An enhanced intelligent-driven learning framework for adaptive multi-objective optimization control of pintle solid-fuel rocket engines

Intelligent learning techniques have shown increasing potential in addressing the complex control challenges of pintle solid-fuel rocket engines (PSREs), which play a key role in energy conversion applications for advanced aerospace propulsion systems. However, Intelligent thrust control in PSREs remains challenging due to strong nonlinearities and uncertainties, which adversely affect the rapidity and stability of energy output. In this paper, an enhanced intelligent-driven learning framework is proposed for achieving the adaptive multi-objective optimization control of a PSRE. The optimization objectives focus on improving the dynamic response of energy output, minimizing the finite entry time, and ensuring system safety by limiting thrust undershoot, which are key indicators of combustion instability and energy loss. To efficiently optimize the five-dimensional controller parameter space, a hybrid intelligent-driven approach is developed through a structured procedure. On this basis, optimal Latin hypercube sampling (OLHS) is first used to establish a well-distributed sample set. Moreover, a high-fidelity surrogate model is built using a radial basis function neural network (RBFNN), with its parameters refined by iterative optimization algorithm. Finally, this iterative optimized RBFNN model enables the multi-objective whale optimization algorithm (WOA) to effectively identify the pareto-optimal solutions. Results show that the optimized controller achieves about 27% reduction in finite entry time and maintains undershoot within safe margins, significantly enhancing the energy output response speed and control stability of the propulsion system. The proposed enhanced intelligent-driven learning framework offers a reliable and efficient solution for the performance optimization of energy-intensive combustion systems.