An optimization neural network model for bridge cable force identification

Accurate determination of cable force values is the most important technical means to avoid damage to the cable bridge. In order to avoid the influence of the difficulty in distinguishing the boundary conditions and the lack of low-order natural frequency on the cable force determination results, an intelligent method for determining the bridge cable force based on the vibration method is proposed. With the cable length, linear density, flexural stiffness and input frequency as input units and the cable force as output unit, a neural network is established to identify the cable force by combining the finite element simulation data, and the model is optimized using the intelligent swarm optimization algorithm. The results show that compared with the cable force prediction models using generalized regression neural network (GRNN) and GRNN optimized using particle swarm optimization (PSO-GRNN) and canonical identification methods, the GRNN optimized using sparrow search algorithm (SSA-GRNN) proposed in this paper has a better identification effect. The prediction error of short cables is essentially within 10%, and that of long cables is within 5%. It can not only realize the accurate identification of bridge cable force by ignoring the boundary conditions and vibration frequency order of cables, but also has a wide range of applications.

Automated summary

Accurate determination of cable force values is crucial for preventing damage to cable bridges. This paper proposes an intelligent method for determining bridge cable force based on the vibration method, which overcomes the challenges of distinguishing boundary conditions and low-order natural frequency. By using cable length, linear density, flexural stiffness, and input frequency as inputs and cable force as the output, a neural network is established to identify the cable force and optimize the model using an intelligent swarm optimization algorithm. Results show that the proposed GRNN optimized by sparrow search algorithm achieves better identification performance compared to other prediction models, with prediction errors within 10% for short cables and within 5% for long cables. This method allows for accurate identification of cable force, disregarding boundary conditions and vibration frequency, and has wide-ranging applications.