Distributed Monitoring With Integrated Probability PCA and mRMR for Drilling Processes

Abstract

Geological drilling processes involve many variables, and their relationships and dynamic characteristics are highly complicated. In the geological drilling processes, the changes in the operating performance may be invisible under non-optimal conditions, while the data distribution may have significant deviations. The quality of the data collection is difficult to guarantee due to the underground measurement and transmission environment, which increases the uncertainty of the operating condition. Operators have struggled to manage performance monitoring for drilling processes over the decades. To improve the utilization of the instrument measurement data, this paper develops a distributed monitoring method with integrated probability principal component analysis and minimal redundancy maximum relevance. The related process variables are divided into sub-blocks by minimal redundancy maximal relevance algorithm. Then, local detection is gathered integrated probability principal component analysis to formulate global monitoring statistics to realize the whole monitoring scheme. Lastly, real-world production processes are used to verify the feasibility and superiority of the new method. The proposed novelty involves constructing local monitoring models for independent variable sub-blocks taking into account the minimal redundancy maximum relevance of the variable space.