A New Model for Predicting the Hardness of Carbonate Mudrocks Through Elemental Compositions Employing Artificial Intelligence Techniques

The expansion of unconventional resource exploration emphasizes understanding source rock geomechanical properties for better development of these resources. Rock hardness, a critical factor, indicates compressive strength and influences various properties like Young’s modulus. It is pivotal in drilling, aiding in estimating bit wear and drilling speed. Additionally, rock hardness is crucial in engineering projects such as dams, tunnels, and slope stability assessments. In this study, new artificial intelligence models were developed to predict the rock hardness based on the rock composition of carbonate mudrocks. More than 200 samples were used to construct and validate four artificial intelligence models which are artificial neural network method (ANN), fuzzy logic system (FL), and support vector machine (SVM). The AI models showed reasonable prediction performance. Correlation coefficient values of 0.90, 0.85, and 0.82 were obtained for the ANN, FL, and SVM models, respectively. Also, the average errors are 5.9, 4.7, and 5.4% for the ANN, FL, and SVM, respectively. However, ANN provides a better option because an equation could be developed based on the optimized ANN model which would allow an easy and fast prediction approach. For example, the ANN shows predictions of 409.8, 531.1, and 677.8, while the actual rock hardness values are 407.4, 521.5, and 674.6, respectively. Furthermore, a new equation was developed based on the optimized ANN model, and the proposed equation can predict the rock hardness with an average error of 5.7%. Overall, this research offers a dependable and fast method for assessing the hardness of carbonate mudrocks, aiding in their characterization and the development of unconventional carbonate formations.