基于序列Kriging-GA的薄板件定位点优化方法

为降低车身薄板焊接装夹时的重力以及夹具定位点偏差对柔性薄板件定位精度的影响，提高薄板定位方案的稳健性，基于有限元分析方法对薄板件第一基准面的定位点布置进行了优化。为了建立能反映薄板定位精度对定位点偏差敏感度的定位方案评价指标，引入影响系数法，建立了定位点偏差对薄板件定位精度的影响系数矩阵。基于影响系数平方矩阵的迹，并考虑重力对薄板定位精度的影响建立了多目标优化模型。采用Kriging代理模型构造出优化模型的目标函数，为了解决代理模型预测精度较差的问题，提出了基于序列Kriging-GA的薄板件定位点优化方法，将最优点选点准则和期望改善准则融合使用，反复在兴趣域增加样本点，更新代理模型，提高了代理模型的预测精度并保证了算法的全局搜索能力。结果表明：序列Kriging模型在兴趣域的预测误差可低至1%，具有很高的预测精度；与采用遗传算法相比，所提出的方法能减少75%仿真次数，显著提高设计效率；在不同定位点偏差组合下，零件各关键点位移量的均值与方差都显著减小，定位精度对定位点偏差的敏感度降低，优化后的定位方案具有较高的稳健性。

Locator Optimization for Sheet Metal Components Based on Sequential Kriging-Genetic Algorithm

In this study, to reduce the interference of gravity and the fixture locating error in the location accuracy of sheet metal components, and to improve the robustness of the locating scheme during the welding clamping process, the locator layout in the first datum plane of sheet metal components was optimized on the basis of finite element analysis. The influence coefficient analysis method was introduced to establish the evaluation index of a locating scheme that could reflect the sensitivity of the location accuracy to the fixture locating error. A multi-objective optimization model was created based on the trace of the influence coefficient square matrix and the effect of gravity on the location accuracy. The Kriging surrogate model was adopted to establish an objective function. To improve the prediction accuracy of the surrogate model, the sequential Kriging-GA method was proposed. By combining the optimal point principle with the expectation improvement criterion, new samples were repeatedly added to the interest-domain in order to update the surrogate model. This improved the prediction accuracy and ensured a global search ability. The results of the optimization examples revealed that the sequence Kriging model had a high prediction accuracy, and the prediction error in the interest-domain could be lowered to 1%. In comparison with the GA, the proposed method could reduce the simulation workload by 75% and greatly improve the design efficiency. Under different locating error combinations, both the mean and variance of the key point deformations decreased significantly, which implies that the sensitivity of the location accuracy to the fixture locating error was reduced, and that the optimized locating scheme had strong robustness.