基于遗传算法及神经网络的梁板结构可靠性优化设计

Reliability Optimization and Design of Beam and Plate Structure Using Genetic Algorithm and Neural Network

采用随机有限元法处理结构分析中所涉及的有关参数的随机性,建立梁板单元的安全余量,采用分枝限界法找出结构系统的主要失效模式,计算其可靠度并进行敏度分析。然后,以质量最小为目标函数,以可靠度大于设定值和设计变量满足一定条件为约束,建立了梁板结构的优化模型,将遗传算法与神经网络相结合来求解建立的优化模型。通过两种算法的结合,利用随机有限元法的基本理论,建立了一套完整的基于可靠性的随机结构系统优化设计方法。通过算例对船舱段结构梁元截面及板元厚度的优化计算结果表明,该算法对多随机因素结构系统的可靠性优化具有良好的收敛效果,用时较少,为原来的60%。

Using stochastic finite element method（SFEM）,we deal with the random variables for structural analysis and set up the safe margin of the beam and plate structure.With the branch and boundary method,we find its main failure modes,calculate its reliability index and analyze its sensitivity.Then,we develop the optimization model for beam and plate structure,using the minimum weight as the objective function and the constraints that the reliability index should be larger than the set value and that the design variables should satisfy certain conditions.We jointly use the neural network and the genetic algorithm to solve the optimization model.We use the SFEM to establish a random structural optimization and design method that is based on the reliability theory.We use the method to calculate and optimize the beams and plates of a ship cabin.The calculation results show that our method is effective for the reliability optimization of structures with multiple random factors and takes time 60% less than before,thus being very efficient.