Flight Parameter Prediction and Fault Propagation based on Machine Learning and Symbolic Directed Graph

Abstract

Fault influence and propagation analysis of flight vehicle systems is an important element of flight vehicle health management and also an important problem to be solved. A fault influence model based on data-driven is established, including a prediction model of flight parameters under fault, a dynamic influence path, and an influence degree model. Based on the historical experimental data, a long and short-term memory neural network (LSTM) model is proposed to predict the time-series data of each flight parameter of the flight vehicle under fault; based on the prediction results, a symbolic directed graph (SDG) is used to describe the fault of the flight vehicle system, and then introduce the concept of a compatible path with time-series characteristics to describe the dynamic propagation process of the fault. The case shows that the method proposed in this paper enables qualitative and quantitative analysis of the fault influence, and can reasonably describe the fault propagation path and influence characteristics.