Fast Optimization of the Antenna Installation Position on the Train’s Roof

In high-speed railway system, the pantograph arcing can easily affect the normal operation of the positioning system through satellite navigation antenna coupling. A predictive model utilizing Multi Modular Neural Network (MMNN) is developed in this study to decrease the coupling coefficient between pantograph and satellite navigation antennas, allowing for efficient and accurate determination the coupling coefficients corresponding to different antenna installation positions on the train’s roof. The Radial Basis Function (RBF) and Generalized Regression Neural Network (GRNN) are selected as sub neural networks for modeling and prediction. Moreover, with the cooperation integration strategy, the overall prediction accuracy of the MMNN prediction model is improved. The accuracy and rapidity of the MMNN model are also verified through the comparative analysis. Finally, the installation position with a smaller coupling coefficient can be found through comparative analysis. That is to achieve the fast optimization of antenna installation position on the train’s roof.