isolation, and recovery (FTMR) are therefore critical for ensuring mission success and extending operational lifetime. Conventional nanosatellite fault detection and diagnosis (FDD) systems predominantly rely on ground-station-based telemetry analysis. In such architectures, raw sensor data is downlinked to Earth, where mission control engineers analyze trends, apply fixed threshold checks, and manually diagnose anomalies. This approach introduces significant limitations: (i) communication latency ranging from minutes to hours depending on orbital passes and ground station availability; (ii) vulnerability during communication blackouts caused by orbital mechanics, antenna misalignment, or interference; (iii) inability to respond to fast-evolving faults in real time; and (iv) high operational costs associated with continuous ground monitoring. Furthermore, fixed threshold-based detection methods lack adaptability to dynamic operating conditions arising from orbital position changes, solar activity variations, battery state-of-charge fluctuations, thermal cycling, and payload activation patterns. These static thresholds often result in false positives during nominal transients or false negatives when faults manifest subtly. Recent advances in data-driven fault detection using machine learning (ML) and deep learning (DL) offer promising alternatives. Techniques such as autoencoders, Long Short-Term Memory (LSTM) networks, convolutional neural networks (CNNs), and hybrid architectures have demonstrated success in anomaly detection for industrial ToT, aerospace systems, and satellite telemetry. However, deploying such models onboard nanosatellites faces severe , constraints: limited computational resources (typically 1 MB RAM, 100 MHz CPU), strict power budgets ( 500 mW average for subsystems), minimal storage capacity, and the absence of labeled fault datasets for supervised training. Additionally, continuous high-frequency inference on resourceconstrained microcontrollers leads to excessive energy consumption, reducing mission lifetime. These challenges necessitate a lightweight, adaptive, event-triggered, onboard fault detection and localization system capable of processing multi-modal telemetry in real time while operating within nanosatellite resource envelopes. This invention addresses these gaps by proposing an intelligent, multi-dimensional fdult localization framework that fuses inertial measurement unit (IMU) and barometric pressure sensor telemetry using an Unscented Kalman Filter on Manifolds (UKF-M), followed by anomaly detection via a hybrid Neural Network—LSTM (NN—LSTM) Autoencoder optimized for Tiny Machine Learning (TinyML) deployment on an ESP32-S3 microcontroller. The system achieves autonomous fault detection, subsystem-level fault isolation, alert generation, and safe-mode initiation without continuous ground- station dependence.
Disclaimer: Curated by HT Syndication.
