Estimation of Durability Benchmark on Concrete Samples Using Artificial Intelligence

Abstract

This paper proposes a deep learning model for predicting the durability benchmark on concrete specimens. The durability benchmark on concrete samples is commonly estimated throughout the Ultrasonic Pulse Velocity measurements. This test establishes a relationship with concrete durability taken into consideration the material's homogeneity. The model proposed in this paper is feed by standard laboratory tests as input parameters, making the model a practical and efficient alternative to predict durability concrete benchmark, saving time, short-cut laboratory work, and avoiding sophisticated instrumentation use. Furthermore, it is an attractive alternative to the need for sophisticated instrumentation for estimating the Ultrasonic Pulse Velocity. The outcomes depict a high predictive accuracy about of 96% in the validation stage. In addition, the model was tested by a new dataset with different properties to demonstrate robustness and certainty in the model. Finally, the model achieves an impressive accuracy of 95.89% in the new validation dataset.