Abstract
Engine transient control has been challenging due to its stringent requirements from both performance and safety. Many methodologies have been proposed such as conventional schedule-based methods, linear parameter varying, multiobjective optimization and evolutionary computations etc. These approaches have been well-established and led to a series of significant results. However, they are either not providing limit protection or requiring exhaustive computational resources, particularly when generating results into full flight envelope applications. Consequently a compromise between limit protection and computational complexity is necessitated. This note considers a sequential quadratic programming (SQP)-based method for full flight envelope investigations. The proposed method can provide important design guidance and the corresponding claims are validated through detailed analysis and simulations.
A Reinforcement Learning Approach for Transient Control of Liquid Rocket Engines
