Hybrid electric vehicles often experiences key-on/off conditions, to increase vibration isolation performance of the Powertrain Mounting System (PMS) under key-on/off conditions is one of the important measures to enhance NVH (Noise , Vibration and Harshness) performance of the vehicle. A model with 13 DOFs for PMS was established , and acceleration of the mounts at active side was estimated under the key-on/off condition. Also vibration acceleration of the mounts and the seat rail were measured. The comparison between calculations and experiments validated the proposed model. Considering or not considering the uncertainty of mount stiffness and locations, two optimization methods for reducing the vibration dose value of longitudinal acceleration of the powertrain were proposed. When the mount parameters were interval uncertain, the upper and lower bounds of the vibration response were calculated using the perturbation center difference method and then used for uncertainty optimization. The experimental results show that if mount parameters are optimized using deterministic methods, the vibration performance of the PMS is the best but the robustness is insufficient, while the vibration isolation performances and robustness of the PMS are compromised using uncertainty optimization methods.