Assessment of damage in hydraulic concrete by gray wolf optimization‐support vector machine model and hierarchical clustering analysis of acoustic emission

Abstract

Summary Acoustic emission (AE) is a useful method for recording fracture processes in concrete. In this work, AE data are recorded during three‐point bending tests to fracture of hydraulic concrete. First, AE data are used to analyze concrete's damage development using hits distribution, b‐value, Ib‐value, and average frequency versus RA value. Second, clustering analysis of AE signals is performed by hierarchical clustering. Third, a support vector machine model based on the gray wolf optimization algorithm is proposed to quantify the degree of damage. Via b‐value analysis it is shown that the fracture process of hydraulic concrete can be divided into three stages: microcracks nucleation; microcracks coalescence into macrocracks (macrocrack nucleation); and macrocrack propagation. Further, it is shown that rise time, ringdown counts, energy, duration, amplitude, and central frequency can be used to characterize the failure modes. Specifically, it is found that microcrack nucleation stage is dominated by tensile failures, macrocrack nucleation stage is characterized by rapid increase of shear failures, which become dominant over tensile failures, and macrocrack propagation stage is dominated by shear failures. Via hierarchical cluster analysis, it is found that the fracture process can be divided into three clusters, which corresponds to the three stages obtained via b‐value analysis. Finally, the proposed support vector machine model based on gray wolf optimization is found to predict the degree of damage in excellent agreement with experiment. This offers an effective practical method for damage assessment by combining AE with machine learning.

Self‐adaptive real‐time clustering analysis and damage pattern recognition for steel tube confined reinforced concrete structures through acoustic emission signals

“…In practical application, Equation ( 2) is vulnerable to sample mean-value, by eliminating the mean-value, L r can be calculated by Equation (3).…”

“…2 A common target of AE monitoring and inspection is to distinguish between the sources of different origin and to get a deeper insight into the interrelation between the underlying processes such as plastic deformation, crack initiation, and cracking. 3 During the AE monitoring process, large quantities of heterogeneous data can be acquired; real-time insights and accurate abnormal diagnosis and recognition are conducive to reduce maintenance cost, increase operating safety, and enhance structural reliability. 4 Clustering is such one of the functionality of data mining, which can extract intrinsic and hidden patterns or characteristics from very large datasets.…”

“…Acoustic emission (AE) is an advanced intelligent sensing technique which has been extensively applied in civil engineering for nondestructive SHM, 1 damage diagnosis, and recognition 2 . A common target of AE monitoring and inspection is to distinguish between the sources of different origin and to get a deeper insight into the interrelation between the underlying processes such as plastic deformation, crack initiation, and cracking 3 . During the AE monitoring process, large quantities of heterogeneous data can be acquired; real‐time insights and accurate abnormal diagnosis and recognition are conducive to reduce maintenance cost, increase operating safety, and enhance structural reliability 4 …”

Damage evolution analysis of concrete based on multi-feature acoustic emission and Gaussian mixture model clustering

“…Das et al (2019) identified the crack modes in reinforced concrete beams using a machine learning-based method, with RA and AF as input features for clustering unlabeled AE data. Li et al (2022) predicted the damage evolution of hydraulic concrete based on the gray wolf optimization (GWO) algorithm, selecting six AE parameters as input features for clustering. It is worth noting that AE parameters collected in experiments often contain redundant and irrelevant features, which can result in discrepancies between the real damage behaviors and AE responses (Li et al, 2014).…”

“…Das et al (2019) identified the crack modes in reinforced concrete beams using a machine learning-based method, with RA and AF as input features for clustering unlabeled AE data. Li et al. (2022) predicted the damage evolution of hydraulic concrete based on the gray wolf optimization (GWO) algorithm, selecting six AE parameters as input features for clustering.…”