Identification Of Cable Force Based On Artificial Neural Network

Accurate identification of cable forces is the key to ensure the reliability of structural stress analysis in the construction control,operation and maintenance stages of cable system bridges.Frequency method is a widely used method of cable force identification.Based on the theory of frequency method,the complex law between cable force and frequency of cable-stayed cable and suspender in practical engineering is discussed and analyzed,and an accurate identification method of cable force is found out in this paper.In this paper,the cable and suspender are taken as the research objects.Artificial neural network is used to study the identification of multi-parameter cable force based on frequency method.For this purpose,the following research is carried out:(1)The existing analytical solutions of cable force identification based on frequency method,practical formulas derived by predecessors and finite element numerical simulation solutions are compared and studied.Based on the relationship between cable forces and natural frequencies in the existing string vibration theory and the differential equation of transverse vibration of cables and suspenders considering bending stiffness,the analytical solutions of cable forces and natural frequencies and the numerical approximate solutions studied by predecessors are given respectively,considering three boundary conditions: hinged at both ends,consolidated at both ends and articulated at one end.The data are obtained by a large number of finite element calculations,which are verified by theoretical analytical and semi-analytical formulas and empirical regression formulas derived from previous theories.Based on numerical analysis and engineering practice,the factors influencing cable force identification based on frequency method are summarized comprehensively.The algorithm of cable force identification based on artificial neural network is established.(2)Based on the numerical simulation data of finite element method,the influence of bending stiffness,boundary condition,length and linear density on the identification of cable force in practical engineering is studied,and the parameters affecting the identification of cable force by frequency method are defined,which lays a theoretical foundation for the parameter analysis by artificial neural network method.(3)The method of artificial neural network is used to identify multi-parameter cable forces.Tens of thousands of simulated data samples are obtained to train the BP neural network for multi-parameter cable force identification,and the predicted results are compared with the calculated results of the existing theoretical formulas,which proves that the method has applicability and high accuracy.The BP neural network is used to study how the selection of multi-parameters affects the effect of the neural network.It is found that when the network is trained with parameters that vary greatly in the range of cable length,the accuracy of cable force prediction is disadvantageous.It is concluded that the selection of length,boundary conditions,linear density and bending stiffness,frequency and cable force will affect the prediction effect of the neural network.