Non-Destructive Concrete Strength Prediction Using AI: A Comparative Study of Machine Learning and Deep Learning Models

Accurate prediction of concrete's mechanical properties is a crucial aspect of civil engineering, ensuring the structural integrity and durability of constructions. Traditional destructive testing methods, while reliable, are time-consuming and resource-intensive. This study presents a novel, non-destructive approach for predicting compressive, tensile, and flexural strengths of concrete using only two input parameters: Ultrasonic Pulse Velocity (UPV) and Electrical Resistivity (ER). A comparative analysis was conducted utilizing five machine learning and deep learning models: Support Vector Regression (SVR), K-Nearest Neighbors (KNN), Extreme Gradient Boosting (XGBoost), Multi-Layer Perceptron (MLP), and Convolutional Neural Networks (CNN). The results demonstrated that CNN outperformed all other models, achieving the lowest Root Mean Square Error (RMSE) and Mean Relative Error (MRE) across all three concrete strength predictions. Specifically, CNN achieved an MRE of 1.37% for compressive strength, 1.25% for tensile strength, and 1.76% for flexural strength, highlighting its superior predictive accuracy compared to traditional machine learning models. CNN's strong performance stems from its ability to learn deep, non-linear feature hierarchies from minimal inputs. By capturing complex spatial and functional dependencies between UPV and ER, CNN can model the intricate mechanical behavior of concrete more effectively than shallow models. This makes it particularly suitable for tasks involving highly non-linear physical phenomena, such as predicting strength characteristics from indirect measurements. This research highlights the potential of AI-driven non-destructive testing as an efficient alternative to traditional methods, offering significant advantages in terms of cost reduction, speed, and sustainability in the construction industry.