The biggest technology revolution we’ve seen, you’ve probably heard, “AI is the future”, but what can it do? How does this apply to the seismic interpretation workflow? And what are the implications for your people, your business and the energy industry as a whole?
AI or artificial intelligence isn’t really a technology. Designed to do the sort of things a mind can do, the key to AI is intelligence. Just like human intelligence isn’t one single dimension, AI is equally varied and structured in the same way. AI is many advanced technologies brought together to enable a machine to act with human-like levels of intelligence; providing context and meaning to the information it learns, imitate it and act accordingly.
To generalise, deep learning is a subfield of machine learning, and machine learning is a subfield of artificial intelligence. Not a definitive explanation, the best way to get a true understanding of AI is to recognise what technologies underpin it and how they can be applied to seismic interpretation. We’ll focus on the principal areas which are relevant to enhancing subsurface understanding and how they can be used to advance the seismic interpretation workflow we know today.
“Geoteric’s AI-Fault interpretation had an immediate impact on our subsurface understanding, enabling an improvement to our well planning process.”
Wellesley Petroleum |
AI’s ability to learn is essential. The ability to learn from experience without being explicitly programmed is arguably what makes machine learning (ML) intelligent and differentiates AI from other forms of computer automation.
Whilst traditional computers cannot fix problems on their own, ML can. Not only can ML learn from past data results. It can arguably make a better decision and do so at an incredible speed. This is often illustrated in popular culture examples, when AI bots take on the professionals.
However, ML is a large topic. It has several different learning styles which are important to define.
Much like we learn and understand ourselves, ML must be trained too. In subsurface understanding, getting it right in our domain is more than just data science, it takes a blend of everything. From maths, physics, geology, geophysics, crucially user experience with current domain knowledge all must work together. Without that, these capabilities on their own are not enough.
Supervised learning takes labelled data which has been organised and described. It deduces the relevant features which characterises each label and learns to recognise those features in unseen data.
For example, you may show the algorithm a large number of labelled faults in seismic data. It would learn how to recognise a fault and spot it on any number of unseen seismic cubes and classify it.
Unsupervised learning doesn’t require pre-defined labels. Instead it takes an unlabelled data set, find similarities and variances within the data set, and categorises them into its own grouping.
Self-organising maps (SOM) is possibly the most common type of unsupervised learning in subsurface understanding. Classifying the data into clusters to identify important geological features during exploration, production and development.
Reinforcement learning is more of a trial and error feedback loop. Much like our popular culture example , the algorithm creates a “gaming environment”. After each action, it is told if it has “won” or “lost” so it can build a picture of actions which results in success.
As described below, interpreter led changes to the results can be captured within deep learning algorithm which runs on top of the seismic data. This will be used for quality control (QC).
The AI fault interpretations are proving to be very useful inputs for well planning. They are full-field from the shallow overburden to the reservoir, our most comprehensive fault interpretations to date. This has improved the rigor of our work so that our decisions are based on a better understanding of where faults may introduce drilling risks, and so that we can better understand lateral changes in our reservoir. They have also make it easy to compare quite different fault patterns defined using different seismic volumes. That this type of information can be provided in 2 weeks’ time is a real game changer.
Lead Geophysicist, Valhall | Aker BP
A technique to enable the application of ML is deep learning (DL). Practised in image recognition or speech, in subsurface understanding we use the imaging side.
For DL to learn significant features within the seismic, it applies a layer (sometimes referred to as a convolution) to gain an understanding of all the features within that layer. The clue is in the word deep, the output of each filter provides the input for the next. As the processes continues to repeat; detecting both low and high-level features, the number of layers increases and as a result can be built up of countless layers.
To achieve this learning, it is powered by a neural network (most commonly CNN). A conceptual model which is based on the brain’s biology and mimics the network of neurons. As a result for an interpreter, removing the considerable amount of manual effort and time intensive feature extraction and classification.
Much like seismic interpretation requires the application of many different methods to complete the workflow, such as high-level screening to build a structural understanding, and correlation of RGB colour blending to fill out the picture - DL works in the same way.
It will gather insights across the seismic cube to draw a more comprehensive picture to aid better decision making. Not only can it create a straightforward view of the structure, it can combine multiple techniques and run several tasks at once.
This means more sophisticated problems can be solved, for example identifying subtleties across the structural geology which will help interpreters effectively target areas for investment through a much cleaner picture than traditional methods.
Geoteric delivered to the Valhall asset several fault confidence cubes for various types of seismic data, in particular 2 compressional PP images with different acquisition and processing sequence and their corresponding converted wave, PS, images. These confidence cubes are a great information for integrated well planning, helping for example assessing fault confidence, positioning uncertainty and ultimately the risk associated to the fault if intersected by the well. Summarising the analysis of 4 seismic volumes into a single fault plane and a risk description can’t be done in an automatic manner and still requires the interpreter judgement, but the Geoteric workflow is in itself a huge step of data integration.
Senior Geophysicist, Valhall | Aker BP
As highlighted above, with such enormous potential it is no wonder AI remains high on the agenda. But whilst the benefits are so great, widescale adoption across the subsurface arena has been slow to date as new technology does not come risk free.
As with any effective risk management, it pays to know what the risks are upfront and how they can be mitigated. We’ve considered the top five risks, defined what we consider these to be and how they have been addressed in the design of Geoteric’s AI offering.
When the stakes are high, how can I trust a machine to make a decision? What happens, if you don’t agree with the result? Do users have the ability to take back control if they need to?
As we’ve said previously, DL layers can be countless. By nature, countless layers can be difficult to understand, but they don’t need to be. Trusting results isn’t knowing how many layers there are, for AI to become viable in the subsurface the result needs transparency and auditability driven by the user. You want to have the ability to check the accuracy of the results and understand how those results can be further developed overtime. Thankfully, there are practical steps operators can take to understand AI solutions and demonstrate safety of use.
Back in 2018, Gatner identified the real power of AI rests in the ability to augment the human to enable effective decision making. AI’s greatest value and impact comes from augmenting human interaction. Therefore, a “human plus machine” concept is paramount as opposed to a to “human versus machine”.
Developed by geoscientists and software developers, Geoteric’s AI Seismic Interpretation is fully integrated within the workflow. In short, steered by three core elements in our mission statement; science, people and technology, Geoteric designed an AI- driven Seismic Interpretation platform with the interpreter in control.
Rather than designing a platform to perform those “out of the box” set tasks, it is advanced in such a way that results are easily integrated into the traditional workflow today for thorough quality control. Results can be evaluated by the interpreter on any inline, crossline and timeslice in real time, fault sticks can be automatically extracted on the AI fault attribute to allow users to edit and effectively QC results.
With people at the heart of the design, it combines both the strengths of human knowledge plus the capability of the machine.
Trust can be best illustrated through the design of Geoteric’s AI Fault Interpretation offering. Geoteric’s AI Fault Interpretation has been designed to build a level of trust and confidence across three steps which are fully integrated into the users workflow;
1. Regional focus
2. Reservoir focus
3. Pre and post drilling well focus
Much like the traditional workflow, which requires the application of different methods to develop a rich understanding, the same concept applies to this offering. Designed to improve knowledge and understanding at each decision point, knowledge can be effectively built throughout the workflow. It is built in such a way that results are easily integrated into the traditional workflow today for thorough quality control at each step.
The first step of an integrated AI interpreter workflow is evaluating the accuracy of the “regional focus” result on any slice within the seismic cube in real time. Therefore, there is a necessity to calibrate results, to ensure users do not solely have to rely on the out of the box results. This means, customers will want to develop and train their own interpretation, capturing the detail of their data in their own unique network.
Networks can then be applied across different datasets, transferred between regions, basins and epochs for company specific enhanced networks to help deliver more accurate results. Multiple networks can be used to develop a more efficient and impactful understanding whilst building confidence to ensure care is taken to avoid misinformation and unintended bias which could hinder as opposed to enhance the process.
Geoteric’s AI Fault Interpretation has been designed so fault sticks can be automatically extracted or grouped (depending upon business need) to allow further user edits. As projects start to focus on features of interest, these edits will be used to feed into the “reservoir and well focus” steps if required.
Results from the network not only improve, those improvements can be applied to future datasets. Augmenting human interaction, you can ensure AI is safe, fair and explainable. Resulting in Geoscientists being able to explore vast amounts of seismic data, gain confidence in the AI output to derive insights and test their assumptions to build a greater understanding of the Earth.
The real value of Geoteric’s AI Seismic Interpretation isn’t just performing set tasks or saving time fault picking.
Subtle details and trends can be difficult to interpret manually. Added to that, varying data quality, can be an issue all too real in subsurface understanding.
ML’s ability to learn and improve over time, means that poor data can be processed with far greater accuracy than is often possible to see as a human interpreter, for example data degraded by the presence of a gas cloud as illustrated in our AkerBP case study. Geoteric’s AI platform can identify more events quicker with a greater level of accuracy. It can see beyond false signals which give unclear or disappointing results in traditional fault detection attribute analysis.
Geoteric’s AI platform is capable of working with a range of data qualities, new high-quality data works well, but so does readily available off-the shelf data, as shown by our case study work on publicly available release data sets. It not only scales well, with ever increasing amounts of data volumes which by its nature will have poor or missing data. Its’s tolerance for imperfect data provides an advantage especially when working with older data sets and is illustrated in our GoM case study.
Using the data you have collected, the data conformant approach identifies faults in a fraction of the usual time with a greater level of accuracy. Results can then be optimised in Geoteric’s standard software to add in greater clarity of detail and increase confidence levels when used in combination with traditional lines of geophysical imaging, such as spectral decomposition. Allowing interpreters to combine their knowledge with the best possible picture, informed decisions can be made much faster throughout exploration, development and production.
Added to that, interpreters are susceptible to unconscious bias. This has big implications on fault sizing and is often driven by lack of time or poor image quality. To avoid the dangers of anchoring or simplified views of complex systems, the initial unbiassed AI assessment of the structural framework provides a detailed answer. It also provides a great catalyst to discuss, without significant interpreter bias or reputational investment, encouraging a fast, open and comprehensive assessment of the structure.
Given the nature of the industry, the sector is unsurprisingly exposed to a specific set of risks across IT, people and processes across the full supply chain. You only need to look at risk management procedures across the sector to know this is top of mind. Today, awareness of the threat is known, cybersecurity is paramount.
As our operators continue to rightfully integrate security efforts across every facet of their organisation, effective integration needs to be transparent. Leveraging the latest AWS technology to ensure our customer’s data is fully isolated and secure, we use a multi-layer approach to security.
Forget human versus machine, Geoteric’s AI Seismic Interpretation platform presents us with an opportunity to do more, learn more and flourish in ways we can’t yet even imagine.
Upstream companies still need do more to thrive and control their costs to ensure they continue to take investments forward, grow and commercialise them. Operators continue to look to understand what is the opportunity that AI presents to their workflow, to understand the impact it can have on their results, their processes and their people.
As we operate in a time of transformational change, expectations are higher and far reaching across the natural, political and social environments. In an uncertain price environment, businesses look towards what they can control, staying resilient whilst strengthening their long-term sustainability. These challenges and expectations reaffirm the need for our industry to reinvent. Not just to keep costs low today whilst maintaining health, safety and environment (HSE) standards, but to continue to reduce risk, protect their people and the environment, all whilst unlocking new resources and producing more affordable energy with minimum investment.
There is certainly no doubt any business with large amounts of data can use AI to replace those monotonous, repetitive tasks to enable cost reduction in the short term. However, a process easily automated only scratches the surface of AI potential. By nature, AI offers a radical new reality. Designed to be helpful, the ultimate purpose for AI adoption in the seismic workflow is to provide a greater understanding of the Earth.
If you think of AI as replacement to your current workflow, your focus is misguided. AI isn’t just about generating new levels of efficiencies; its real power is opening new ways of working and generating value beyond what was expected. This means processes shouldn’t simply be replicated, instead, processes should be re-imagined and re-engineered to utilise AI to its fullest potential.
One truth remains: no tool will fully replace the art of interpretation. The best integrated AI seismic solutions harness knowledge, present the opportunity to learn more, do more, providing support when users need it most. In this way, interpreters can add more value to the business by a more thorough understanding, drive operator growth in new ways through new levels of decision making and more confidently outline the difference between a good and bad prospect?
In the Valhall seismic obscured area, where classical fault interpretation is very difficult even on converted wave data, all discontinuities are captured easily by the fault confidence cubes and the interpreter has the ultimate decision to interpret the discontinuity as a fault or leave it. This process can be supported by AI using automatic fault pick interpretation in Geoteric. Though not fully automatic, going through this workflow is also a way to get into the data, revealing the main structural trends and the average size of fault planes.
Senior Geophysicist, Valhall | Aker BP
Over the past century, technology has helped us gain a better understanding of our world. We believe a thorough understanding of the Earth can shape new perspectives and provide solutions to some of the greatest challenges we face today, which is why we’ve spent over 30 years expanding what’s possible in geological analysis.
As we journey down from the surface, we want to help geoscientists unleash the potential of their data to gain a deeper understanding. So, no matter what you’re looking to achieve, be it powering the industries of today, to protecting the world of tomorrow. Anything is possible when you can see beyond the surface with Geoteric.
Geoteric enables geological evaluation by integrating interpreter knowledge with the data through a suite of modules designed to improve quality, speed and understanding.
