论文部分内容阅读
针对实际工程中不确定性因素与产品质量特性之间不具有显式函数关系的稳健优化问题时,代理模型的精度成为关键。本文提出一种基于支持向量机代理模型和粒子群算法的稳健优化方法,采用拉丁超立方试验设计采样布点,优化问题的目标性能函数、约束函数的均值和标准差由具有自动参数优化的支持向量机模型替代,采用粒子群优化算法对稳健优化模型进行求解。以典型的两杆结构优化为例,结果表明支持向量机代理模型的综合性能比常用的响应面、BP神经网络和Kriging模型更优越,稳健优化结果比较理想,为复杂产品的不确定性设计优化提供了一种新的思路。
In order to solve the problem of robust optimization that does not have an explicit functional relationship between the uncertainties and product quality characteristics in practical engineering, the accuracy of the proxy model becomes the key. In this paper, a robust optimization method based on SVM agent model and particle swarm optimization algorithm is proposed. The Latin hypercube test is used to design the target performance function of the sampling points and the optimization problem. The mean and standard deviation of the constraint function are determined by the support vector with automatic parameter optimization Machine model instead, using particle swarm optimization algorithm to solve the robust optimization model. Taking the typical optimization of two-bar structure as an example, the results show that the comprehensive performance of the SVM agent model is superior to the commonly used response surface, BP neural network and Kriging model, the robust optimization results are ideal, and the design and optimization of the uncertainty of complex products Provides a new way of thinking.