针对整车空气悬架系统在高维不确定参数影响下动态特性响应不易求解的问题，提出了一种基于自适应多项式混沌展开的系统动态特性分析方法。首先，建立了整车七自由度空气悬架系统动力学模型，基于该模型求解系统动态特性响应；然后，为缓解传统多项式混沌展开模型在处理高维不确定参数时引发的“维度灾难”问题，将系统动态特性响应函数按照基于切割点的高维模型进行分层处理，将响应函数分解为常数项、一阶和二阶组成函数；接着，应用多项式混沌展开模型对每个组成函数进行处理，以获得组成函数的具体表达式；进一步，为解决传统均匀采样所导致的低概率区域样本冗余与关键敏感区域采样不足的问题，引入自适应矩形分割方法对不确定参数取值空间进行采样，并结合回归法求解多项式系数；最后，结合算例分析验证方法的有效性。考虑24维不确定参数的某整车空气悬架系统动态特性分析结果表明，多项式展开阶数及收敛阈值的合理选取能有效保证所提方法的拟合精度与效率；所提方法在系统动态特性响应分析中的综合性能整体优于传统多项式混沌展开方法；部分参数的不确定性对整车系统动态特性响应存在较大影响，且不同响应的关键影响参数存在一定差异，这些在优化设计中应予以重点关注。

To address the difficulty in calculating the dynamic characteristic responses of the full-vehicle air suspension system considering high-dimensional uncertain parameters, a dynamic characteristic analysis method based on the adaptive polynomial chaos expansion is proposed. First, the seven-degree-of-freedom dynamic model of full-vehicle air suspension system was established, and the system dynamic characteristic responses were calculated based on this model. Then, to alleviate the curse of dimensionality problem encountered by the traditional polynomial chaos expansion models when handling high-dimensional uncertain parameters, the system dynamic characteristic response functions were processed in a layered manner based on high-dimensional models with cutting points. The response functions were decomposed into constant terms, First-order and second-order composition functions. Next, polynomial chaos expansion models were applied to handle each composition function to obtain their specific expressions. Further, to address the problem of sample redundancy in low-probability regions and insufficient sampling in critical sensitive regions caused by traditional uniform sampling, an adaptive rectangular partitioning method was introduced for sampling in the uncertain parameters space, and the polynomial coefficients were solved by using regression methods. Finally, the effectiveness of proposed method was verified through numerical example. Analysis results of a full-vehicle air suspension system with 24-dimensional uncertain parameters show that, the reasonable selection of the orders of polynomial expansion and the convergence thresholds can effectively ensure the fitting accuracy and efficiency of the proposed method. The comprehensive performance of the proposed method in the analysis of system dynamic characteristics responses is overall superior to the traditional polynomial chaos expansion method. The uncertainty of some parameters has a significant impact on the dynamic characteristics responses of system, and the key influencing parameters for different responses have certain differences. These should be given special attention in the optimization design.