Structural damage detection is still a challenging problem owing to the difficulty of extracting damage‐sensitive and noise‐robust features from structure response. This article presents a novel damage detection approach to automatically extract features from low‐level sensor data through deep learning. A deep convolutional neural network is designed to learn features and identify damage locations, leading to an excellent localization accuracy on both noise‐free and noisy data set, in contrast to another detector using wavelet packet component energy as the input feature. Visualization of the features learned by hidden layers in the network is implemented to get a physical insight into how the network works. It is found the learned features evolve with the depth from rough filters to the concept of vibration mode, implying the good performance results from its ability to learn essential characteristics behind the data.

No abstract

Summary This paper proposes a data‐driven damage detection method based on fixed moving principal component analysis (FMPCA) to analyze structural dynamic responses and monitor the bridge operational condition and the damage occurrence. The damage indices based on principal components (PCs) and eigenvalues can be calculated continuously by applying a fixed moving window. The length of the moving window is determined by using a new criterion based on the convergent spectrum of cumulative contribution ratio. Numerical simulations and experimental tests in the laboratory on beam bridge models subjected to stochastic loads are conducted to investigate the accuracy and effectiveness of the proposed approach. Both simulation and experimental results indicate that using the FMPCA can well analyze the dynamic vibration data to detect damage or abnormal vibration behavior during the operational condition. It can be used to accurately monitor the time instant of damage occurrence, which is very important in long‐term monitoring of civil engineering structures. The proposed method is successfully applied to analyze the data recorded during an incident that a real large‐scale suspension bridge was slightly scraped by the mast of a sand ship, which further verifies the effectiveness and feasibility of this method in engineering applications. The results also indicate that the bridge was not damaged after the incident but presented a short time abnormal vibration behavior owing to the impact of the ship mast.