Deformation Reconstruction Calibration Of Integrated Antenna Based On Network Strain Sensor

The integrated antenna has the advantages of optimizing the use of the load platform to increase detection performance,excellent fit ability,and maintaining good aerodynamic characteristics of the aircraft.Therefore,the wing integrated conformal antenna has been widely studied and adopted.The real-time high-precision deformation reconstruction of the integrated wing antenna is of great significance to ensure the electrical performance of the antenna and the safety of aircraft navigation.In this paper,the calibration problem of integrated antenna deformation reconstruction based on network strain sensor under small sample data is studied and verified by experiments.The elastic displacement is separated from the multi-section structural displacement by derivation,and a fuzzy self-calibration network is constructed to correct the displacement.The non-uniform rational B-spline(NURBS)function is proposed to enlarge the dimension of small sample data,so as to improve the calibration performance of the network.The reconstruction theory is abstracted as a linear model,and a linear model parameter estimation method based on Bayesian estimation is proposed to realize the correction of the strain coefficient.The reconstruction and calibration algorithms are integrated in a self-developed software system.It provides a basis for calibration experiment verification and concise analysis.First of all,in view of the problem that when the integrated wing antenna is reconstructed in sections,the angle deflection of the front section will affect the displacement of the rear section.By deducing the inverse finite element theory,the displacement is decomposed into two parts: elastic displacement and projection displacement.Elastic displacement and projection displacement are caused by strain and angle,respectively.Then,the elastic displacement error of each segment is distributed to the degrees of freedom through the error distribution algorithm.So as to realize the preliminary calibration of displacement.Secondly,the error distribution algorithm requires actual displacement data and cannot deal with new data.Therefore,a self-calibration network(SCN)is proposed to train the error distribution value and strain.According to the generalization ability of the fuzzy network,the fuzzy relationship is constructed between the strain and the calibration value of the degree of freedom.SCN provides guarantee for real-time calibration.However,the calibration accuracy of SCN is seriously affected by the small sample data volume.Therefore,the NURBS function is proposed to fit small sample data.The sample points are interpolated by changing the function parameters to realize the expansion of sample data.Then,strain sensors have different strain coefficients at different positions on the surface of anisotropic material.Therefore,a Bayes-based coefficient calibration algorithm is proposed.According to the inverse finite element reconstruction theory,a multi-parameter linear model between strain and structural surface displacement is constructed.Then,the posterior distribution of the parameters in the multivariate linear model is derived according to the Bayesian algorithm.The stationary distribution values of the parameters are obtained by Gibbs sampling.The parameters are compared with the model coefficients to calculate the strain calibration coefficients of each sensor.Finally,the construction of the real-time reconstruction and calibration software system is completed.Real-time calibration of displacement is achieved by integrating the SCN algorithm and parameter estimation algorithm in the software platform.Based on the ANSYS modeling of the wing-like three-section truss,the calibrated displacement is processed by the VTK visualization toolkit to realize the three-dimensional deformation display.By comparing the displacement of the target with the measured displacement of the NDI instrument,it is verified that the SCN algorithm and the strain coefficient estimation algorithm based on small samples are effective in improving the reconstruction accuracy.