Structural Damage Identification Based on Multi-Head Convolutional Neural Network and Improved Beluga Whale Optimization Considering Ambient Temperature

Accurately structural damage identification remains a significant challenge due to dynamic response fluctuations caused by temperature variations. To address this problem, a novel two-stage damage identification method based on Multi-head Convolutional Neural Network (MCNN) and Improved Beluga Whale Optimization Algorithm (IBWO) is proposed to precisely predict temperatures, localize the damage, and quantify damage severity. First, the MCNN is exploited to forecast the ambient temperature and detect the number of damaged locations. The predicted values are then fed into an iterative optimization process of the IBWO to identify the damage location and severity. Finally, numerical models of simply supported beams and a 3-storey bookshelf structure are employed to verify the effectiveness and robustness of this method under measurement noises. Results demonstrate that the two-stage diagnosis method can accurately identify both single and multiple damages under varying ambient temperatures. The comparison study with two other state-of-the-art methods also demonstrates the superior performance of the two-stage method in damage localization and quantification.