This paper presents a robust optimization method to decrease the variations in the performance of the designed system caused by the unavoidable manufacturing, installation or measurement errors of the design variables. Generally, it is difficult and costly to determine statistical information with sufficient precision for uncertain design variables; in this study, interval numbers are used to describe the uncertain design variables, and only the bounds of these variables are required. An improved interval truncation method is presented for estimating the variation ranges of the system performances. The robustness estimations of the system performances are incorporated into the optimization formulation to obtain the nominal design variables, which could make the system performances relatively robust; therefore, the design robustness is estimated and improved in the optimization iteration process. The robust optimization method is applied to a general powertrain mounting system (PMS) to improve the design robustness of the PMS decoupling layout and frequency allocation. The optimization results show that the robust optimization method could effectively increase the decoupling ratios in the interested vertical and pitch directions, and the frequency allocation is more robust than that obtained using the traditional deterministic optimization.