History Predicting the Future

Using data science, AI and historical data to create a more sustainable future for late life assets

Only a few years ago, many within the oil and gas industry perceived the use of data science and moves towards using Artificial Intelligence (AI), to be not much more than a novelle, highly technical phenomenon; too expensive and to challenging to implement, especially on existing and ageing assets.

Fast forward to 2021 and the technology is being adopted at an increasingly rapid pace. This has been partly driven by a much lower cost base and the need for more effective technology adoption to facilitate savings, but also because there is a renewed and urgent requirement to reduce physical POB on assets for reasons such as minimising the risk of physical interactions on plant and other unforeseen curveballs such as Covid-19.

Rather than lacking the horsepower to produce valuable data, operators are now struggling to compute the large and sometimes overwhelming screeds of information being churned out every minute of every day; a change from the challenges of the previous decade where asset owners struggled to obtain readily available quality historical information that could be turned into anything of significant or alternative benefit.

So what can be done with all this data – and how can it be turned into a valuable commodity for asset owners?

Machine learning and data science are providing useful ways to compute and digest previously unused data, enabling faster and more improved decision-making processes.

At Imrandd, our software uses machine learning to collate and then monitor complex operations, identifying risks before they occur, as well as providing confidence in where, when and how to optimise inspection and integrity scopes. Gaining control and visibility of asset performance through the collection, interpretation and management of data facilitates better planning, cuts out the fat in integrity management and helps maximise remaining life.

Our engineers and data scientists use an in-house suite of cutting-edge analytical tools and predictive modelling to manage the vast amount of inspection and maintenance information that was previously locked away, inaccessible or too hard to decipher. This enables customers to leverage information across other similar assets and equipment types to help 'future-proof' inspection and maintenance regimes by discovering patterns based on various inputs and then testing potential impacts through use of simulation.

By using our own data harvesting software to search for useful information in big data sets, we can also extract information from previously unusable or disparate data sources, sort them into one central register or database and then link related pieces together to provide a much clearer picture of an asset’s performance – quickly piecing together a complex puzzle if you will.

This helps to identify trends as well as gaps and inconsistencies, which can then be addressed to generate key data outputs across assets. By harnessing previously unused and ignored historical data we are improving equipment maintenance and inspection strategies and significantly reducing our customers’ OPEX costs, whilst, most importantly, improving overall risk management.

Visibility is key

To challenge, review and analyse all aspects of an asset – Imrandd can apply the following solutions to support engineering, commercial and integrity teams.

We can provide any of these solutions as stand-alone services, however when combined, our customers achieve a much more effective asset management programme, one that’s truly aligned with commercial, safety and operational considerations.

Perception Vs reality – effective data analysis does not cost the earth

While some still believe implementing advanced data analysis techniques on older assets is costly and a time-consuming process, this is not the case, in fact all though it sounds complex, when applied correctly, it makes complex tasks much simpler. Typically we help our customers utilise the value in information they already have, combining maintenance, inspection, integrity, engineering, document control and process data and eliminating the low value tasks of collecting, checking and cleansing this manually. By doing this we free up technical teams to focus on much higher value areas, supporting virtual collaboration and quicker access to information also.

Trust and accountability - the future of machine learning and AI

While we must accept that currently, AI and deep learning does have its limitations, processes are being refined, improved and advanced at a rapid pace. Trust in the process, trust in the outcomes, and trust in the results - these are the concerns facing our industry as AI becomes further imbedded into our everyday activities. To trust is to be vulnerable, and no person, or company wants vulnerability in their critical infrastructure.

So how do we combat this? How do we address concerns around AI provability and ensure the accuracy of machine learning outcomes? First, we have to accept the outcome of some AI systems simply isn’t provable. However, we can develop special purpose machine learning tools based on deep domain knowledge and intricate understanding of the industry they’re applied to. This way we can add control and checks to ensure accountability in the decision-making process.

At Imrandd we’re building complex training data sets and developing machine learning tools for specific purposes, creating systems that draw on our deep understanding of materials and process engineering, as well as the latest developments in data science. This enables us to devise solutions with provable outcomes, and in turn manage the apprehension some operators still have about the use of machine learning and data analysis.

Concerns around liability, accountability and provability must be addressed if systems are to be useable and adopted with confidence. However, these concerns can be overcome by using problem-specific techniques where processes can be explained, and the outcomes easily understood. This allows us to use latest data science thinking, applied to current and future challenges in the most reliable and cost-effective way.

“Trust in the process, trust in the outcomes, and trust in the results - these are the concerns facing our industry as AI becomes further imbedded into our everyday activities. To trust is to be vulnerable, and no person, or company wants vulnerability in their critical infrastructure.”

Read the latest issue of the OGV Energy magazine HERE.