PREDICTION OF CONCRETE COMPRESSIVE STRENGTH USING SUPERVISED MACHINE LEARNING MODELS THROUGH ULTRASONIC PULSE VELOCITY AND MIX PARAMETERS

As a non-destructive technique for concrete compressive strength assessment for existing concrete structures, Ultrasonic Pulse Velocity (UPV) test method has been widely used. Since the UPV affected by many factors, it is not easy to accurately assess the concrete compressive strength. Effect of some factors which are coarse aggregate grading type, slump, the water-cement ratio (w/c), sand volume ratio, coarse aggregate volume ratio, testing age, concrete density, and pressure of steam curing, were analyzed on the relationship between ultrasonic pulse velocity and concrete strength. 436 records of data, extracted from published research work, were used to build seven supervised machine learning regression models which are; Artificial Neural Network (ANN), Support Vector Machine (SVM), Chi-squared Automatic Interaction Detector (CHAID) decision tree, Classification and Regression Trees (CART) decision tree, non-linear regression, linear regression, and stepwise linear regression models. Also, the independent variable importance for each predictor was analyzed and for each model. With an adequate tuning of parameters, ANN models have produced the highest accuracy in prediction, followed in sequent with SVM, CHAID, CART, non-linear regression. Linear and stepwise linear regression models have present low values of predictive accuracy. w/c was observed to be the highest importance factor in prediction of concrete strength, and the forecasting of the concrete strength was efficient when using w/c and UPV only as predictors in any of the used predictive models.