Knowledge embedded wind tunnel typical pressure vessel design optimization method research.

Abstract

Wind tunnels serve as an essential infrastructure for aerodynamic research and aerospace vehicle development, and pressure vessels are the most important type of structure in transonic and supersonic wind tunnels. Conventional structural design methodologies exhibit critical limitations including over-reliance on empirical specifications, computational inefficiency, and excessive conservatism. Although data-driven surrogate-based optimization approaches partially mitigate these issues, their generalizability variable operation condition remains limited. This study proposes a knowledge-embedded hierarchical Kriging (KEHK) framework that synergistically integrates the pressure vessel design specification with adaptive multi-fidelity modeling. The methodology introduces three key innovations: a knowledge-embedded sequential sampling method based on pressure vessel design specification, an adaptive hierarchical Kriging architecture incorporating multi-fidelity training samples, and a novel condition-mapping protocol to enhance cross-scenario generalizability. The experimental validation of a transonic wind tunnel acceleration section demonstrated a 26.2% structural weight reduction while maintaining operational integrity, coupled with a 150 × computational efficiency improvement over conventional finite element analysis for single-iteration simulations. Comparative evaluations revealed the KEHK model's superior generalization capability, achieving a prediction error of <15% across 0.1-2.0 the operational pressure ranges, significantly outperforming the conventional hierarchical Kriging, BPNN and nonadaptive KEHK variants. These advancements have established the framework as a robust solution for next-generation wind tunnel engineering applications, effectively bridging the gap between the computational efficiency and operational reliability.