Observation-informed modeling of artificial neural networks to predict flow and bleeding of cement-based materials

Abstract

Workability of concrete, an essential aspect in construction, determines the efficiency of pumping and placement processes, as well as the strength and durability properties after hardening. The slump test is the most widely used method for evaluating workability, and technicians roughly estimate the workability using their human senses. It may be challenging to unexperienced technicians and vulnerable to mistakes. This study aims to substitute human sensory and slump tests with artificial intelligence. An artificial neural network was adopted to predict both the fluidity and bleeding of the mortars. The observation-informed modeling acquires input of the measurement, the viscosity curve in this study, for the prediction. The resultant network yields a high accuracy for predicting the channel flow and bleeding rate of the mortar samples. This approach can improve the quality and efficiency of construction processes by reducing errors caused by human-sensory based decision.