Application of supervised learning for classification of cracking and non-cracking major damage in TRMs based on AE features

Abstract

Textile reinforced mortar composites (TRMs) experience various types of damage. In this study, these damage mechanisms (such as cracking and non-cracking) were understood or distinguished with the help of acoustic emissions (AE). Four types of TRMs made of a single mortar matrix and two types of carbon textiles (coated and noncoated) were evaluated under uniaxial tension. Meanwhile, the acoustic emissions were monitored during the tensile test.The study showed that AE features are sensitive to damage evolution but more importantly that TRMs are complex and various modes can coexist. Similarly, the evolution of improved b-value (Ib) also demonstrated sensitivity to the existence of damage. Finally, the supervised learning models (trained on AE data) demonstrated that information on the instant of cracking obtained through digital image correlation (DIC) can be successfully coupled with AE features to separate the cracking and non-cracking damage types with accuracy of above 80 %.