Abstract: In power infrastructure projects, the quality of reinforced concrete is related to the safety and durability of the project, and the actual effective diameter of steel bars will be reduced due to factors such as corrosion, which directly affects the seismic and compressive properties of reinforced concrete. A method for detecting the diameter of concrete steel bars based on OPTINET-BP was proposed. Firstly, for concrete specimens with different steel bar diameters, the voltage response of the eddy current detection coil under 1 kHz pulse excitation was collected, and Savitzky-Golay filtering was performed to eliminate noise. Secondly, the exponential function fitting of the detection coil voltage was carried out to extract the characteristic quantity of the voltage signal. Finally, the voltage signal characteristics were preprocessed to obtain data samples, and BP network and OPTINET-BP network were used for training to predict the diameter of steel bars. By comparing the experimental results of reinforced concrete samples with different specifications, it was found that the mean square error of the improved model was 0.990 18, which was about 65 % lower than that of the traditional BP model. The experimental results showed that the improved model could significantly improve the accuracy of steel bar diameter detection and had high practical value.