Operating one of the largest and most complex oil refineries in the world, Motor Oil Group can’t afford any unexpected downtime.
The organization’s Corinth Refinery, with widespread ancillary plants and fuel distribution facilities exporting to more than 45 countries, forms the largest privately-owned industrial complex in Greece. Even a brief shutdown can heavily impact the Corinth Refinery’s production.
So, after suffering an outage, the organization realized it needed to look beyond standard preventative maintenance for its equipment. “We needed to do something over and above traditional methods,” explains Nick Giannakakis, CIO at Motor Oil.
Motor Oil required a solution that would accurately predict and address potential equipment problems before they occur – critical for the refinery to maximize production, improve safety and lower maintenance costs.
The answer lay in data-driven predictive maintenance. The proliferation of sensors and other data-gathering systems across its production line means companies like Motor Oil have access to more operational data than ever before. Using technologies like machine learning (ML) and predictive analytics, organizations can drill down into the data to extract insights to inform decision-making and drive operational improvements.
Motor Oil also wanted to ensure its maintenance workers and technicians had access to the data in real-time, with updates and automatic notifications to alert them to any problems.
“The value of machine learning models now is the computational power. The way that we can leverage all the data which is stored makes this time in history unique. All of us are in a position to leverage all this technology, and all this data,” Giannakakis tells SAPinsider.
SAP BTP delivers data-driven predictive maintenance
The company turned to the SAP Business Technology Platform (BTP) on SAP HANA Cloud and SAP Analytics Cloud, which was delivered by Accenture Greece.
“Usually, traditional manufacturing applications are well positioned on-premise or even at the edge,” says Giannakakis. “Our business was one of the first to put critical manufacturing applications in the cloud.”
The refinery’s maintenance department collaborated with SAP’s Data Science group on a data value workshop to identify opportunities to enhance existing maintenance processes through machine learning and predictive analytics.
A successful proof-of-concept project analyzed four years of data on pressure, temperature and vibration sensors from three compressors. The model was fed with the sensors’ alarm and trip thresholds to predict when these would be exceeded.
So in addition to the new data that the refinery’s sensors collect hourly, SAP HANA Cloud enables Motor Oil to use root-cause analysis algorithms from the SAP Predictive Analysis Library on abnormal events by utilizing existing data to explain them with more than 77 percent accuracy. This root-cause analysis provides up to 120 hours advance warning of any upcoming technical issues.
Next, Motor Oil collaborated with SAP and Accenture’s Applied Intelligence unit in Greece to implement an intelligent data application. The solution offers users intuitive dashboards where sensors and equipment can be monitored in real time.
Giannakakis stressed the importance of collaboration around reporting to ensure any alerts for abnormalities reach the right recipients in the most suitable way, allowing time for technicians to solve the problem before any kind of shutdown occurs.
“[We are] leveraging all the collaboration capabilities that we have now to help us reduce the time to reaction from a human factor standpoint,” he says.
“When you have an organization operating 24/7, you may have an early warning alert at a time that is not sufficient. So, you need to…reach the head engineer or engineer team that most probably at that time are doing an inspection in the field. The alert goes through the dashboard, through the collaboration engine that we have [to their] mobiles.”
Creating a holistic view of the refinery’s equipment
By having access to all the data about its critical equipment together, Motor Oil can now analyze the refinery’s operations, revealing hidden patterns about how different sensors affect each other. The platform also continuously improves as it keeps retraining in the view of new data, allowing predictions to become more accurate.
As abnormal events can now be predicted in advance, technicians are able to simply repair sensors instead of replacing them entirely, saving on maintenance costs. This is in addition to the massive cost savings of avoiding production shutdowns and creates a safer working environment for employees at the Corinth Refinery.
“We are continuously focusing on how we will leverage the technology and maximize the benefits,” says Giannakakis.
Moreover, the CIO says Motor Oil is now looking at where the company can extend the platform across other parts of the refinery’s business.
“We’ve had very interesting discussions of upgrading our estate, modernizing it…How [can we] expand this solution to more equipment? How will we expand the solution with additional intelligent models? Especially now that we possess experience from our users, from our use cases,” says Giannakakis.
Using a digital twin to build a hydrogen electrolyzer
However, predictive maintenance isn’t the only investment Motor Oil is making in Corinth Refinery’s digital transformation. Sustainability and decarbonization are key drivers of transformation for the oil and gas industry. Companies seek to increase their compliance with energy transition mandates and open up new value chains for carbon and hydrogen – and Motor Oil is no different.
The company is currently working with the EU on a new project to install a 30-MW electrolyzer to produce hydrogen for the Corinth Refinery. Motor Oil said the system, EPHYRA, will produce around 4,500 tons of renewable hydrogen to refinery processes and other users. It is scheduled to become operational in 2026.
“We’ll get a complete change of the nature of this plant,” says Giannakakis.
Motor Oil is building a full-scale digital twin as part of the process, which was specified as part of a deal to ensure funding by the EU. Again, machine learning means data collected over long periods of time can be used to create data-driven models to make predictions about the future behavior of the electrolyzer.
“This is machine learning by design. It will help us become more innovative,” reveals Giannakakis. “It’s one of the most aggressive plans that we have.”
Use cases beyond manufacturing
Alongside the EPHYRA project, Giannakakis and his team are continuing to collaborate with SAP on other ideas for BTP across the business.
“We are thinking of quite a few use cases beyond traditional manufacturing,” he explains. “We’re working with SAP on some of those use cases. We have been discussing how we can provide the capabilities of continuous intelligence, for example, in our retail business. We want to be able to identify potential changes before opportunities are lost.”
He adds: “The capabilities of embedded AI, provided by the platforms, is a key enabler for us, a key differentiation.”
