Development of an Automated Inflection Point Detection Algorithm for Deformation Rate Analysis

Abstract Accurate estimation of in situ stresses is essential for safe design of underground civil and mining engineering structures. Traditional in situ stress measurement methods often require extensive time, resources, and access to test location. Moreover, with these methods, it is challenging to consider the effect of high uncertainty due to limited number of tests. A laboratory-based method deformation rate analysis (DRA) is considered as promising approach allowing the use of core samples to infer in situ stress conditions and the consideration of variability and uncertainty. The key to the successful use of DRA method is the accurate detection of inflection point on stress vs strain-difference curve signifying the transition associated with prior stress states. Traditionally, this point has been determined through subjective visual assessment. Therefore, recent studies have searched for standardized methods to reduce subjectivity. This paper introduces an objective and transparent repeatable mathematical model for the detection of inflection point. First model parameters were fine-tuned using the results of DRA experiments conducted on sandstone samples. Then, the constructed model was implemented on unseen data or validation data collected from the current literature. The results show that the model presents acceptable accuracy for the detection of inflection point independent from lithologies, measurement tools, stress path, and preloading conditions.