Robust design solutions are necessary for vehicle NVH refinements. Variations in the vehicle-level performance need be cascaded to its subsystems or components. Selection of a control parameter is crucial for reducing influence of the production variations on the vehicle vibration or noise responses. In order to accurately analysis the vehicle NVH performance, this work presents the framework of NVH analysis platform including excitation calculation, vehicle modeling and results post analysis though developed coding. The developed excitation tool evaluates the engine excitation using direct method based on internal data of the engine and the cylinder pressure of measurements. And then the excitation is loaded on the vehicle finite element model for vehicle NVH performance simulation. For root cause of vehicle NVH, mind map analysis was drawn to find out the reasons for the idle vibrations of a vehicle, and a parameter diagram analysis for PTMS was constructed based on experience of vehicle product. A novel robust optimization strategy called SRBF-based Robust Optimization (SRBF-RBO) is developed in this paper. SRBF-RBO applies two-steps for “shifting” performance and “tightening” distribution. Based on the previous work, the proposed method exerts the benefits of SRBF on excellent efficiency and high approximation accuracy around optimum to search a robust design. For estimating the robustness, a novel robustness objective is first proposed in this paper. A realistic vehicle application of PTMS robust design was implemented to validate the capability of SRBF-RBO. The comparison results between baseline and optimum design show that the vibration of steering wheel reduced 41% and the noise at driver’s ear reduce almost 6 dB(A), as well as the dispersion of the performance distribution has also been reduced about 60%. The results indicate the proposed method improves merit of vehicle NVH and solution robustness.