**5. Conclusions**

Currently, it can be difficult or impossible to obtain sufficient specimens by destructive tests according to the ASTM C469 and C39 for the measurement of the Ec and fc required for the design, construction, and maintenance of a structure. Thus, it is necessary to predict them by the non-destructive test (NDT), and one of the most reliable NDT methods is to estimate Ec and fc from the dynamic elastic modulus by ultrasonic pulse velocity or resonance frequency tests. The Ec and fc have usually been predicted from Ed by classical equations or the regression analysis, but these cannot not give an accurate estimation due to the intrinsic effects of each test method and nonlinearity between Ec or fc and Ed. In this regard, this study presented a possibility to improve the estimation of Ec and fc from Ed by a machine learning (ML) algorism based on the results of four types of test on the 282 specimens, comparing with the classical equations such as the Lydon and Balendran equation, Popovics equation, and BS 8110, as well as the regression method. The optimum relationship between Ec or fc and Ed

was derived by using the characteristics of Ed from ultrasonic velocity and resonance frequency test methods associated with four ML algorisms. The accuracy of predicting Ec and fc was compared with the general regression and ML based on four reliable Ed measurements, for the complete range of experimental data. The following conclusions have been drawn:


**Author Contributions:** J.Y.P. and S.-H.S. conceived and designed the experiments; Y.G.Y. and T.K.O. performed the experiments and analyzed the data; J.Y.P. and S.-H.S. contributed device/analysis tools; Y.G.Y. and T.K.O. wrote the paper. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was supported by Incheon National University Research Grant in 2019.

**Conflicts of Interest:** The authors declare no conflict of interest.
