Emil Laftchiev

Machine Learning Methods for Predicting the Field Compressive Strength of Concrete

August 2019

This study evaluates the efficacy of machine learning (ML) methods to predict the compressive strength of field-placed concrete. We employ both field- and laboratory-obtained data to train and test ML models of increasing complexity to determine the best-performing model specific to field-placed concrete.

The ability of ML models trained on laboratory data to predict the compressive strength of field-placed concrete is evaluated and compared to those models trained exclusively on field-acquired data. Results substantiate that the random forest ML model trained on field-acquired data exhibits the best performance for predicting the compressive strength of field-placed concrete; the RMSE, MAE, and R2 values were 730 psi, 530 psi, and 0.51, respectively.

We also show that hybridization of field- and laboratory-acquired data for training ML models is a promising method for reducing common over-prediction issues encountered by laboratory-trained models that are used in isolation to predict the compressive strength of field-placed concrete.