基于神经网络理论的船舶结构可靠性优化设计

THE RELIABILTY-BASED OPTIMUM DESIGN OF SHIP'S STRUCTURE BASED ON NEURAL NETWORK THEORY

本文将人工神经网络理论应用于船舶纵向结构的可靠性优化设计中，采用模拟退火和玻尔茨曼机原理，在优化过程中可避开局部最优解，取得系统的近似最优解。本文提出了优化过程中设记忆器的方法，有效地提高了优化精度。本文还提出了几种改进措施，使计算时间明显减少。通过实际船舶的舯剖面可靠性优化设计，证明了上述方法和作者开发的计算机程序的有效性和通用性。

This paper applies neural network theory to the reliability-based optimum design of a ship's longitudinal structure and adopts simulated annealing method and Boltzmann machine principle by which we can avoid the local optimum and find the approximate global optimum of the system. Since only one approximate global optimum can be reached using neural network theory for optimum designs, an approach which sets a memory in optimization process is developed and precision is improved efficiently. Some new improvements are introduced to the simulated annealing method, which reduce design time greatly. The validity and general applicability of the above methods and program are verified by the results of the reliability-based optimum design of real midship cross section.