When unplanned disruptions occur, an offshore drilling platform can struggle with multiple weeks of non-productive time. Our team recently talked to one of these platform operators and they have engaged a leading analytics company to address this specific cost. They are slowly working to solve this problem using predictive analytics. In today’s oil field, efficient operation depends on 3 types of predictive analytics.
This blog explains each of these 3 types in the oil and gas context. For an explanation of the analytics system itself and how it works, you can reference “How Predictive Maintenance Works – 5 Steps“.
Simple Analytics
The best picture of simple analytics is predictive maintenance. Data from a vibration sensor or temperature sensor directly predicts the most common failure scenarios. Common scenarios include a worn bearing or a physical stress point.
Sometimes a few data points work together to provide a conclusive prediction. For example, a noisy or vibration prone piece of equipment is more complex. It may only identify a failure scenario after assessing multiple data points in combination. Even with multiple data points in combination, a prediction still only requires relatively simple analysis.
Simple analytics consists of simple mathematical averaging and comparison. Most equipment monitoring systems have this type of calculation and alerting available for built in sensors. When new sensors are added to equipment that is already operational, a 3rd party system processes the data and generates the required alerts. In some cases the alerting platform can even integrate with service software to set up service tickets for scheduled maintenance activities.
Examples of companies for each type of analytics can be seen in the video “15 Predictive Maintenance Systems for Oil & Gas“
Process Analytics
An automated process combines multiple inputs and outputs into an interdependent web of operation. This means that an alert in the system consists of multiple interdependent parts that all contribute to that outcome.
With the increasing maturity of artificial intelligence, control system platforms now include advanced analysis capability. These platforms already have access to many intermediate data points in the process. Using this data with artificial intelligence can predict failure scenarios.
For example a 3 phase separator in an oil field may have increased pressure readings in one part of the tank. If those readings are outside normal operation there could be multiple causes. The control system software considers multiple readings and combines them. Using an analysis framework or artificial intelligence the system can then identify recommended corrective actions.
System Analytics
Now let’s go back to the offshore platform scenario we started with right at the top. On an offshore platform the dependencies reach much further than one individual process. Although all the processes may interact with each other, there are also multiple elements outside any of the processes that could be factors as well.
For example, high winds from a specific direction could interact with a process that is within the acceptable operating levels to create a failure scenario outside that process and cause an alarm. An alarm can scale down or halt operation until a root cause is identified and resolved. These interdependent system level issues can be very difficult to diagnose. Predicting patterns before a failure is even more challenging.
When these complex system failures represent a high enough cost, the price of a comprehensive data science platform can be justified. This platform analyzes seemingly unrelated data points to identify a data correlation for prediction. In reality, some scenarios are not predictable enough to avoid costly downtime but when they are, the cost benefit is significant.
Foundational Requirements
All types of analytics contain important common elements. All analytics is dependent on good data sources. For failure prediction it needs to be live data wherever possible. The live data requirement implies a solid network connection between the source of the data and the recommendation engine that does the live analytics. Cisco Systems has a comprehensive capability to provide secure and hardened network infrastructure for the most difficult environments.
Summary
In summary, the data that is available in today’s oil & gas infrastructure often shows us clear patterns that predict failures. Sometimes these patterns are simple, single sensor data streams. At the other end of the spectrum the pattern may be disguised in a complex relationship of variables and interactions. In either case, predictive analytics can make oil and gas operations more safe, more efficient and more sustainable.
Resources
Video Resources:
How Predictive Maintenance Works
15 Predictive Maintenance Systems for Oil & Gas
This article was helpful in our search to understand our analytical data. Do you have any recommendations on system analytics integrations? Thank you.
I’m glad that you found this blog helpful. I review 15 different analytics systems in the link at the bottom of the blog. It’s a youtube video that maps company names onto the 3 types of predictive analytics. For system analytics I believe the ones I called out were SAS and Trendminer as the big complex systems and then I identify 5 smaller systems that could work as well.