PSO训练的弹性RBFNN在船型优化中的应用研究

Ship hull optimization based on PSO training FRBF neural network

针对基于SBD技术的船型优化设计问题,提出了新型神经网络近似技术。通过将粒子群算法用于FRBF神经网络的权值训练,提出了PSO-FRBF神经网络算法,通过对比分析多种方法建立的兴波阻力系数近似模型,验证了新算法的适用性与优越性。以Wigley船型为算例,以船舶主尺度和船型修改系数为设计变量,以排水体积的变化量为约束条件,引入PSO-FRBF兴波阻力系数近似模型,建立了船体总阻力优化模型,并采用模拟退火优化算法实现了主尺度与船型的优化设计,得出了可靠合理的优化船型。实例证明新型神经网络可为船舶优化设计相关阶段提供良好的技术支持。

A new neural network approximation technique is proposed, which aims at optimizing the design of a hull form based on SBD technology. Using a PSO algorithm in the weight training of the FRBF neural network,a PSO- FRBF neural network algorithm is proposed. The applicability and superiority of the new algorithm was proved by comparing the wave-making resistance coefficient approximation models were established using different methods. Then,Wigley hull is taken as example,with the principal ship dimensions and hull revision parameters as design variables and the displacement volume variation as a constraint condition; the PSO-FRBF wave-making resistance coefficient approximation model is introduced and the total resistance optimization model is established. In addition, the simulated annealing optimization algorithm is used in the optimization design of the main dimension and hull, and a reliable and reasonable optimized hull is obtained. A practical case verifies that the new neural network can provide good technical support for the related stages of ship design optimization.