September 5, 2023
At Africa’s largest oil refinery, even a single day of unscheduled downtime can translate into millions of dollars lost. The Dangote Petroleum Refinery, a sprawling 650,000 barrels-per-day facility built at a cost of roughly $20 billion, is poised to transform Nigeria’s fuel supply. But achieving that promise hinges on an often overlooked factor: maintenance. Ensuring equipment reliability in a refinery on this scale is a high-stakes challenge; one that engineers like Gbenga Ajenifuja have tackled with data-driven precision.
Analysts have warned that for the Dangote Refinery to meet its targets, it “must be accompanied by a leap forward in refinery maintenance capability”. Nigeria’s past refineries suffered from poor upkeep, and stakeholders quietly wonder if this new private behemoth can break that cycle. As a Reliability & Inspection Engineer from 2021 to 2023, Ajenifuja was on the front lines of building that new maintenance culture. His mission: predict and prevent problems before they halt operations. “Our mandate was zero unplanned shutdowns,” he says, standing amid towering distillation units and maze-like piping. “We had to identify weaknesses early and act fast, because reactive fixes are just too costly here.”
Ajenifuja and his team embraced an approach rooted in data analytics and predictive maintenance. He implemented Cox proportional hazards modeling—a sophisticated statistical tool more commonly used in medical research—to analyze equipment failure patterns. By feeding in historical failure data and real-time sensor readings, the model could estimate the “survival” probability of critical components. In practical terms, it meant the refinery could anticipate which pumps, valves, or compressors were most likely to fail next and when. “Cox regression helped us move from reacting to failures to preventing them,” Ajenifuja explains. If a set of pumps showed a rising hazard rate, his team would schedule a preventive overhaul during a planned maintenance window rather than waiting for a breakdown. These predictive insights were invaluable in a facility where an unexpected outage of a single unit can cascade into a plant-wide shutdown.
But data alone wasn’t enough – it had to be translated into action on the ground. Each week, Ajenifuja convened cross-functional meetings with process engineers, operators, and maintenance crews to review the latest analytics. If the models flagged an elevated risk on, say, a high-pressure distillation column’s heat exchanger, inspectors were dispatched immediately. Often, their non-destructive testing (NDT) tools would confirm subtle signs of trouble: a hairline crack forming, or unusual vibration indicating an incipient bearing failure. By combining statistical forecasts with on-site inspections, the team created an early warning system to catch issues weeks or months before a catastrophic failure. In one instance, ultrasonic scans detected a tiny crack in a vital pipe weld that could have led to a major leak—a fix was made during a scheduled turnaround, averting an outage that might have cost millions in lost production. It’s a stark illustration of the stakes: a single undetected cracked pipe at a refinery once led to an unplanned shutdown costing $4 million.
Ajenifuja helped develop rigorous inspection routines using advanced NDT methods, from ultrasonic flaw detection on pipelines to radiographic imaging of critical welds, all aimed at catching defects without taking equipment out of service. “We shifted to preventive inspections, not just routine schedules,” he notes. Rather than relying on fixed calendar intervals, inspection frequency was dynamically adjusted based on risk. High-temperature reactors, for example, got extra scrutiny if data showed even minor deviations in performance. This risk-based strategy meant resources were focused where they were most needed.
Fostering close collaboration between departments. Mechanical engineers provided him with design data and material stress limits; operators shared anecdotal observations of equipment behavior; data scientists helped refine the predictive algorithms. “It wasn’t siloed – reliability touches every aspect of operations,” he says. That teamwork paid off when subtle signs of trouble emerged. In late 2022, a series of small pressure fluctuations in a critical processing unit set off alarms in the analytics model. Operators had barely noticed anything amiss, but Ajenifuja’s reliability group took it seriously. They worked with the operations team to perform a controlled slowdown and inspection. Sure enough, they discovered early-stage coking buildup in a heat exchanger that could have led to a clog. The unit was cleaned during a short maintenance stop, preventing a far more severe fouling incident.
Such stories underscore how the refinery’s management increasingly prioritized proactive maintenance. It’s a marked change from the old paradigm of running equipment to failure. “In the past, maintenance was like firefighting – wait for something to break, then scramble. We can’t afford that here,” Ajenifuja says. At Dangote’s refinery, the goal was to shrink that percentage significantly using continuous monitoring and analysis. Each avoided failure not only saved money but also kept the tight project timeline on track.
By mid-2023, the maintenance team reported a noticeable reduction in surprise equipment failures compared to the initial startup phase. While exact figures remain confidential, Ajenifuja hints that the frequency of critical incidents dropped markedly thanks to their interventions. Equally important, when inevitable scheduled overhauls did occur, the team had data to justify them, which helped management see maintenance not as a cost center but as a vital part of production.
For Ajenifuja, who has since moved on to new endeavors, the experience was transformative. “It felt like we were pioneers, proving that our local industry can adopt world-class maintenance practices,” he reflects. Standing under the shadow of the refinery’s colossal distillation tower, he and his colleagues demonstrated that data and diligent inspection could tame even the most complex industrial systems. The lights never went out, the oil kept flowing – and a new maintenance mindset took root, one Cox-model prediction at a time.
