摘要
疲劳累积损伤是一个非稳态能耗过程,可以用遗传算法优化后的3层2 7 1BP神经网络来描述疲劳损伤的非线关系,经仿真验证表明,该神经网络具有较高的精度和泛化能力.通过对材料疲劳损伤临界值和载荷的分散性的分析研究,建立了疲劳失效动态准则,并运用蒙特卡罗随机抽样法对材料疲劳寿命的可靠性进行了仿真验证;对调质45号钢在随机载荷和2级载荷作用下,进行了疲劳寿命可靠性仿真计算,仿真结果与实验结果和理论分析比较吻合.
The fatigue accumulative damage is an unsteady process. Three-layer 2-7-1 BP network, which has been optimized with genetic algorithm, is used to truly describe the complicated relation of the fatigue damage. And the simulation results of BP network are proved exact. The dynamic criterion of fatigue failure considering the randomness of the critical value of fatigue damage and circular load is established. Moreover, the reliability of fatigue life of materials can be simulated with Monte-Carlo stochastic sampling method. For the material of 45# steel under the action of random load or two-level load, the simulation results of reliability of fatigue life is coincident between the values of experiments or theory.
出处
《海军工程大学学报》
CAS
北大核心
2005年第2期67-71,共5页
Journal of Naval University of Engineering
关键词
疲劳损伤
疲劳寿命
可靠性
神经网络
遗传算法
fatigue damage
fatigue life
reliability
neural network
genetic algorithm
