Structural health monitoring (SHM) integrates advanced sensor networks and machine learning technologies, aiming to automatically extract and identify damage features from sensor data of engineering structures, thus enabling real-time assessment of structural integrity and early diagnosis of potential damage. However, these damage features often include redundant or irrelevant features, which poses challenges for effective feature extraction and damage diagnosis. To solve these problems, a feature selection algorithm based on multi-layer cooperative particle swarm optimizer (MCPSO) is proposed. In MCPSO, the three learning strategies of midpoint sample, random sample and comprehensive sample are skillfully mixed into the particle swarm optimizer, and the hierarchical structure is used to update the population. The damage feature subset is optimized by simulating the search process of multi-layer particle swarm, and the feature set most sensitive to structural damage is identified to improve the accuracy and reliability of damage detection. Taking the multi-damage state monitoring of bolted structure as a verification case, the ultrasonic guided waves signals of bolted structure in different states are collected by lead zirconate titanate sensors. The experimental results show that compared with the machine learning algorithm, MCPSO can select a stable and effective feature subset from the noise data, and realize the identification and quantification of various damage states such as health, crack, loosening and loosening-crack composite damage, which provides a universal method for the technical development and engineering practice in the field of SHM.
12 October 2023 in Multimedia Tools and Applications
Structural health monitoring (SHM) plays a vital role in promptly identifying structural damage in aircraft, optimizing maintenance, and reducing costs. However, it faces significant challenges in practical applications, mainly in that it needs to process a large number of continuously collected sensor data, which are inevitably contaminated by random noise. Therefore, this study maps the relationship between Lamb wave signal data with noise and the health condition of aircraft structures using an end-to-end approach to construct a deep learning framework. The framework integrates deep residual convolutional networks (DRSN) for feature extraction, efficient channel attention (ECA) for feature enhancement, and long short-term memory (LSTM) in analyzing time series data. Lamb wave signal datasets considering different damage locations and severity are obtained by lead zirconate titanate (PZT) sensors on the aircraft structure, and the datasets are destroyed by using multiple levels of Gaussian random noise to approximate the noise disturbances and unavoidable unpredictability of the industrial environment. The experimental data confirms that the performance metrics of the proposed framework for damage presence, localization and quantification tasks are all above 97 % in a noise-free environment. When dealing with high-noise scenes, the framework provides stronger anti-noise robustness and higher accuracy compared to existing state-of-the-art methods. Quantitative analysis of ablation experiments and visualization of the t-distributed stochastic neighbor embedding (t-SNE) algorithm are applied to reveal the contribution of each component of the designed framework towards feature extraction and damage identification of Lamb wave signals in noisy environments.
1 July 2026 in IEEE Transactions on Industrial Informatics