Load Prediction of Space Deployable Structure Based on FBG and LSTM

To predict the dynamic load of space deployable structure, a novel prediction method based on fiber Bragg grating (FBG) sensors and long short-term memory (LSTM) neural network is proposed. The FBG sensors can measure strain values of the structure. The LSTM establishes the relationship between the strain values and load without a complex dynamic model. The LSTM first works in training mode, its input is strain values and output is the load. Then, the LSTM works in the prediction mode, which takes the current strain values as input and uses the trained network model to predict load. The network of LSTM is simple, which can achieve good prediction effect and shorten the running time. To evaluate the effectiveness of the proposed method, numerical simulations and experiments of space deployable structure with single-point input load are employed. The results demonstrate that compared with the method based on traditional neural network, the max error, root mean square error, and consuming the time of the proposed method are reduced by more than 43%, 21%, and 67%, respectively.