摘要
在结构参数化有限元分析的基础上,获取结构随机设计变量与功能函数的关系,构建随机设计变量到功能函数的神经网络模型。由神经网络表达式得到功能函数和梯度显式表达式,进而计算可靠度以及可靠度对随机变量的灵敏度。以计算可靠度为非线性约束方程,以计算可靠度灵敏度为目标函数,采用遗传算法建立优化模型,得到灵敏度最小化的随机设计变量。导弹发射装置锁制钩优化设计实例表明,该方法在提高锁制钩概率可靠度的同时,能够降低可靠度灵敏度,为实施发射装置结构可靠性优化和稳健设计提供通用、有效的方法。
Based on the finite element analysis of structural parametrization,the relationship between structural stochastic design variables and performance function is obtained,and a BP neural network model is set up to provide the explicit performance function and gradient expressions. Thus,the probabilistic reliability and reliability-based sensitivity can be calculated. Taking the probabilistic reliability as nonlinear constraint equation,and the reliability-based sensitivity as target function,the optimal model is established based on genetic algorithm to obtain the stochastic design variables of minimization sensitivity. The missile launcher detent hook optimization example indicates that the probabilistic reliability is increased,meanwhile,the reliability-based sensitivity is significantly decreased. The proposed methodology provides a general and effective method for launcher structural reliability optimization and robust design.
出处
《航空兵器》
2017年第5期54-59,共6页
Aero Weaponry
关键词
结构稳健可靠性
优化设计
有限元模型
神经网络
遗传算法
发射装置
structural robust reliability
optimization design
finite element model
neural network
genetic algorithm
launcher
