车辆的振动与噪声(NVH)性能正在成为国内外汽车行业竞争焦点之一。子结构频响函数综合方法(FBS)，因为分析频带宽、可应用于模态密集与大阻尼结构等诸多优点，在整车及总成NVH分析与优化中得到越来越多的应用。然而，该类方法在实际工程应用中也存在一些困难，如自由状态下子结构频响函数(FRF)的准确获取、实测FRF中的噪声处理、病态矩阵求逆，以及与界面转角自由度相关FRF的缺失问题等。本文以某微型轮毂驱动电动汽车(简称微车)为研究对象，将其拆分为若干子结构，并尝试解决FBS中的关键技术问题，以建立较为准确的整车NVH分析模型。研究工作可为FBS方法的完善与工程应用提供参考。在模态模型方法中引入振型修正，提出了一种惯性参数识别方法，通过仿真与实验对辨识效果进行了验证。进而将该方法应用于辨识带内饰车身惯性参数，并据此对车身低频带FRF进行了修正，获得了完全自由状态下车身低频带FRF。基于子空间的特征值分解推导了一种实测FRF的降噪方法。对具有噪声的FRF仿真信号进行了子空间降噪处理，证明了算法的有效性。最后将降噪算法应用于微车实测FRF中，取得了较好的降噪效果，以提高FRF矩阵逆运算的精度。考虑到实测子结构FRF中的界面转角自由度信息的缺失问题，提出了界面自由度柔性等效方法。连接界面被分成若干子界面，每个子界面采用刚性等效。板型结构的仿真结果证明了该方法能明显改善FBS精度。进而利用柔性等效方法对微车的前悬架子系统进行了处理，获得了与其子界面有关的转角自由度的FRF。将截断奇异值分解与支持向量机响应面方法相结合，提出了一种有限元模型修正方法。利用傅里叶反变换将FRF变为时域内的脉冲响应函数，并进行相空间重构与截断奇异值分解以表征原FRF，采用支持向量机响应面模型替代有限元模型，利用遗传算法求解模型修正量。通过对某车型扭力梁模型的修正，证明了方法的有效性。进而对微车的后悬架有限元模型进行修正，取得了较好的修正效果。将微车划分为车身和前后悬架等子结构。根据微车结构对称性，分析了各子结构FRF之间的关系，推导了整车FBS综合方程。将经过修正和降噪后的车身子结构的实测FRF，和基于有限元模型计算得到的前后悬架系统FRF，带入整车综合方程，建立了整车FRF分析模型。与实测结果的对比验证了所建模型的可靠性。

Automobile NVH (Noise, Vibration and Harshness) performance is becoming one of the important competitive focus and hotspot for automobile manufacturers across the world. The frequency response function based substructure approach (FBS) has gained more and more momentum for analysis and optimization on the NVH performance of a whole vehicle or its sub-systems, due to its many advantages such as applicable in a wide frequency range, and suitable for structures with high damping or high modal density. However, there are some difficulties occurred to the method during actual engineering applications, mainly including problems as follows: How to obtain the frequency response functions (FRFs) of substructures under a completely free boundary condition? How to eliminate the noise in the measured FRFs? How to solve the inversion of an ill-conditioned matrix of FRFs? How to deal with the absence in the FRFs of rotational degrees of freedom? Taking a micro-vehicle with in-wheel motor drives as a target, this dissertation addresses the aforementioned problems with the FBS methods, and establishes a model for NVH analysis on the full vehicle which is numerically divided into several substructures. The outputs of this study are valuable with improvement and engineering application of the FBS method. A method for inertia properties identification is proposed by incorporating modal shape modification into the modal model method. Numerical simulation and physical experiments are carried out to validate the applicability and efficiency of the proposed method. Inertia properties of the trimmed body of the target vehicle are identified by means of the method, which are then used to update the FRFs of the body in low frequency band. The FRFs of the body are then obtained under a completely free condition.Employing the eigenvalue decomposition (EVD) of the measured FRFs, a method is derived based on signal subspace approach to eliminate noise from measured FRFs. The efficiencies of the methods are demonstrated by reducing noise from a set of simulated noisy FRFs. The method is then used to successfully eliminate noise from the measured trimmed body FRFs, which will improve the accuracy of calculating the inversion of the FRFs matrix. A method of interface flexible equivalence is proposed to deal with information absence of the rotational degrees of freedom of the interface in the measured FRFs. In the method, an interface is divided into several sub-interfaces, each of which is treated as rigid. Simulation on a plate-like structure shows that the proposed flexible equivalence can significantly improve the synthesizing accuracy of the FBS method. In sequence, the method is applied to the interface between the front suspension and body of the target micro-vehicle, yielding corresponding FRFs related to the rotational degrees of freedom in the interface.A method for finite element model updating is put forward by combining the truncated singular value decomposition (SVD) and the support vector machine (SVM) response surface approaches. In the method, the impulse response functions are obtained through inverse Fourier transformation on the measured frequency response functions. By the phase space matrix re-construction and truncated SVD procedure, the features of the original FRFs are reflected. A support vector machine (SVM) response surface is constructed to replace the finite element model. An optimization problem is defined and solved by a genetic algorithm for parameter updating. The effectiveness of the method is validated by model update with a twist beam. Finally, the method is used to update the finite element model of the rear suspension of the target micro-vehicle.The target micro-vehicle is divided into three substructures, namely the trimmed body, the rear suspension and the front suspension. Regarding structural symmetry exists with the micro-vehicle, analysis was performed for the relationship among the FRFs of the related substructures, and the governing equations for FBS approach were derived for quick assembling and modeling. The FRFs of the trimmed body are obtained by physical testing with modification according to measured inertia parameters and noise elimination. Finite element models of front and rear suspensions are respectively established, and the corresponding FRFs of selected points are calculated. Using the FRFs of all the three substructures, a hybrid FBS model is constructed using the governing equation for the micro-vehicle. Physical experiments are carried out to validate the established whole vehicle model by comparing the calculated FRFs with their tested counterparts.