LSTM-Based Post-Mission Assessment of Reusable Rocket Engines

Reusable rocket technology is having a significant impact on the economics of spaceflight, with the condition of the engine being a fundamental factor in determining its reusability.This study proposes a novel fault diagnosis framework for the evaluation of reusable rocket engines post-mission, with the aim of determining their suitability for continued operational use or the necessity for maintenance.The study utilises deep learning, specifically Long Short-Term Memory (LSTM) networks, to analyse time-series data, including vibration signals, thermodynamic parameters and operational history. This analysis captures temporal patterns that are crucial for assessing the health of the engine. This approach facilitates precise identification of degradation and failure modes that are critical to the decision-making process concerning reuse. The present methodology is centred on the "Reusability Readiness Index" (RRI), a metric that quantifies engine condition by integrating fatigue damage, thermal stress accumulation, and operational anomalies detected via LSTM's sequential data processing.The present framework was validated using post-mission data from ground tests and simulated flight profiles of a liquid rocket engine. The results obtained from this study indicate that the LSTM-based method achieves a classification accuracy of over 90% in determining the reusability or repair requirements of engines, thereby reducing false positives by approximately 15% in comparison with traditional threshold-based approaches. This enhanced precision facilitates reliable post-mission evaluations, optimises maintenance schedules, and streamlines reuse planning. At present, the framework is tailored for post-mission analysis; however, it provides a robust foundation for retrospective engine assessment, offering actionable insights into component wear and performance trends. However, the potential of this framework extends beyond the scope of this study.Future work will build upon this LSTM-based architecture to enable real-time fault monitoring during missions.The incorporation of live sensor feeds, such as pressure transients and thermal fluctuations, and the refinement of the model's computational efficiency, will facilitate a transition from post-flight diagnostics to in-flight health tracking.This evolution will enhance mission safety and operational flexibility, allowing for proactive responses to emerging issues. The adaptation of the framework to diverse engine types will be a key focus area to ensure broader applicability. This research underscores the potential of deep learning in evaluating reusable rocket engines, thereby integrating post-mission analysis with the promise of real-time capabilities.By refining reuse assessments and establishing the foundations for continuous monitoring, our work contributes to the advancement of the reliability and sustainability of reusable spaceflight systems.