How to Apply Artificial Intelligence in the Oil and Gas Industry
Some might say artificial intelligence (AI) is the biggest buzzword of this decade. But, I can say AI is real, revolutionary, and has disrupted many industries. And I believe, its applications have only scratched the surface, with wider applications coming in the near future. Banking and finance, retail and e-commerce, healthcare, and logistics are a handful of industries that have tasted the benefits of using AI in the business.
While many industries have been revolutionized by AI, the application of AI in the oil and gas (O&G) industry has been limited. One of the factors that cause slow adoptions of AI is the unwillingness of O&G industry players to share and open their operational data. Sharing these data is considered as taboo and unthinkable for most of O&G companies because they think these data are sensitive and proprietary. But, data is the core of AI. Researchers and scientists won’t be able to find the use case of AI if the data are not shared.
This is contrast with what the other industries have done. Some industries, like e-commerce and finance, willingly share their data to be used for research in AI in the corresponding industries. Some companies publish their data in online communities, such as Kaggle, to enable data scientists and machine learning engineers to solve AI problems and challenges in their industries and to speed up innovations.
I have been blessed with the opportunities to work in the O&G industries for a few years. I have also been grateful to have chances to explore how various AI approaches can be used to improve efficiencies in various O&G operations, particularly in the inspection and maintenance area, during my PhD study. Hence, I would like to share in this article, how AI can be applied in the O&G industries, particularly in the inspection and maintenance sector. These are discussed in the following.
1. AI can automate the assessment of inspection and maintenance results
During inspection and maintenance, we aim to detect any anomalies that may threaten the operational integrity of O&G assets by using various techniques, such as ultrasonic testing, radiography, magnetic flux leakage, etc. Inspection is normally performed by authorized personnel who have undergone a rigorous training program. But still, errors and mistakes occur during inspection; not because the personnel is lack of the required training and education, but because humans are inherently full of bias in decision making.
Basically what humans do during the assessment of inspection and maintenance results are pattern recognitions. We identify patterns that are not supposed to occur and give them flags. Pattern recognition is the core of all machine learning algorithms, and these algorithms can do this job better and faster than humans. But, how we can enable these algorithms to identify anomalies in inspection results? Simply speaking, we just have to have a bunch of inspection data with the corresponding label on them (defect or non-defect) and train the selected machine learning algorithms with these data.
To illustrate further about this, see the image above. It is an example of object detection performed by a deep learning technique. Deep learning is a recent development of AI that is based on artificial neural networks. I won’t discuss deeper about deep learning here. But, anyone who is curious about this technique can read this awesome article. The same technique can be applied to, for instance, radiography images generated during inspection activities. Imagine we have an algorithm that can determine if the existence, the location, and the type of defects in radiography images. Isn’t it awesome?!
Here is the fundamental: machine learning algorithms enable computer systems to learn from and interpret data, refining the process through iterations to produce programs tailored to specific purposes.
2. AI can be a part of an asset’s surveillance system
O&G assets operate 24/7. That’s why we need to monitor these assets 24/7. For instance, an O&G pipeline normally has a system to detect any oil leakage. Currently, this system is powered by human operators who have to be standby monitoring the state of the pipeline system. They have to look at various operating parameters of the pipeline, such as pressure, temperature, and flow rate, to identify any irregularities that may indicate any failures or leakages. But, again, this type of task can be performed better and faster by AI because it involves pattern recognition type of task. AI can monitor several parameters at the same time, and combine this information to decide the state of the pipeline system. Human operators can only monitor a maximum of 3 to 4 parameters. Otherwise, they will get confused and the probability of making errors will be higher. Another application of machine learning in a pipeline surveillance system is to detect and determine the type of threats by third party works in the vicinity of pipeline (e.g. excavator, pneumatic hamper, plate compactor, etc.).
3. AI can optimize inspection and maintenance plan
O&G companies need to perform optimization of inspection and maintenance plan because performing these activities are costly. Consequently, they need to decide which assets need to be prioritized to be inspected and maintained and which assets can be excluded for later maintenance. The trend is to use a risk-based assessment approach to optimize the plan. But, because this assessment is performed manually, it is deemed time-consuming, makes great demands on efforts, and vulnerable to human biases and errors.
AI can tackle the problems faced by manual assessment. Machine learning algorithms provide increasing levels of automation in the knowledge engineering process, replacing much time-consuming human activity with automatic techniques that improve accuracy or efficiency by discovering and exploiting regularities in data. The evolution of the machine learning approach has attempted to reduce/eliminate the costly and laborious knowledge-engineering process involved in the development of knowledge-based systems. Accordingly, machine learning systems are capable of converting data and information into knowledge and enable cost-effective exploitation of knowledge resources.
I have written a research paper about the application of AI in the risk-based approach assessment in the following link.
In this article, I have given three general applications of AI in the O&G industry, particularly in the inspection and maintenance sector. Although AI can be used to perform various tasks in O&G inspection and maintenance, it doesn’t mean to substitute human personnel entirely. Even the best machine learning model can’t achieve 100% accuracy and precision. AI system is meant to augment human capabilities. A mixture of machine and human intelligence is needed, such that inductive learning can be complemented by the tacit knowledge of human workers. I believe the inclusion of human intelligence can refine the insights provided by the machine intelligence. This is in line with the concept of jidoka (or autonomation), derived from the Toyota Production System (TPS), which can be translated as automation with a ``human touch”.