Fatigue damage detection from imbalanced inspection data of Lamb wave
Jingjing He,Ziwei Fang,Jie Liu,Fei Gao,Jing Lin
DOI: https://doi.org/10.1177/14759217211015243
2021-05-24
Structural Health Monitoring
Abstract:The core of structural health monitoring is to provide a real-time monitoring, inspection, and damage detection of structures. As one of the most promising technology to structural health monitoring, the Lamb wave method has attracted interest because it is sensitive to small-scale damage with a long detection range. However, in many real-world structural health monitoring applications, the nature of the problem implies structures work under normal condition in most of its operating phases; therefore, classes of data collected are not equally represented. The predictive capability of damage detection algorithms may significantly be impaired by class imbalance. This article presents a damage detection method using imbalanced inspection data which is collected through Lamb wave detection. Aiming at maximizing detection accuracy, an improved synthetic minority over-sampling technique using three-point triangle (triangle synthetic minority over-sampling technique) is proposed to conduct the over-sampling procedure and increase the number of minority samples. The iterative-partitioning filter is employed to remove the noisy examples which may be introduced by triangle synthetic minority over-sampling technique. Three conventional classification methods, namely, support vector machine, decision tree, and k-nearest neighbor, are used to perform the damage detection. A fatigue crack detection test using Lamb wave is performed to demonstrate the overall procedure of the proposed method. Three damage sensitive features, namely, normalized amplitude, correlation coefficient, and normalized energy, are extracted from signals as datasets. A cross-validation is performed to verify the performance of the proposed method for crack size identification.