In the product design process, early system design often plays an important role. Although the existing system design methods are very effective, the preference information of different stakeholders and the subjective uncertainty of their existence have not been processed well. Because different stakeholders have different backgrounds and different personalities, the language preferences are expressed differently. Therefore, it is difficult to select a system scheme that meets the performance requirements and is approved by various stakeholders. To solve this problem, this paper proposes a system design method based on heterogeneous language information environment. First, the theoretical system schemes set is constructed through the morphological matrix, and the multi-objective programming model is constructed through the performance indicators to select the system schemes set that optimizes each performance objective and meets the performance constraints. Then, a trapezoidal asymmetric cloud model is used as an intermediate for heterogeneous language information to integrate each stakeholder’s preference information expressed by their favorite linguistic term set to select the best system solution. The effectiveness of the method is verified by a system design example of a horizontal directional drilling machine (HDDM).