AI spots what your sensors miss and your eyes can't see turning maintenance into money and downtime into data across coal operations


, , , , , , , , , , , , , , ,
, , ,
,
In an industry where every unscheduled shutdown translates into lost revenue, wasted resources and mounting frustration, one company is making a compelling case for using artificial intelligence to turn maintenance from a cost centre into a strategic advantage.
In a recent Austmine webinar titled “Data-Driven Maintenance Decisions: What We Learned from Global Deployments in Coal Operations”, Razor Labs' Co-Founder and Chief Technology Officer Michael Zolotov, alongside Vice President of Sales Andrew Kaushal, shared firsthand insights into how their AI-powered predictive maintenance platform, DataMind AI™, is transforming the way coal operations approach reliability, efficiency and safety.
A smarter way to see what’s coming
The premise behind DataMind AI™ is straightforward: collect data from multiple sensors - vibration, temperature, current, oil condition, pressure and visual feeds - and use machine learning to detect the early warning signs of equipment failure. But what sets Razor Labs apart is what happens next.
“We’re not just looking at vibration data and flagging thresholds,” explained Michael. “We’re diagnosing specific faults and their root causes - like identifying fluting in a bearing and tracing it back to an electrical grounding issue, not just telling you the bearing is failing.”
In one example from an Australian coal site, the system detected bearing fluting - a phenomenon caused by stray electrical currents damaging the bearing race. While traditional condition monitoring would have prompted a simple bearing replacement (which wouldn’t have resolved the underlying problem), DataMind AI™ instead pointed the site team to the real culprit: a grounding fault. Once corrected, not only did the bearing last longer, but future failures were prevented entirely.
“That’s what sensor fusion and deep diagnostics enable,” said Michael. “We're not just delaying failure - we’re breaking the cycle of recurring faults.”

From black boxes to business value
Andrew, who has spent years working with mine sites across Australia and Southeast Asia, noted that the shift from reactive to predictive maintenance isn’t just about fancy tech - it’s about outcomes.
“Every mining operation is under pressure to do more with less,” he said. “Especially in today’s coal market, where prices have softened, our customers want to maintain productivity while reducing costs. That’s exactly what DataMind AI™ is delivering.”
Across multiple deployments, Razor Labs has documented measurable improvements: reductions in unplanned downtime, fewer catastrophic equipment failures and significantly extended asset life. At one coal handling and preparation plant (CHPP), the platform paid for itself in just four months. In that period, only half the machinery had been brought online with the system, yet over $1 million in downtime and spare parts costs were avoided.
“The business case is solid,” Michael added. “You’re not just buying software - you’re buying peace of mind, reliability and longer life from your critical assets.”
Seeing the invisible
One of the most compelling elements of Razor Labs’ approach is its use of AI-enabled vision systems to monitor conveyor systems and detect visual faults that no vibration sensor could ever catch.
Using deep learning algorithms trained to analyse raw video feeds in real time, the system can identify out-of-spec ore, belt misalignments, tracking issues, splice clip degradation and even blockages forming under a material stream - all without human intervention.
In one case, oversized ore - black rocks on a black belt - was automatically measured down to the millimetre. When the system detected particles exceeding acceptable size ranges, it sent alerts to site teams, who then adjusted crusher gaps or repaired screens to prevent downstream blockages.
Michael explained, “We’re not just flagging anomalies - we’re giving operators enough lead time to act. Sometimes, that's a few hours. Other times, it's days. Either way, the downtime is avoided.”
Some sites have even gone a step further, integrating DataMind AI™ directly into SCADA systems. In extreme cases, the system can autonomously stop conveyors when it detects a major hazard, such as a 500-millimetre boulder or a significant belt rip, preventing more serious mechanical damage.
Diagnosis that learns
The platform’s real power lies in its ability to filter out operational noise, one of the biggest challenges in a mining environment. By synchronising load (via motor current), speed (via tachometers), and system context (from maintenance records), the platform ensures diagnostics aren’t clouded by normal production variability.
“This isn’t about throwing more data at the problem,” Michael said. “It’s about using the right data in the right context. Just like a doctor looks at your blood test and MRI together, our platform fuses data streams for more accurate conclusions.”
When a failure is corrected - say, a misaligned belt is realigned or a blocked chute is cleared - the system doesn’t stop there. It verifies whether the condition has actually improved. If not, it flags the team that further action may be needed.
“That closed-loop feedback is a game changer,” said Andrew. “It gives maintenance teams a real sense of control and confidence that their fixes are working.”

1. Mobile Asset Fleet – Monitor engine health, fuel efficiency, powertrain integrity, oil cooler, and braking systems.
2. Conveyors – Detect belt misalignment, splice wear, and ore flow disruptions.
3. Ore Sizing for Crusher & Mill Optimization – Identify oversized material and optimise feed for downstream processing.
4. Fixed Assets – Ensure reliable operation of stackers, reclaimers, and car dumpers.
5. Extraction – Monitor mill and crusher performance to prevent mechanical failures.
6. Flotation Cell Monitoring – Track changes in flow, froth stability, and mechanical efficiency for better recovery outcomes.
Case in point: Centrifuges and consequences
Michael also shared a striking example involving a centrifuge that was performing normally - until scheduled maintenance introduced a new fault. Post-maintenance, DataMind AI™ flagged severe imbalance, traced back to uncleaned residue in the drum.
“Without that early warning, the centrifuge would have continued operating under extreme stress,” Michael said. “That shortens lifespan dramatically. But because we caught it within a day, they cleaned it, rebalanced it, and it was back to optimal within 48 hours.”
It’s this kind of rapid response that prevents minor oversights from snowballing into capital write-offs.
A new chapter in maintenance strategy
Beyond conveyors and centrifuges, Razor Labs is applying similar diagnostic techniques to pumps, flotation cells and other critical infrastructure. By monitoring parameters like pump curves and flow pressures, the system can detect cavitation and impeller wear - issues that often elude vibration-only monitoring.
At one site, the platform revealed that pumps were being routinely replaced after 24 months based on policy - not actual condition. With DataMind AI™, those same pumps lasted five to nine months longer, with full performance verified. In an era of tight budgets and capital constraints, that kind of extension is pure gold.
“In some cases, there’s no downtime impact but there’s still major savings,” said Michael. “Just by extending useful life, we’re improving ROI on every asset.”
Looking ahead
While mining remains Razor Labs’ primary focus, the team is also extending its reach into energy, utilities and even oil and gas. Their hardware is already certified for explosive environments, and partnerships with groups like Siemens Energy are expanding their presence beyond resources.
But for now, coal is where the most dramatic transformations are taking place - and DataMind AI™ is helping operators see what used to be invisible.
As Andrew put it: “The shift from reactive to predictive maintenance isn’t theoretical anymore. It’s happening right now, and it’s delivering real results at scale.”