Making informed decisions requires clear, moment-to-moment understanding of the situation. Exebenus Current ML provides real-time predictive situational awareness aimed at avoiding or reducing costly operational downtime and lost time.
During planning, our machine learning agents can be used for offset well analysis to evaluate risks and prepare ready-for-action contingency plans. This reduces potential downtime and helps eliminate the hesitations that can add up to invisible lost time.
During execution, our agents use real-time data to recognize deteriorating conditions that may lead to stuck pipe and other issues. This gives your crew a heads-up 30 minutes to hours ahead of potential events, prompting them to implement mitigation steps and avoid downtime.
- With Exebenus Current ML, insight propels productivity.
- Predict hazardous events
- Optimize offset well and root cause analysis
- Provide real-time situational awareness
Exebenus Current ML Stuck Pipe agents are designed to predict, in real time, high-risk conditions related to pressure differentials, hole cleaning conditions and wellbore geometry—conditions that, without intervention, typically result in costly stuck pipe situations. Warnings are provided 30 minutes to hours prior to potential events, giving rig crews sufficient time to take mitigating actions.
When used on historical real-time data as part of offset well analysis, the agents can identify unreported near misses and provide guidance for optimizing performance in the future.
At Exebenus, we have chosen to develop targeted machine learning models rather than complex models. Why this approach?
Complex models consume vast amounts of data, and take longer to set up, train and run. In contrast, our targeted models solve well-defined problems and deliver more accurate predictions. They use data that’s always available in real time on the rig, which means our agents can be used anytime, anywhere, easily.
Our generic “out of the box” models are adaptive enough to be used in any geographic area. They predict stuck pipe situations in various well operations. All it takes is some manageable fine tuning and model training using past offset well data.
Our robust agents return useful predictions based on a range of data quality. In fact, they do their best work when consuming raw, unfiltered data, and even handle data gaps.
Exebenus Current ML provides reliable predictions within a useable timeframe, requiring only a connection to the existing WITSML system.
Exebenus Current ML agents consume real-time or historical WITSML data that is readily available in all well operations. Minimal human intervention and no data filtering or cleaning required.
In trials, our agents have performed “out of the box” using raw data that was not prepared in any way.
For ease and speed, the agents output WITSML data to integrate with your operation center’s workflows and familiar real-time WITSML viewers. Within only a few days, your teams can be monitoring and analyzing data, and advising rig crews.
Exebenus Current ML is cloud-based, and agents can be deployed stand alone or as a package.
“We’re seeing machine learning coming into the real-time operations space primarily to do two things: to predict and to optimize. Often it’s about detecting anomalies early and avoiding hazards… uses where machine learning is improving safety and reducing operating costs.”