Investigation on the data augmentation using machine learning algorithms in structural health monitoring information

Abstract

Structural health monitoring system plays a vital role in smart management of civil engineering. A lot of efforts have been motivated to improve data quality through mean, median values, or simple interpolation methods, which are low-precision and not fully reflected field conditions due to the neglect of strong spatio-temporal correlations borne by monitoring datasets and the thoughtless for various forms of abnormal conditions. Along this line, this article proposed an integrated framework for data augmentation in structural health monitoring system using machine learning algorithms. As a case study, the monitoring data obtained from structural health monitoring system in the Nanjing Yangtze River Tunnel are selected to make experience. First, the original data are reconstructed based on an improved non-negative matrix factorization model to detect abnormal conditions occurred in different cases. Subsequently, multiple supervised learning methods are introduced to process the abnormal conditions detected by non-negative matrix factorization. The effectiveness of multiple supervised learning methods at different missing ratios is discussed to improve its university. The experimental results indicate that non-negative matrix factorization can recognize different abnormal situations simultaneously. The supervised learning algorithms expressed good effects to impute datasets under different missing rates. Therefore, the presented framework is applied to this case for data augmentation, which is crucial for further analysis and provides an important reference for similar projects.

Machine learning and structural health monitoring overview with emerging technology and high-dimensional data source highlights

“…Other methods such as k-nearest neighbors (kNN), stochastic regression, extrapolation and interpolation, and many others can be employed to correct or remove irrelevant data. Recently, Tan et al 84 investigated the effectiveness of multiple supervised learning methods (Ridge, RF, SVR, MLP, and XGBoost) for data augmentation under different missing rates of the inputted database. All models were able to capture the missing trend when the missing data are uniformly distributed, and SVR and MLP performed best on average with root mean square error (RMSE) of less than 2.…”

Non-parametric empirical machine learning for short-term and long-term structural health monitoring

“…Depending on the type of training data and the availability of damaged information, a machine learning algorithm can be categorized as supervised, semisupervised, and unsupervised learning classes. 7,8,[11][12][13] For complex and full-scale civil engineering structures, it may not be practical to impose intentional damage patterns in an effort to prepare full-labeled training data including undamaged and damaged information. Therefore, it seems that unsupervised learning is more suitable than the other machine learning algorithms for SHM of civil structures, particularly for early damage detection.…”

From Cognitive Bias Toward Advanced Computational Intelligence for Smart Infrastructure Monitoring

“…In the same direction, to overcome the aforesaid decision errors or biases, advanced computational intelligence techniques are gaining increasing attention for developing the SHM strategies. For instance, data mining techniques (Gordan et al, 2017(Gordan et al, , 2018(Gordan et al, , 2021cTan et al, 2021), cloud computing (Abdulkarem et al, 2020), and deep learning (Wang et al, 2021) have recently been used in SHM. Both AI and machine learning are also considered as emerging technologies in the 2020s.…”