Abstract:A method of missile test fault diagnosis based on the long and short time neural network is proposed for the problems of complex mechanical structure which has variable shape missile, high probability of failure in the process of variable shape and high cost of sensor measurement. First, an actuator fault characteristics database of the variable shape vehicle is constructed based on the aerodynamic parameter model and the nonlinear dynamics model of the variable shape vehicle. Then, an actuator fault diagnosis framework based on the long-short time neural network is proposed for the serialized feature data in the event of a failure of the variable shape vehicle. An ant colony optimization algorithm is used to optimize the hyperparameters of network training to improve the accuracy and generalization of fault diagnosis. It is verified through simulation that the method can achieve low-cost, high-efficiency, and high-accuracy fault rapid localization for variable-profile missiles.