摘要
螺旋输送器的动态特性设计可归结于特征值反问题的求解。针对结构参数到结构响应之间的非线性映射关系,通过一种基于神经网络代理模型的优化策略,采用正交试验设计在设计空间中选择初始样本点,构造神经网络代理模型,神经网络结合遗传算法求解,利用神经网络的非线性拟合能力和遗传算法的非线性寻优能力,引入训练后的BP神经网络预测结果作为个体适应度值,获得全局最优值及对应输入值。解决了遗传算法能获全局最优解与有限元大量结构重分析之间的矛盾,是结构反问题的一种有效求解策略。
The design of the helical conveyor with dynamic properties can be classified as the solution to inverse eigenvalue problem. In order to deal with non-linear mapping function between structural parameters and mechanical properties, the optimization strategy using BP neural network surrogate model is proposed. The surrogate model is constructed with initial sampling points generated by orthogonal experiment design. A strategy of combining genetic algorithm （GA） and BP neural network was proposed, optimization solution can be solved by using the nonlinear approach capability of BP neural network and the nonlinear search operation of GA by em- ploying the individual fitness value coming from the forecast evalua- tion based on the BP neural network system to the optimization of a genetic algorithm which deal with the defects between genetic algo- rithm and structural reanalysis. This method is effective to structur- al inverse problems.
出处
《机械设计》
CSCD
北大核心
2013年第5期18-20,37,共4页
Journal of Machine Design
基金
国家自然科学基金资助项目(50745018)
湖南省教育厅科研资助项目(10C0552)
关键词
螺旋输送器
反问题
BP神经网络
遗传算法
特征值
helical conveyor
Inverse problem
BP neuralnetwork
genetic algorithm
eigenvalue
