The factory that hears failure coming: How SKF is turning machine vibrations into a reliability business

A bearing costs a few hundred dollars. The line it sits inside can cost far more every minute it stops.
That asymmetry is the quiet economics of modern manufacturing. For India’s manufacturers, the stakes are rising in step with ambition. Make in India and the production-linked incentive schemes are pushing factories to scale faster, run tighter and compete on cost and quality against the world’s best. In that environment, a failed gearbox on a steel plant’s overhead crane or a seized bearing in a wind turbine is not a maintenance event. It is a P&L event.

This is the territory in which SKF, the Swedish bearings and rotating-equipment major, is making a strategic bet. Its Industrial business in India, now a separate entity following SKF India’s demerger of its automotive and industrial businesses, has outlined industrial investments of ₹800 to ₹950 crore through 2030, including a new plant in Pune by 2028. But the more interesting shift is in what SKF is selling alongside the steel.

Wuttipun Phukoed, the ISEA Service Head for SKF’s Industrial business, describes a move from supplying components and maintenance support towards selling reliability itself.

“Unplanned downtime remains one of the biggest hidden costs in industrial operations, making reliability a strategic priority for manufacturers focused on productivity, efficiency, and competitiveness,” Phukoed told Express Computer.

From preventive to predictive: the sensor on the shop floor
The core of the offer is condition monitoring: instrumenting critical rotating assets so that their health can be read continuously rather than inspected on a calendar. “At SKF, we are helping customers move from preventive to predictive maintenance through a combination of connected technologies, advanced analytics, AI-enabled diagnostics, and engineering expertise,” Phukoed said. A key product is the wireless Enlight Collect IMx-1, which tracks vibration and temperature on critical assets.

Adoption, he says, follows the economics of failure. “Adoption is strongest in industries such as mining, cement, steel, power generation, and other process industries, where reliability directly impacts output and profitability.” Renewable energy is catching up quickly, because “maximising uptime across distributed assets is critical.”

The obvious objection from any CIO is that plants already drown in data. Phukoed agrees, and that is precisely his point. “Today the real challenge isn’t collecting machine data, but turning it into timely, actionable intelligence,” he said. SKF’s equipment in India generates terabytes of signal data, which the company processes with an AI platform using automated machine learning to catch bearing damage, misalignment, lubrication problems and imbalance early.

The human layer matters here. At SKF’s Global Reliability Centre, AI helps specialists scan and prioritise thousands of vibration datasets from customers’ online monitoring systems, so experts spend their time on the machines that need them most. From the India centre alone, SKF says it is monitoring and sharing detailed analysis for more than 150,000 process-critical customer assets. And because the analytics run in the cloud, “customers don’t need to hire dedicated data scientists or invest in local hardware and software to benefit from it.”

Phukoed is careful to frame the technology as a force multiplier. “This is about empowering existing maintenance and reliability teams with AI, not replacing them, so they can predict failures before they happen.”

Prediction is not the finish line
The industry conversation has already moved beyond predicting failure to prescribing the fix. Phukoed says leading customers are making that leap. “The objective is no longer just to identify that a failure might occur, but to understand the optimal intervention, when it should be performed, and what business impact it will have.”

Fully autonomous maintenance, where systems trigger repairs with minimal human involvement, is another matter. “Truly autonomous maintenance … is still at an early stage for most industries,” he said. He sets out three conditions before it can become mainstream: “robust machine condition sensing and connectivity across critical assets, high-quality operational data integrated with maintenance workflows, and confidence in AI-driven recommendations built through proven reliability outcomes.”

The third condition is the telling one. It is not a technology gap. It is a trust gap, and trust is earned one avoided breakdown at a time.

What it looks like on the ground
Phukoed offered three examples, without naming the customers. In steel, SKF helped a leading manufacturer improve the reliability of extremely critical applications such as overhead cranes, converters, ladle turrets and gearboxes through sensor-based monitoring and expert analytics. In wind energy, it assessed the health of a major operator’s turbine fleet and set up a structured condition-monitoring programme across more than 100 turbines, paired with vibration analysis training. For a large industrial producer, it combined vibration analysis, thermography, motor current signature analysis and non-destructive testing into one condition-based programme.

The common thread, he said, is combining sensors, analytics, diagnostics and engineering expertise to “move customers from reactive maintenance to data-driven reliability.” There is a sustainability dividend as well: spotting progressive failures early lets customers “maximise asset life, minimise waste, and advance their circularity ambitions” through remanufacturing and life-extension.

The hard part is people, not pixels
If the technology is maturing, why does predictive maintenance still stall at the pilot stage in so many plants? Phukoed’s answer will resonate with any CIO who has watched a promising proof of concept die in the field.

Technically, scale demands standardised sensor deployment, reliable connectivity across sites, and analytics that flag the right issues “without overwhelming teams with false alarms.” But then he turns to the factor he thinks gets underestimated: “The bigger, often underestimated, challenge is organisational.”

“Maintenance teams used to fix what’s broken need to be trained to act on predictive insights, plan interventions around production schedules, and trust the data enough to change how they work,” he said. His prescription is a phased rollout that proves value on critical assets first, paired with a structured skills and change-management plan. It succeeds “far more reliably than trying to deploy technology across every site at once.”

SKF is also changing how it sells. “That’s why we are focusing on value proposition and performance-based contract rather than traditional sales approaches,” Phukoed said, an arrangement that puts the supplier’s commercial outcome next to the customer’s uptime.

An alert that goes nowhere is just noise
The integration question is where the CIO’s remit meets the reliability engineer’s. Phukoed insists that digital reliability “only creates value when it plugs into the systems our customers already run their operations on.” Sensor and condition data is designed to feed customers’ ERP, MES and plant automation environments, so that “a bearing health alert doesn’t just sit with a reliability engineer, it can inform a maintenance work order, a production schedule adjustment, or an asset-performance report at the plant level.”

The aim, he said, is “a genuinely unified view of asset performance,” in which reliability data is part of the same operational picture used to run production, plan maintenance and make investment decisions.

The next three to five years
Looking ahead, Phukoed sees three shifts. Intelligence will move to the edge, closer to the asset, for faster detection and less reliance on constant connectivity, which matters for remote assets like wind turbines. Systems will graduate from flagging to acting: “Systems are becoming less about flagging and more about acting,” with AI recommending and eventually initiating maintenance actions. And automated, cloud-delivered AI will bring advanced predictive maintenance within reach of mid-sized operations, not only the largest plants.

Then comes the caveat that separates a sober industrial player from an AI evangelist. “What won’t change is the foundation. AI is only as good as the engineering understanding, bearings, sealing, lubrication, behind it,” he said. The next phase, in his view, is “engineering and AI advancing together, not AI replacing engineering judgement.”

What CIOs must learn from the shop floor
The first lesson is to treat reliability as a business outcome, not a maintenance project. When a supplier is willing to be paid on uptime and performance, the conversation changes from what a sensor costs to what an hour of avoided downtime is worth. CIOs should insist that their plants can state that number, because without it no predictive programme can be judged.

The second lesson is that data without a workflow is decoration. The value in Phukoed’s account lies in the moment an alert becomes a work order and a schedule change. That makes the integration of condition data with ERP, MES and plant systems a core architecture decision, not an afterthought, and it makes data quality in maintenance records a strategic asset.

The third is to start narrow and earn trust. The plants that succeed prove value on the most critical assets, then widen. A rollout that attempts every site at once will generate more false alarms than faith. Trust in a model’s recommendation is the true gating factor for prescriptive and autonomous maintenance, and it can only be built on a track record.

The fourth is to invest in the people who will act on the signal. Technicians who have spent careers fixing what breaks must be trained, and given the authority, to plan work around what the data says. A dashboard nobody acts on is the most expensive display in the factory.

The fifth is perhaps the most important: do not let AI substitute for engineering knowledge. A model that flags an anomaly is only useful if someone understands the physics of the machine well enough to know what to do about it.

The bottom line
Every failure on a factory floor announces itself first. A bearing runs a little hotter. A vibration signature shifts by a fraction. For a hundred years those whispers went unheard until the day they became a shutdown, an incident report and a very bad quarter.

What has changed is that the whisper can now be captured, analysed and acted on before the line stops. The technology is no longer the scarce resource. The scarce resource is an organisation willing to believe the signal and change how it works because of it.

The machines have been telling the truth all along. The CIOs who win the next decade of Indian manufacturing will be the ones who finally build a company that listens.