论文部分内容阅读
基于Kriging代理模型提出了一种同时考虑预测响应值及其不确定性的多点加点准则,并基于该准则发展了一套序列近似优化方法。多点加点准则基于初始样本信息和所预测的对象函数特征增加新样本集,以在寻优迭代过程中自适应地提高代理模型的精度。该文方法依据多点加点准则在一次迭代中增加多个空间无关的新样本点,适用于多机同时计算或并行计算,从而提高计算效率。以两个经典的数学函数为例,将该优化方法与期望提高准则方法进行了比较,结果表明该文提出的优化方法能够有效地提高最优解的全局性。将方法用于一盒式注塑件的成型工艺优化设计,优化结果也表明了该方法的有效性。
Based on the Kriging proxy model, a multi-point plus criterion is proposed which considers both the predicted response value and its uncertainty. Based on this criterion, a set of sequence approximate optimization methods is developed. The multi-point plus criterion adds a new sample set based on the initial sample information and the predicted object function features to adaptively improve the accuracy of the proxy model during the optimization iteration. According to the multi-point plus criterion, this method adds many new sample points that are not related to space in one iteration, which is suitable for simultaneous calculation of multiple machines or parallel computation, so as to improve computational efficiency. Taking two classical mathematical functions as an example, this optimization method is compared with the expectation raising criterion method. The results show that the optimization method proposed in this paper can effectively improve the overall quality of the optimal solution. The method is applied to optimize the molding process of a box-type injection molding. The optimization results also show the effectiveness of the method.