The Use of Artificial Intelligence in High Strength Hybrid Fiber Self Compacting Concrete—An Approach to Function Approximation of Flexural Strength

Abstract

This paper focuses on artificial intelligence methods employed to generate a function for flexural strength of High Strength Hybrid Fiber Self Compacted Concrete which is one of the special concretes that tackles with the workability and durability without affecting the strength of concrete. Its outstanding deformability in the fresh state not only leads to high resistance to segregation resulting in better homogeneity and enhancement but also improves productivity by decreasing the time of construction. Quartz powder, glass fibers, steel fibers, carbon, and synthetic fibers when incorporated in plain concrete results in improvement of the toughness, post cracking resistance, ductility, strength, etc. Additionally, it also controls the detrimental effect of shrinkage. The current study includes the Development of HSHFSCC (High Strength Hybrid Fiber Self Compacting Concrete) and Employment of various machine learning models compared to categorize the collected data into test, train, and validation followed by 19 kinds of various machine learning regression models along with artificial neural network for advancing a function to approximate the flexural strength of HSHFSCC. Here, three parameters were used as input which includes the setting time, percentage of quartz sand, and the percentage of river sand. A total number of 25 datasets were utilized with fivefold cross-validation using MATLAB Machine Learning and Deep Learning toolkit and the optimized performing models were then validated in python. The assessment factors including R-square and root mean square suggests a level of reliability and accuracy of the model.