Determining the shear strength of geomaterials is critical in stability analyses for geotechnical design, which can also be costly and time-consuming if traditional laboratory methods are used to determine the parameters experimentally. This can become even more challenging when assessing waste/marginal materials such as coal wash, steel furnace slag, and rubber due to their variability and nonlinear nature of properties. To address this, this study uses two nonlinear machine learning techniques, namely artificial neural network and multivariable regression, to predict the peak friction angle ( \({\phi }^{\prime}_{peak}\) ) of several rubber-incorporated granular waste mixtures, whereas a bulk proportion of prior research was focused on just a single material or mixture type. To keep the model relatively simple for more straightforward implementation in practice, five key parameters are considered including the type of mixture (rubber content; median particle size), its basic physical properties (initial void ratio; dry unit weight), and the applied stress state (confining pressure). In this respect, this study innovatively captures the underlying geotechnical principles that govern the frictional resistance of granular media to efficiently predict the shear strength of a variety of material and mixture types, as opposed to just one. This study reveals that an ANN model trained with Bayesian regularisation (ANN–BR) with 7 hidden nodes is the best to predict \({\phi }^{\prime}_{peak}\) , achieving a coefficient of determination of 0.94. The current analysis proves that ANN–BR can be used as a powerful and reliable model to predict the shear strength of various granular materials, with and without waste rubber inclusion.
