为了求解需要消耗大的计算资源的非线性约束的工程优化问题，提出一种基于多代理模型的自适应约束优化算法。首先给优化问题中的目标函数及每一个约束函数分配一个代理模型候选集，其次通过交叉检验确定每一函数相应候选集内代理模型对研究问题拟合性能排序,并根据排序结果构造一系列原优化问题的近似模型并通过序列二次规划算法求解，当候选集内代理模型数目不一致时，优先选择具有最优性能的代理模型。候选集内代理模型是保留或删除基于代理模型拟合性能评价结果确定，新样本通过求解近似模型及组合应用非均匀变异算子和 杂交算子共同确定。最后应用提出的算法求解四个典型的数学优化问题，结果表明基于自适应优化算法得到的近似优化解均较好的逼近于理论最优解；同时应用提出算法对汽车轮毂轴承单元轴铆工艺中的铆头成形曲面进行优化设计，优化结果较好.

In order to solve the nonlinear constraint engineering optimization problem which requires high computational resource，a self-adaptive constraint optimization algorithm based on multiple surrogates is proposed. First a candidate set including surrogate model will be assigned to the objective function and each constraint function in the optimization problem, then based on the cross-check, the ranking of the corresponding surrogate model in the candidate models will be determined by the fitting performance of the function; The surrogate model is selected to construct approximate model of the original optimization problem according to the result of the ranking, solved by the sequential quadratic programming algorithm. When the number of surrogate models is different in the candidate set, the priority is selected. The surrogate model will be kept or deleted in the candidate set is determined by the evaluation results of the fitting performance of the function, and the new sample in the algorithm is got by solving the approximate model and the combination of the application of the inhomogeneous variation operator and the hybrid operator. Finally application of the proposed constraint optimization algorithm to solve four representative mathematical optimization problems, the results show that the approximate optimization solutions based on the adaptive optimization algorithm are better approximated to the theoretical optimal solution; And used to optimize the design of the rivet head forming surface in the shaft clinching process of the vehicle wheel bearing unit, the optimization result is better.