Research on corrosion damage and load amplitude prediction of bridge steel via MMM detection technology and GA-BPNN

This study aims to address the challenge pertaining to the evaluation of bridge steel plates under corrosive conditions by integrating metal magnetic memory (MMM) detection technology with a genetic algorithm-optimized backpropagation neural network (GA-BPNN) to predict quantitatively the corrosion damage and load amplitudes of bridge steel plates. Bridge steel plates with different corrosion defects were obtained using electrochemically accelerated corrosion tests, and MMM signal detection was conducted under uniaxial tensile conditions to study the influences of corrosion defects and tensile loads on MMM signals. A stress-related force-magnetic coupling magnetic charge model considering end effects under weak magnetic conditions was proposed to analyze the effects of corrosion depth, width, and stress on MMM signals. Finally, by constructing a GA-BPNN model and integrating nine types of magnetic signal characteristic parameters, quantitative predictions of the corrosion defects and load amplitude were performed. The results indicate that the constructed model can accurately predict the corrosion depth, width, and load amplitude, with R² values of 0.911, 0.980, and 0.903, respectively, for the test set. This study offers a novel solution for bridge health monitoring and quantitative evaluation of corrosion damage.