Prediction of compressive strength of granite: use of machine learning techniques and intelligent system

The accurate determination of uniaxial compressive strength (UCS) plays a vital role in the initial design phase of rock engineering and rock geotechnics. Traditionally, this assessment entails costly, time-intensive and labor-demanding experimental tests. Consequently, there is significant promise in exploring machine learning techniques for UCS prediction, warranting further investigation. This study aims to introduce an innovative machine-learning approach and an intelligent system for forecasting UCS based on various granite rock datasets. To achieve this, a novel hybrid model is proposed by combining Marine Predators Algorithm (MPA) and artificial neural network (ANN), and then resulting in an intelligence system. Additionally, forty-nine empirical formulas, including fourteen developed in this study and thirty-five from prior literature, are considered. The input variables for the model comprise the Point load strength index (Is(50)), Schmidt hammer rebounded number (RL) and P wave velocity (Vp), while the UCS serves as the output variables. The obtained results show that the MPA-ANN model exhibits superior performance compared to other prediction models. Furthermore, a user-friendly intelligence system is developed using MATLAB programming. This research stands as a compelling demonstration of the efficacy of a combined supervised learning approach and swarm intelligence algorithms in addressing engineering challenges, such as UCS prediction. It has the potential to offer valuable support for practical applications in the field and further explorations in the domain of rock mechanics studies.