Closing the Integration Gap for the Next Generation of Drilling Decision Support Systems

Abstract

This paper identify the lack of integration between drilling decision support systems and their users as a barrier towards better decision support and increased drilling automation. In part one of this paper we outline the workings of a next generation of decision support systems that reduce this gap. In part two we present our preliminary results with a time series anomaly detection technique, which the decision support systems we have described would require. Part I -IntroductionThroughout the development of integrated operations and intelligent energy, there is a shared expectation of better decision support in the future. This is to be achieved both through better collaboration between people and through improved data analysis and modelling. For the latter, the most frequently cited challenges are data quality [1, 2] and breaking down "data silos" to make data easier to find and integrate with other data sources, or to make different computer systems talk to each other. Another expectation is increased levels of automation, where the aforementioned decision support is gradually expanded to let the computer not only advice on a decision, but to actually make the decision and carry it out. This too requires a close integration of sensors, different computer systems and machinery.

Risk Factors Evaluation for Monitoring of Well Drilling

“…Nybø [63][64] solved a similar problem. In this work, a hybrid system is developed that includes a physical model and AI.…”

Data-driven indicators for the detection and prediction of stuck-pipe events in oil&gas drilling operations

“…In (Roberts et al, 2016) the task of detecting and recognizing a kick is studied from a cognitive point of view; a decision tree is designed to aid the driller in the detection and reaction to the kick. In (Nybo and Sui, 2014) a clustering method is employed to recognize anomalies in the mud log data. Finally, some works are concerned with the prediction of stick-slip phenomena as well, see (Efteland et al, 2015).…”

TiWS-iForest: Isolation Forest in Weakly Supervised and Tiny ML scenarios

“…The relaxation of the unsupervised settings is typically reasonable in the context of Decision Support Systems (DSS). In recent years DSSs became pervasive and today are applied to all fields where complex and delicate decision have to be made to assist human operators in the decision-making process like in medicine [11], precision agriculture [12], energy [13], environment [14] and security [15,16]. These systems are equipped with anomaly detection functionalities that allow to automatically monitor the process, giving feedback and alerting the human user to make a possible action; in this context end-users can provide feedback on anomaly detection module suggestions [17], making the weakly-supervised scenario that will be considered in this work reasonable.…”