A study on the application of an improved adaptive neural network in prestressed bridge engineering inspection

In recent years, there have been mixed evaluations of the performance of pre-stressed bridges in society. Based on this, this study proposes to integrate adaptive neural networks with BP networks to build a bridge tolerance detection model and combines support vector machines and radial basis function networks to build a bridge wind vibration detection model. The results showed that in the detection results of angle adjustment and detachment, Sample 1 was the closest to the true value, with a difference of only 0.01. As the number of samples increased, the difference became larger, and the difference in sample 5 reached its maximum value of 0.3. The turbulence level of 0.5% had the lowest initial vibration wind speed at a wind attack angle of 10°, with a maximum value of 21m/s. This indicates that the proposed combination model should be more accurate in detecting the tolerance of bridges and more timely in detecting wind-induced vibration risks. In general, research methods have a significant technical value for the safety maintenance of bridge engineering.