针对故障注入实验所需故障样本容量大、样本区分性不高等问题,建立了一种基于符号有向图的复杂系统故障样本选取方法,该方法通过对系统建立符号有向图(Signed Directed Graph,SDG)模型并引入相容通路概念完备描述了故障的传播过程,并采用去毁度评价计算模型节点重要性,提出了一种考虑节点重要性及故障传播的故障样本选取方法.案例分析表明,该方法选择的样本数量以及需求的数据信息更少,提高了样本完备性且更能覆盖整个系统.“,”In this study, a theoretical framework for fault sample selection based on signed directed graphs is developed and applied to a stable tracking platform. Compared with other selection methods, the proposed one can overcome the problems such as requirements of massive fault samples and low distinction of the samples. First, the SDG mode is established. Second, compatible pathways are introduced, which can describe the fault propagation completely, and the importance of the model node is calculated by destroying degrees. Next, the fault samples selection method is designed with consideration of the importance of components and fault propagation. Finally, simulation results are given to show that the proposed method can take less fault samples and data information, the sample completeness is improved tremendously and the system is covered entirely.