A decentralized operating performance assessment for geological drilling process via multi-block total projection to latent structures and Bayesian inference

Abstract

Geological drilling process is an industrial process that contains a lot of variables, and the relationships and characters among them are also complex. Accordingly, implementing operating performance assessment by conventional methods has always been a problem that operators are dedicated to solving. In fact, geological drilling process is a complex industrial process involving multiple systems and stratigraphic uncertainties, which makes assessing its operating performance difficult. A decentralized operating performance assessment based on multi-block total projection to latent structures (T-PLS) and Bayesian inference is proposed for the geological drilling process. Utilizing the variational trends of the detection variables, the most related variables can be grouped in the same block, and the process capability index determines the operating performance grade. This is followed by a T-PLS algorithm-based operating performance assessment model and Bayesian inference. Each sub-block’s assessment results are integrated to achieve a comprehensive performance assessment. Last but not least, drill data show that the proposed method is effective and superior. The proposed method provides better accuracy and generalizability in assessing drilling performance. The novelty of the assessment scheme involves that a decentralized framework is proposed for operating performance assessment by identifying normal operating conditions first and then constructing the multi-block T-PLS-based monitoring model on the local sub-blocks.