Abstract: The modern weapon systems become more and more complex with the performance improvement. High integration and miniaturization have increased the risk of the potential fault, while the number of possible test points is reduced, which reduces the capability of test and diagnosis and affects the system security. Therefore, testability modeling and analysis technology are more and more emphasized by product designers. A system with adequate testability can diagnose and isolate faults promptly, rapidly, and accurately, with improved reliability and safety for specific tasks. In order to apply the testability technology to the development of weapons, an approach is proposed in this paper to analyze and evaluate the system testability based on the multi-signal model, as a hierarchical model. On the basis of the system structure, a hierarchical directed graph is used to indicate the direction of signals, the component of individual unit and the mutual relationship. The characteristics of the system can be represented by defining its function, the pattern of fault, and by testing the relationship between signals. Muti-signal modeling does not need detailed information about the failure mode, so it can be used to diagnose and isolate unexpected failure mode. It does not need detailed information about the design of the system, so it can be used to model even a conceptional product. In this paper, a multi-signal flow graph model of fuze is set up and the system testability is evaluated quantitatively and some effective strategies for diagnosis are proposed, by use of the fault detection rate, fault isolation rate, fault isolation of ambiguity group, unidentified fault and unused test. Meanwhile, testability suggestions such as the location of the test point, the location of isolation point to eliminate the feedback loop can be tagged directly in the functional module. It also provides accurate graph or text format testability reports. Simulation results indicate that a fast and accurate fault detection and isolation can be achieved by using this method, with the following features: easy modeling, low distortion and high generality. Its application can improve reliability and save maintenance cost of products.