An Efficient Meta-Ensemble Paradigm for Modelling Poisson’s Ratio and Maximum Horizontal Stress in Casing Collapse Hazard

This study develops an efficient meta-ensemble paradigm for predicting geotechnical-geological parameters namely static Poisson’s ratio (PR) and maximum horizontal stress (MHS) significant to casing collapse hazards. Engineers strive to reduce the risk of casing collapse due to drilling through the meticulous design of wells and operation practices. Thus, identifying casing collapse hazards in oil and gas wellbores necessitates a thorough geotechnical examination. For this purpose, a stacking ensemble paradigm was developed to predict PR and MHS, the two most important variables of casing collapse hazards. Three conventional models (viz., gradient boosting regressor, decision tree regressor, and k-nearest neighbour regressor) and linear regressor were used as base models to construct a stacking ensemble paradigm (ENSM) in which a random forest regressor was used as the meta-model. The performance of the ENSM model was compared with two additional standalone models namely feed-forward neural network and extreme learning machine. Experimental results exhibit that the ENSM paradigm delivers higher prediction accuracy than all other employed models with R2 values ranging from 0.9755 to 0.9991. The experimental results confirmed that the proposed ENSM paradigm achieves high prediction accuracy in predicting the PR and MHS of soils and can be considered an effective approach to aid engineers in assessing the risk of casing collapse hazards in oil and gas wellbores and operation practices.