Optimized Design for Armature Lightweighting and Contact Pressure Distribution Uniformization Based on BP Neural Network and Genetic Algorithm

Abstract

The variation in armature structural parameters has a profound impact on electromagnetic railgun performance. In order to enhance the electrical contact performance of armature and rail interfaces and achieve armature lightweighting, a novel approach is proposed that combines backpropagation (BP) neural networks with genetic algorithms for optimizing armature structures. Employing Latin hypercube experiments, structural dimension samples were extracted for both flat armatures and convex-arc armatures, and a training dataset was generated through finite element simulations. Mapping models were constructed based on BP neural networks to relate the contact pressure distribution coefficient, maximum stress, mass, and total contact force individually to the armature structural parameters. Incorporating a comprehensive evaluation index denoted as “W” as the fitness function for the genetic algorithm, global optimization of armature size was carried out. The optimized results were subsequently validated through finite element comparative analysis. The outcomes revealed that, following optimization of flat armatures, a 10.4% reduction in mass and a 55.7% decrease in the contact pressure distribution coefficient were achieved. For convex-arc armatures, a 25% reduction in mass and a 46.5% decrease in the contact pressure distribution coefficient were observed. Simultaneously achieving armature lightweighting and improving the uniformity of contact pressure distribution on armature and rail interfaces, this methodology offers a novel perspective and serves as a valuable reference for addressing the optimization of armature structures under complex operating conditions.