The establishment of accurate and intelligent real-time onboard diagnostic models is essential for improving the reliability and safety of weapon system.Aiming at the limitations of the existing diagnostic model development methods, a new method for developing the onboard diagnostic models is proposed.By use of the graphical modelling approach and based on the data of the system or onboard equipment, the implementation procedures for the model development are introduced, including data preparation, system model extraction, fault characteristic matrix construction, and diagnosis reasoning.Taking the model development process and diagnosis results of a certain type of fuel tank as an example, it is verified that the real-time onboard diagnostic model based on system data analysis has high detection rate, high isolation rate and high accuracy.