Estimation of Concrete Mechanical Properties by Acoustic Emission with Random Forest Algorithm

Abstract

The concrete structures are damaged by environmental factors and external loads. To maintain such ages structures, it is necessary to properly determine the mechanical properties and accumulated damage level suffered by concrete using core tests. In this study, relationship between the stress level of damaged concrete under uniaxial compression test and acoustic emission (AE) parameters was investigated using clustering and regression analysis with random forests to identify the most important AE parameters. For clusters 1 and 3, R2 is higher, and RMSE and MAE are lower than for non-clustered cases. Therefore, cluster analysis can be expected to improve the accuracy of AE testing. Thus, the most important parameter is determined to be rise time, the secondly is detected the centroid frequency. These two parameters can be used to clarify compressive fracture behavior of damaged concrete.