论文部分内容阅读
针对复杂的实际工程多目标优化问题,提出代理模型引导采样的多目标优化方法。通过自适应的加强径向基函数构造的代理模型寻找Pareto优化解集,从找到的解集中通过一定策略筛选出部分作为样本点加入到下代样本空间中,样本点随着迭代的进行越来越靠近全局Pareto最优解集。从当前所有样本点中获得Pareto解,并根据其分布情况作为收敛条件。该方法中代理模型仅仅用来引导采样,也不需要反复计算大量样本点验证代理模型的精度,得到的解都被实际模型验证过。在典型多目标测试函数中体现了精度和效率。最后成功应用于薄板冲压成形变压边力优化中,表明了具有解决多目标实际工程优化问题的能力。
In order to solve the complex multi-objective optimization problem of practical engineering, a multi-objective optimization method of agent-based guidance sampling is proposed. Pareto optimal solution set is searched through an adaptive agent model constructed by strengthening radial basis functions. From the found solution set, a certain strategy is used to select the part as the sample point to be added to the next-generation sample space. As the iteration progresses The closer to the global Pareto optimal solution set. Pareto solutions are obtained from all the current sample points, and the convergence condition is based on their distribution. In this method, the proxy model is only used to guide the sampling and does not need to repeatedly calculate the accuracy of a large number of sample points to verify the proxy model. The obtained solutions are validated by the actual model. Accuracy and efficiency are demonstrated in typical multi-objective test functions. Finally, it has been successfully applied in the optimization of the pressure side force of sheet metal stamping, which shows that it has the ability to solve the practical engineering optimization problem of multi-objective.