Mission Re-Planning of Reusable Launch Vehicles Under Throttling Fault in the Recovery Flight Based on Controllable Set Analysis and a Deep Neural Network

• Published in: Aerospace Vol. 12; no. 3; p. 166
• Main Authors: Li, Keshu, Zhang, Wanqing, Yuan, Han, Zhou, Jing, Ma, Ying
• Format: Journal Article
• Language: English
• Published: Basel MDPI AG 20.02.2025
• Subjects: Algorithms; Angle of attack; Approximation; Artificial neural networks; Atmosphere; Computational efficiency; Control algorithms; Controllability; controllable set; Convex analysis; Convexity; deep neural network; Effectiveness; Efficiency; Feasibility; Flight; Landing; Launch vehicles; mission re-planning; Neural networks; Optimization; Recovery; reusable launch vehicle; Reusable launch vehicles; Space transportation; Throttling; throttling fault; Vehicles
• ISSN: 2226-4310; 2226-4310
• DOI: 10.3390/aerospace12030166

The frequent launches of reusable launch vehicles are currently the primary approach to support large-scale space transportation, necessitating high reliability in recovery flights. This paper proposes a mission re-planning scheme to address throttling faults, which significantly affect the feasibility of powered landing. To quantify the influence of throttling capability, the concept of “controllable set (CS)” is introduced. The CS is defined as the collection of all feasible initial states that can achieve a successful powered landing and is computed using polyhedron approximation and convex optimization. Based on the CS, the physical feasibility of a power landing problem under deviations from the nominal conditions can be evaluated probabilistically. Besides, a deep neural network (DNN) is constructed to enhance the computational efficiency of the CS analysis, thereby meeting the requirements for online applications. Finally, an effective re-planning scheme is proposed to deal with throttling faults in recovery flight. This is achieved by adjusting the designed angle of attack during the endo-atmosphere unpowered descent phase and selecting the associated optimal handover conditions to initiate the powered landing. The optimal re-planning parameters are determined through a comprehensive investigation of the design space, leveraging probability-based CS analysis and computationally efficient DNN predictions. Simulations verify the accuracy of the CS computation algorithm and the effectiveness of the re-planning scheme under different fault conditions. The results indicate high feasibility probabilities of 99.97%, 98.12%, and 78.52% for maximum throttling capabilities at 65%, 75%, and 85% of nominal thrust magnitude, respectively.