A Framework of Structural Damage Detection for Civil Structures Using Fast Fourier Transform and Deep Convolutional Neural Networks

Deep Learning for Structural Health Monitoring: Data, Algorithms, Applications, Challenges, and Trends

Environmental effects may lead to cracking, stiffness loss, brace damage, and other damages in bridges, frame structures, buildings, etc. Structural Health Monitoring (SHM) technology could prevent catastrophic events by detecting damage early. In recent years, Deep Learning (DL) has developed rapidly and has been applied to SHM to detect, localize, and evaluate diverse damages through efficient feature extraction. This paper analyzes 337 articles through a systematic literature review to investigate the application of DL for SHM in the operation and maintenance phase of facilities from three perspectives: data, DL algorithms, and applications. Firstly, the data types in SHM and the corresponding collection methods are summarized and analyzed. The most common data types are vibration signals and images, accounting for 80% of the literature studied. Secondly, the popular DL algorithm types and application areas are reviewed, of which CNN accounts for 60%. Then, this article carefully analyzes the specific functions of DL application for SHM based on the facility's characteristics. The most scrutinized study focused on cracks, accounting for 30 percent of research papers. Finally, challenges and trends in applying DL for SHM are discussed. Among the trends, the Structural Health Monitoring Digital Twin (SHMDT) model framework is suggested in response to the trend of strong coupling between SHM technology and Digital Twin (DT), which can advance the digitalization, visualization, and intelligent management of SHM.

A Novel CNN-LSTM Hybrid Model for Prediction of Electro-Mechanical Impedance Signal Based Bond Strength Monitoring

The recent application of deep learning for structural health monitoring systems for damage detection has potential for improvised structure performance and maintenance for long term durability, and reliable strength. Advancements in electro-mechanical impedance (EMI) techniques have sparked attention among researchers to develop novel monitoring techniques for structural monitoring and evaluation. This study aims to determine the performance of EMI techniques using a piezo sensor to monitor the development of bond strength in reinforced concrete through a pull-out test. The concrete cylindrical samples with embedded steel bars were prepared, cured for 28 days, and a pull-out test was performed to measure the interfacial bond between them. The piezo coupled signatures were obtained for the PZT patch bonded to the steel bar. The damage qualification is performed through the statistical indices, i.e., root-mean-square deviation (RMSD) and correlation coefficient deviation metric (CCDM), were obtained for different displacements recorded for axial pull. Furthermore, this study utilizes a novel Convolutional Neural Network-Long Short-Term Memory (CNN-LSTM)-based hybrid model, an effective regression model to predict the EMI signatures. These results emphasize the efficiency and potential application of the deep learning-based hybrid model in predicting EMI-based structural signatures. The findings of this study have several implications for structural health diagnosis using a deep learning-based model for monitoring and conservation of building heritage.

A Systematic Review of Structural Health Monitoring Systems to Strengthen Post-Earthquake Assessment Procedures

Structural health monitoring (SHM) is vital to ensuring the integrity of people and structures during earthquakes, especially considering the catastrophic consequences that could be registered in countries within the Pacific ring of fire, such as Ecuador. This work reviews the technologies, architectures, data processing techniques, damage identification techniques, and challenges in state-of-the-art results with SHM system applications. These studies use several data processing techniques such as the wavelet transform, the fast Fourier transform, the Kalman filter, and different technologies such as the Internet of Things (IoT) and machine learning. The results of this review highlight the effectiveness of systems aiming to be cost-effective and wireless, where sensors based on microelectromechanical systems (MEMS) are standard. However, despite the advancement of technology, these face challenges such as optimization of energy resources, computational resources, and complying with the characteristic of real-time processing.

M1M2: Deep-Learning-Based Real-Time Emotion Recognition from Neural Activity

Emotion recognition, or the ability of computers to interpret people's emotional states, is a very active research area with vast applications to improve people's lives. However, most image-based emotion recognition techniques are flawed, as humans can intentionally hide their emotions by changing facial expressions. Consequently, brain signals are being used to detect human emotions with improved accuracy, but most proposed systems demonstrate poor performance as EEG signals are difficult to classify using standard machine learning and deep learning techniques. This paper proposes two convolutional neural network (CNN) models (M1: heavily parameterized CNN model and M2: lightly parameterized CNN model) coupled with elegant feature extraction methods for effective recognition. In this study, the most popular EEG benchmark dataset, the DEAP, is utilized with two of its labels, valence, and arousal, for binary classification. We use Fast Fourier Transformation to extract the frequency domain features, convolutional layers for deep features, and complementary features to represent the dataset. The M1 and M2 CNN models achieve nearly perfect accuracy of 99.89% and 99.22%, respectively, which outperform every previous state-of-the-art model. We empirically demonstrate that the M2 model requires only 2 seconds of EEG signal for 99.22% accuracy, and it can achieve over 96% accuracy with only 125 milliseconds of EEG data for valence classification. Moreover, the proposed M2 model achieves 96.8% accuracy on valence using only 10% of the training dataset, demonstrating our proposed system's effectiveness. Documented implementation codes for every experiment are published for reproducibility.

Structural Assessment under Uncertain Parameters via the Interval Optimization Method Using the Slime Mold Algorithm

Damage detection of civil and mechanical structures based on measured modal parameters using model updating schemes has received increasing attention in recent years. In this study, for uncertainty-oriented damage identification, a non-probabilistic structural damage identification (NSDI) technique based on an optimization algorithm and interval mathematics is proposed. In order to take into account the uncertainty quantification, the elastic modulus is described as unknown-but-bounded interval values and the proposed new scheme determines the upper and lower bounds of the damage index. In this method, the interval bounds can provide supports for structural health diagnosis under uncertain conditions by considering the uncertainties in the variables of optimization algorithm. The model updating scheme is subsequently used to predict the interval-bound of the Elemental Stiffness Parameter (ESP). The slime mold algorithm (SMA) is used as the main algorithm for model updating. In addition, in this study, an enhanced variant of SMA (ESMA) is developed, which removes unchanged variables after a defined number of iterations. The method is implemented on three well-known numerical examples in the domain of structural health monitoring under single damage and multi-damage scenarios with different degrees of uncertainty. The results show that the proposed NSDI methodology has reduced computation time, by at least 30%, in comparison with the probabilistic methods. Furthermore, ESMA has the capability to detect damaged elements with higher certainty and lower computation cost in comparison with the original SMA.