Production variation analysis of vehicle NVH performance becomes more and more necessary in development process. In order to accurately evaluate vehicle NVH responses, this paper presents a general framework involving excitation evaluation, vehicle FE model and results post-processing. Through mind map and parameter diagram analysis of idling shake, the major cause was found to be piece to piece of its powertrain rubber mounts system. Along with the vehicle NVH simulation, a novel decoupled robust-based optimization strategy called SRBF-RBO is applied to PTMS robust optimization. SRBF-RBO distinguishingly features two-steps searching strategy, the “shifting” performance and “tightening” distribution. Based on the adaptive metamodeling technique of SRBF strategy, the “shifting” operation can quickly search deterministic optimum, and improves the approximation accuracy of metamodels around the optimum in local region. With the developed criterion of robustness evaluation, the purpose of the “tightening” step is to minimize the robustness objective in the local space based on SRBF models. Applied to practical PTMS robust design problem, the proposed methodology demonstrates its capability of searching optimal and robust solution.