Improving Non-Destructive Concrete Strength Tests Using Support Vector Machines

Materials

Yi-Fan Shih²  Yu-Ren Wang²  Kuo-Liang Lin¹  Chin-Wen Chen² 

[1] Faculty of Civil and Ecological Engineering Department, I-Shou University, Kaohsiung 84001, Taiwan;Department of Civil Engineering, National Kaohsiung University of Applied Sciences, Kaohsiung 80778, Taiwan;

关键词: concrete compressive strength;    non-destructive test;    rebound hammer test;    ultrasonic pulse velocity test;    artificial intelligence;    support vector machines;   

DOI  :  10.3390/ma8105368

来源: mdpi

【 摘 要 】

Non-destructive testing (NDT) methods are important alternatives when destructive tests are not feasible to examine the in situ concrete properties without damaging the structure. The rebound hammer test and the ultrasonic pulse velocity test are two popular NDT methods to examine the properties of concrete. The rebound of the hammer depends on the hardness of the test specimen and ultrasonic pulse travelling speed is related to density, uniformity, and homogeneity of the specimen. Both of these two methods have been adopted to estimate the concrete compressive strength. Statistical analysis has been implemented to establish the relationship between hammer rebound values/ultrasonic pulse velocities and concrete compressive strength. However, the estimated results can be unreliable. As a result, this research proposes an Artificial Intelligence model using support vector machines (SVMs) for the estimation. Data from 95 cylinder concrete samples are collected to develop and validate the model. The results show that combined NDT methods (also known as SonReb method) yield better estimations than single NDT methods. The results also show that the SVMs model is more accurate than the statistical regression model.

【 授权许可】

CC BY   
© 2015 by the authors; licensee MDPI, Basel, Switzerland.