A Data-Driven Method for Typical Landing Gear Structure Optimization Based on Neural Networks

The landing gear, as a crucial component of an aircraft, is pivotal for maintaining the safety and reliability of air travel. This study introduces a data-driven structural optimization method aimed at mitigating the peak strain on the landing gear’s rocker arm. The initial phase involves selecting nine design variables for parametric modeling to generate an initial dataset. Subsequently, the Maximum Information Coefficient (MIC) technique is used to conduct a parameter sensitivity analysis, enabling the identification and elimination of variables with minimal influence. A comparative analysis between the Genetic Algorithm–Backpropagation Neural Network (GA-BPNN) and BPNN reveals that GA-BPNN has a superior fitting capability on the enhanced dataset. By applying Particle Swarm Optimization (PSO), the optimal solution for GA-BPNN is identified. The implementation of this optimized method results in a 38.16% reduction in peak strain, validating its feasibility and reliability in enhancing aircraft safety.