Shape reconstruction of aerial refueling hose based on distributed fiber optic sensors and FSResNet-MLP

In the refueling process, the aerial refueling hose is affected by complex aerodynamic factors, which is prone to whipping phenomenon, which affects docking efficiency and flight safety. Therefore, a frequency shift feature extraction method of distributed fiber optic sensor based on frequency shift residual neural network is proposed (FSResNet) to solve the problem of nonlinear strain response law which is difficult to characterize due to the superelasticity of rubber on the outer layer of refueling hose. On this basis, the physical information of the hose curvature is introduced into the multi-layer perceptron (MLP) neural network, and the neural network is trained with small sample data to realize the adaptive reconstruction of the hose shape, which solves the large error of the shape reconstruction caused by the instantaneous reverse phenomenon of the main normal vector at the inflection point of the Frenet-Serret frame. An experimental system for refueling hose shape reconstruction based on distributed fiber optic sensor with high spatial resolution is constructed. The results show that the average relative error of the two-dimensional shape reconstruction of the hose is 8.7%. The average relative error of Y-axis coordinate reconstruction and Z-axis coordinate reconstruction is 9.27% and 9.39% respectively. The research results can not only help to ensure the safety of aerial refueling, but also provide key technical support for the establishment of "new generation of digital twin system for aerial refueling hose".