Electronic Thesis and Dissertation Repository

Thesis Format

Integrated Article


Master of Engineering Science


Civil and Environmental Engineering


Moncef L. Nehdi


The use of steel fibers for concrete reinforcement has been growing in recent years owing to the improved shear strength and post-cracking toughness imparted by fiber inclusion. Yet, there is still lack of design provisions for steel fiber-reinforced concrete (SFRC) in building codes. This is mainly due to the complex shear transfer mechanism in SFRC. Existing empirical equations for SFRC shear strength have been developed with relatively limited data examples, making their accuracy restricted to specific ranges. To overcome this drawback, the present study suggests novel machine learning models based on artificial neural network (ANN) and genetic programming (GP) to predict the shear strength of SFRC beams with great accuracy. Different statistical metrics were employed to assess the reliability of the proposed models. The suggested models have been benchmarked against various soft-computing models and existing empirical equations. Sensitivity analysis has also been conducted to identify the most influential parameters to the SFRC shear strength.

Summary for Lay Audience

Shear failure of reinforced concrete (RC) beams has been a concern due to its brittle and sudden nature. The use of conventional stirrups to increase the shear capacity of RC beams has been effective in avoiding catastrophic failures. However, using conventional stirrups is laborious, costly, and can be challenging when the structure has a thin or irregular cross-section. The use of steel fiber-reinforced concrete has gained great momentum in recent years owing to the noteworthy increase in the shear capacity and possible replacement of minimum stirrups with steel fibers. Yet, the use of steel fibers remains limited due to the lack of design provisions in building codes. This is mainly associated with the complex shear transfer mechanism and random orientation of fibers inside the concrete matrix. Exiting empirical equations for SFRC beams shear capacity have been developed with a relatively few data samples, which makes their accuracy over new samples that fall outside their range of validity uncertain.

To overcome such drawbacks, the research presented herein suggests two machine learning models developed from an extensive database to predict the shear capacity with high accuracy. The first hybrid model is a combination of artificial neural network (ANN) and atom search optimization (ASO). The second model, which is based on genetic programming (GP), is an alternative approach that aims to generate a transparent equation for estimating the SFRC shear strength. Appropriate tuning of the hyperparameters for each model has been conducted to achieve optimal accuracy. The performance of the suggested models was assessed via several statistical metrics. The accuracy of the models was also compared to that of other widely used machine learning models and empirical equations. Results reflected the superior accuracy of the proposed models in terms of correlation and error between predicted and actual values. In addition, sensitivity analyses were performed to identify the most important parameters affecting the shear strength. It was found that for both models, the shear span-to-depth ratio had the greatest influence on the shear capacity of SFRC beams without stirrups.