Abstract
Reinforced concrete is susceptible to erosion from natural disasters such as earthquakes, floods, and strong winds, as well as its own aging and environmental factors. If minor structural damage is not detected in a timely manner, continuous development may lead to damage to structural components, further causing sudden failure of overall structural stability, resulting in major accidents and causing significant losses to life and property. Therefore, identifying the health status of reinforced concrete structures and promptly identifying initial minor damages has important engineering significance. Intelligent optimization algorithms are a type of efficient optimization algorithm, and using intelligent optimization algorithms to optimize hyperparameters of neural network models has become one of the research hotspots. Cuckoo Search Algorithm (CSA), as a new type of intelligent optimization algorithm, has the characteristics of simple implementation, few parameters, strong optimization ability, and good performance in complex optimization problems. It has the potential to be applied to hyperparameter optimization of neural network models. Therefore, in order to improve the performance of damage identification in reinforced concrete structures, this paper constructs a model based on Improved Cuckoo Search Algorithm (ICSA) to optimize Gaussian Weighted Extreme Learning Machine (GWELM). By leveraging the powerful feature extraction capability of Convolutional Neural Network (CNN), a CNN integrated CNN-ICSA-GWELM model is proposed and applied to damage identification of reinforced concrete structures. The experimental results showed that CNN-ICSA-GWELM achieved an accuracy of 99.81%, a precision of 99.74%, a recall rate of 99.53%, and an F1 score of 0.9960 for identifying damage test sets in reinforced concrete structures. Therefore, the performance of CNN-ICSA-GWELM in identifying damage to reinforced concrete structures is superior to CNN-CSA-GWELM, CNN-GWELM, ELM, and Support Vector Machine (SVM).