Utilização de machine learning para previsão de propriedades de concreto auto-reparável modificado com bactéria e fibra

Abstract

Self-healing concrete has been studied for decades as an alternative material to overcome problems such as cracking and low durability of conventional concrete. However, laboratory experiments can be costly and time-consuming. The use of predictive computational models that can estimate the properties of concrete are being researched to develop new types of concrete and meet the challenging requirements of the construction industry. In this work, Machine Learning (ML) models were developed using Multiple Linear Regression (MLR), Support Vector Machine (SVM) and Random Forest Regressor (RFG) to predict and analyze the repairing rate of the cracked area and capillary water absorption in self-healing concrete modified with alkali-resistant bacteria, polyvinyl alcohol (PVA) fibers and polypropylene (PP) fibers. The results show that the ML models performed better in predicting the repairing rate of the cracked area and that the SVM and RFG algorithms deliver better adjusted and efficient models. Furthermore, the results of this work show that ML can be a valuable tool for the AI-based study and development of improved concrete.