Modern combat aircraft demands efficient maintenance strategies to ensure operational readiness while minimizing downtime and costs. Innovative approaches using Digital Twining models are being explored to capture inter system behaviors and assessing health of systems which will help maintenance aspects. This approach employs advanced deep learning protocols to analyze the intricate interactions among various systems using the data collected from various systems.
The research involves extensive data collection from sensors within combat aircraft, followed by data preprocessing and feature selection, using domain knowledge and correlation analysis. Neural networks are designed for individual systems, and hyper parameter tuning is performed to optimize their performance. By combining those outputs during the model integration phase, an overall health assessment of the aircraft can be generated. This assessment enables advanced fault isolation at the system level by identifying subtle deviations in system behaviors, expediting troubleshooting and corrective actions.
This research lays a solid groundwork for advancing condition-based maintenance in combat aircraft using deep learning techniques. It uncovers several promising directions for further enhancement, encompassing the exploration of multi-system interactions, establishment of real-time health monitoring, anomaly detection, root cause analysis and the incorporation of emerging deep learning methodologies for long-term performance evaluations. The present study covers the development of digital twins for identified aircraft systems with these aspects and the results obtained with outcomes are presented in this paper.