Many unavoidable uncertainties in the engineering structure will affect its performance response. It is necessary to analyze the uncertain responses generated by uncertain factors in the vehicle design. Therefore, this study proposes a feasible identification method of uncertainty responses for vehicle structures. The proposed method consists of radial basis function neural network (RBFNN) and Taylor interval expansion (IE) model. The optimal network parameters of RBFNN are trained by the improved K-means clustering algorithm and singular value decomposition (SVD). Here, the first-order and second-order differential equations are derived from the RBFNN parameters since it is difficult to calculate the partial derivatives of complex systems without explicit expressions. A series of typical functions are tested and the results show that the trained RBFNN parameters can approximate the first-order and second-order partial derivatives of different types of functions. Besides, the subinterval expansion method is further applied to improve the calculation accuracy of uncertain structural responses. Finally, the proposed method is applied to four engineering applications (vehicle hood, mechanical claw, anti-collision structures, and multi-link suspension). Compared with genetic algorithm (GA) and Monte Carlo simulation (MCS), the proposed method can improve computational efficiency while ensuring accuracy. The identification method of interval uncertainty responses can be used as an alternative for uncertainty analysis and subsequent reliability or robustness optimization.

工程结构中许多不可避免的不确定性会影响其性能响应。需要对车辆设计中的不确定因素产生的不确定响应进行分析。因此，本研究提出了一种可行的车辆结构不确定性响应识别方法。该方法由径向基函数神经网络（RBFNN）和泰勒区间展开（IE）模型组成。RBFNN 的最优网络参数通过改进的 K-means 聚类算法和奇异值分解 (SVD) 进行训练。这里，一阶和二阶微分方程是从 RBFNN 参数导出的，因为没有明确的表达式很难计算复杂系统的偏导数。对一系列典型函数进行了测试，结果表明训练后的RBFNN参数可以逼近不同类型函数的一阶和二阶偏导数。此外，进一步应用子区间展开法来提高不确定结构响应的计算精度。最后，将所提出的方法应用于四个工程应用（汽车引擎盖、机械爪、防撞结构和多连杆悬架）。与遗传算法(GA)和蒙特卡罗模拟(MCS)相比，该方法在保证精度的同时提高了计算效率。区间不确定性响应的识别方法可用作不确定性分析和后续可靠性或稳健性优化的替代方法。