Step #1

Based on my knowledge, I know that AI is used in the workplace (i.e., the oil and gas industry) to predict possible failures and minimize the risk.

Since the beginning of the oil and gas industry, multiple failures have occurred across all disciplines. These failures cause multiple deaths, disasters, and loss of huge amounts of money. Therefore, petroleum engineers and all workers in the oil and gas field are always trying to learn from failures, minimize risks, and enhance production. In the last few years, the focus has shifted to implementing AI, ML, and digitalization in this industry to make operations more efficient, reduce costs, make better business decisions, and encourage innovation. One of the new AI and ML applications in the oil and gas industry is an AI-based ESP that predicts upcoming failures and provides recommended actions for engineers.

The Electrical Submersible Pump (ESP) is an artificial lift method that is used to lift oil and gas up to the surface and increase production rate. ESP is well known for frequent and unpredictable failures. In addition to a lack of continuous monitoring and analysis of performance due to manual supervision, leading to short ESP run life. Therefore, an AI-driven ESP event management and surveillance system is implemented to overcome the current challenges.

The new system is an advanced ESP predictive failure analytics, which is a unique technology that combines AI, ML, physics-based rules, and expert knowledge-based approaches. The ML model continuously monitors the ESP sensors, predicts the ESP status using the sensor’s real-time data instead of fixed engineering rules, and provides results and feedback to engineers through a visualization tool. The platform goal is to provide early and interpretable predictions of ESP failures with failure alarms, and to estimate RUL (i.e., Remaining Useful Life) to take preventive actions, aiming to optimize production and extend the ESP run life.

The AI-based ESP systems vary depending on the well location and conditions. Each field has its AI platform depending on its needs and conditions; therefore, different systems are implemented with various ML algorithms used. These platforms vary in prediction accuracy, and some of them exceeded 95% accuracy. In the end, they all have the same approach, which is reducing the unplanned events and operational costs, and enhancing production for a safe and better oil and gas future.

Step #2

Prompt: Suggest ways for AI to be used in the Oil and Gas Industry

ChatGPT Link: https://chatgpt.com/share/69f53424-dd68-83ea-898a-6f74be060a53

Step #3

AUC Professor: Dr. Sherif Fakher

How is AI currently being integrated into petroleum engineering education?

In our Industry, AI is integrated into any data interpretation, either in seismic, well logging, or any process. It helps engineers get faster solutions with minimized human errors. For education in the oil and gas field, AI is integrated into the software. All the software used in our industry includes an AI component. For example, Pipesim Software has an AI component that predicts flow rate based on your input data and physics-based equations, optimizes production, detects anomalies, and speeds up simulations by replacing equations with trained models.

Do you think students using AI tools (like ChatGPT) are learning more or relying too much on shortcuts?

If AI tools are used in a suitable and balanced way, they can be a strong learning platform for students in all majors. But it must include two parameters. First, students must ensure that the data generated from ChatGPT or any AI tool is correct before using it. Second, sometimes the AI tool gives outdated information or gives wrong information when it does not know the answers to the student’s questions. Also, it sometimes generates wrong citations, especially in research. Therefore, students must take care while using AI in learning to make sure they are using it correctly and ethically. In addition, they must ensure that their professors permit them to use the AI or not if they are using it as part of a course.

How can professors ensure academic integrity when AI tools are widely available?

The petroleum courses at AUC are designed to force students to do their work on their own without using AI. Most of the assignments and projects require multiple analyses and interpretations of multiple data files; some of these analyses and interpretations are done on different software. Therefore, even if the students tried to use AI for cheating or getting credits, they would not be able to. In addition, all the problems that require calculations cannot be solved using AI because they are taught to students in unique ways that cannot be done by AI. So, if these problems are solved in different ways, it will be so clear that they used AI. But in general, all the academic integrity rules are listed properly and clearly in the syllabus, so when there is any kind of violation, students will have an academic integrity case.

Are there areas in petroleum engineering research where AI is being used but should not be trusted yet?

There are a lot of untrusted areas in AI data interpretation in research because the oil and gas industry is highly volatile and has high risk. So, unfortunately, we are one of the industries where the answer is not yes, or the answer is not no, in most of our questions. So, we need to use our engineering logic and think about every piece of information properly, and use simulators to simulate well conditions to reduce the error possibility.

What ethical concerns arise when students depend on AI for technical problem-solving?

The major ethical issue is cheating because students can give the AI any question or mathematical problem, and it will solve it without any effort from them. The second concern is the ability of the students to get credit for themselves for information or tasks they did not do, or even for getting wrong citations. Sometimes ChatGPT hallucinates and gets information wrong, and if you use it, you have to cite it. But the problem lies in getting the wrong citations. So, the result is that the AI tool generates information from unknown sites that have a high probability of being wrong, cites it to the wrong people, and gives a wrong citation

Should AI be taught as a core skill in petroleum engineering curricula?

Yes, it should be included in an ethical and well-structured way. AI is part of the future and a huge contributor in our industry today, so students need to know how to use it ethically and properly to be ready for the future industry that is using AI as a crucial aspect.

Step #4

AUC Part-time Professor and Industry Professional: Dr. Mostafa Oraby

What AI tools or systems are currently used in your company?

An AI tool that provides missing information in some of the wells’ data. In many cases, when engineers run wireline logs in wells, some of the logging tools malfunction and do not record data, leading to the loss of some data. So, the AI can be used to fill in the gaps between data. For example, if one of the tools did not record, the AI uses data from other neighbors’ wells and builds the machine learning process. Through the other well data, the ML model finds the missing information. Also, in the data interpretation, when some of the highly experienced petrophysicists, who builds well recognized interpretation, are not available, the AI can use the predetermined and calculated interpretation to fill in the gaps in the wells that do not have an interpretation.

Have you seen AI reduce costs or improve efficiency? Can you quantify that?

AI improves efficiency by reducing mistakes and errors in the data, which leads to saving costs on a large scale. The oil and gas industry is very critical, and any small mistake can lead to disasters. Therefore, in well logging, the cost is not the biggest driver. The availability of data and good interpretation is more important than cost. Sometimes we reduce the amount of data that we acquire if we have multiple wells to avoid repeating all the data in all wells, because that would be very costly. But again, the cost is not the driver. The driver is finding out very well-interpreted data and very well-acquired data because that’s very critical in the evaluation of our reservoirs, where the money value comes later. Having good interpretation and full data available in the field adds more than just spending a few dollars on acquiring data.

Are there situations where engineers ignore AI recommendations? Why?

No, ignoring AI recommendations would be a mistake, especially if there are not enough very highly experienced people. It’s really important not to ignore the importance of AI in our interpretation and our data acquisition. Having a full suite of data and parameters, very well-recognized parameters, is more important than anything else in our daily data acquisition. And the value of the field is dependent on how accurate the data are and how good the interpretations are. So, it’s very important not to ignore the importance of AI in our daily interpretation.

How does your company handle accountability if an AI system makes a wrong prediction?

AI accuracy depends on the amount of data we feed and the learning process. If there is not enough data to feed the AI search engine, and if there is not enough data where we can educate the machine learning, then uncertainties are included in our planning. So, uncertainties are added based on how much data is available, and what the what-if scenarios are. So, it is important to include what-if scenarios when there is a lack of data. We must have uncertainties in anything we generate; the more data we have, the better the results are, and the lower the uncertainties are.

Do you think AI could reduce the number of petroleum engineering jobs in the future?

No, AI will not reduce the manpower. Manpower and reservoir engineers’ brains are always needed. Engineers are assigned to look at the AI-generated data and make sense of what they see. We cannot use the AI right away. We must review the AI results and ensure they make sense in light of their field experience. The human brain is needed to find out if there is anything that does not make sense in the data generated by AI.

Can you describe a situation where human judgment was more important than AI?

In Angola, there was a single well that had only old and unreliable data that did not match the seismic data. In this case, AI is used as it is well known for its ability to identify inconsistencies, such as differences in stress direction, and generate possible scenarios by combining various data sources. However, it could not determine the accurate data. Therefore, engineers used their logic and experience and evaluated the data to choose the most reasonable interpretation and plan the new well accordingly. And as a safety factor, they put the worst-case scenario and included all the additional tools to collect better data and reduce uncertainty. This example proved that human judgment is more important than AI when dealing with uncertainty and incomplete information.

Step #5

My Reflection:

In the first interview with Dr. Sherif, I was surprised by how the AI is actually an important part of our industry and our education in general. As I know as a student, most of the professors hate AI tools because most of the students use it unethical and for cheating. So, after interviewing my professor, I was surprised by how he is encouraging students to use AI as a helpful tool for better learning, even in research. I noticed that AI has become a crucial part of our daily lives, like social media. Even though we know that social media has a lot of harmful effects on us, we still use it because we cannot ignore its existence, and we cannot stop using it, just as with AI. Knowing that AI is part of the software that I am using makes me excited and curious about understanding how this software works. However, I am scared about overrelaying on AI. Previously, in research, as an example, I used to search for references and books manually and read them to identify whether it is helpful for me or not. But recently, I started using AI for researching references and summarizing them instead of reading those papers to identify if I am going to use them or not. This idea scares me to rely too heavily on AI and forget how to research properly; however, I will try to learn more about using AI as a core skill, as my professor advised. Even though there is a risk of relying too heavily on AI in everything, it saves a lot of time in any work, and it has become a crucial part of our present and future. Therefore, I will keep learning about using AI ethically and properly in performing multiple tasks.

In the second Interview with Dr. Oraby, I was surprised by the AI applications in well logging. I knew that AI is applied now in everything, but I did not expect that it is even applied in logging. It is my first time knowing about how AI is used to fill the gaps between the data using previous data, or even predicting the data in the gaps using physics-based equations. As I know before, well logging is just using some tools that are lowered in the well to measure different well parameters, like porosity, saturation, clay volume, etc. Therefore, I did not expect any application of AI in this sector. Also, I was slightly surprised by how efficiency is more important than cost. I know it is more logical to prefer efficiency over cost, but one of the most common sentences that I hear in the oil and gas industry is “Every action and every process depends on cost and economics. It is all about cost”. So, hearing about preferring efficiency over cost made me shift my way of thinking to look further to see that preferring high data efficiency now will cause more earnings, as the value of money comes later. To be honest, that made me excited to learn about economics. I do not know anything about economics, even the basics, and I used to hate economics and financials in general. But now, I am a little curious about them, and I started thinking about taking an economics minor. As part of my dream in the future, I want to be a logging engineer after graduation, so knowing this information pushes me to be more curious about well logging, field data interpretation, and economics.