Structural deformation prediction of vehicle-mounted radar based on deep learning

Abstract

The vehicle-mounted radar antenna boasts a relatively large size, making its structure susceptible to significant wind forces under strong winds. Additionally, since antennas are typically installed on the rooftops or other elevated positions of vehicles, the large wind resistance area may lead to structural deformation. Therefore, this paper proposes a novel approach. Firstly, this paper utilizes SolidWorks to model the antenna's back frame, beams, and element array, employing high-strength aluminum alloy materials to strike a balance between lightweightness and stiffness requirements. Secondly, this paper imports the model into ANSYS Workbench to simulate the pressure exerted by wind speeds ranging from 5 to 35 meters per second, selecting deformation data from 480 monitoring points (442 on the crossbeam and 38 on the back frame). Finally, this paper constructs a strain-displacement BP neural network model. The experimental findings reveal that: ① The deformation of the antenna intensifies with increasing wind speed, with a 2.5-fold increase at 25 meters per second compared to 5 meters per second; ② The predictive accuracy of the model enhances with wind speed, achieving the smallest error at 25 meters per second (a 70% reduction compared to 5 meters per second), with an overall prediction accuracy exceeding 98%; ③ The locations with larger errors vary across different wind speeds, concentrating in specific areas during high-speed winds.This method addresses the challenge of monitoring antenna deformation in strong enhances the precision of health monitoring for vehicle-mounted radar antenna systems but also offers more accurate data for future structural monitoring and the design of vehicle-mounted antennas.