Inspecting mining conveyor belts is a hazardous, slow, and inconsistent job. These belts can stretch for kilometers through open-pit and underground mines, exposing inspection crews to acid dust, extreme heat, UV light, and uncomfortable working positions.
Traditionally, a worker has to walk the entire length of the belt and judge the condition of thousands of components by eye, a process made even harder by harsh conditions and the difficulty of remembering exactly which part was failing and where.
SK Godelius helps mining operators monitor and maintain these critical assets using robotics, computer vision, and multimodal sensing. For instance, using Ultralytics YOLO models, the company's robots can walk alongside conveyor belts and automatically detect components like rollers, flagging which ones are missing, stalled, or overheating.
Bringing robotics and AI to the world's toughest mines#
SK Godelius is an engineering and robotics company that has been operating since 2011, with offices across Chile, Canada, and Australia. It combines robotics, artificial intelligence, engineering, and deep mining domain expertise to build technology for difficult conditions, and over the years, that technology has changed how mines approach safety and efficiency.
Crucially, the company does more than write code; it designs and builds the machines itself: robots that carry their own sensors, cameras, processors, and onboard intelligence, capable of operating with little or no human supervision in open-pit and underground mines.
A central goal behind this work is to take people out of harm's way. Mining inspections often put workers in close contact with hazardous conditions, and SK Godelius aims to replace those high-risk manual tasks with autonomous systems that can do the job more safely and more consistently.
By pairing robotics with AI, the company gives mining operators a reliable, repeatable way to understand the condition of their equipment, turning what was once a dangerous manual chore into a standardized, data-driven process.
The hidden cost of manual conveyor inspections#
Conveyor belts are the backbone of a mine, moving ore continuously across distances that can stretch for kilometers. Keeping them running means inspecting them regularly, and traditionally, that has meant workers walking the full length of the line by hand.
In open-pit mines, that work happens under strong sun, dust, airborne acid, and long stretches in awkward positions. Underground, inspectors face different conditions such as confined spaces, low light, poor ventilation, and limited visibility.
In both settings, inspections place workers close to demanding and sometimes hazardous environments. The manual approach is also difficult to keep consistent.
A single person walking kilometers of belt can’t easily catch and recall every fault, which roller was overheating, which one had stopped, and where exactly the problem was. Different shifts may report different things, so the same belt can be assessed differently from one day to the next, making it hard to track issues or spot patterns over time.
That inconsistency carries a real cost. In a large mine, a failed conveyor can cost up to $100,000 an hour in lost production, and repairing a torn belt can take one or two full shifts. Without complete, consistent data, operators are often left responding to failures after they occur rather than preventing them.
Inspecting conveyor belts with Ultralytics YOLO#
To automate conveyor inspections, SK Godelius uses a robotic platform that travels alongside the line, capturing visual, thermal, and acoustic data as it moves.
Each sensor has a role: the acoustic analysis listens for trouble in the rollers before it surfaces as heat, the thermal cameras reveal which components are running hot, and a vision AI pipeline leveraging Ultralytics YOLO models interprets the visual data in real time.
Fig 1. An SK Godelius robot inspects a conveyor belt using Ultralytics YOLO (Source)
Of the more than 50 conveyor failure modes SK Godelius tracks, around 20 are visual (the ones addressed through the computer vision pipeline), while the remaining failure modes are caught through complementary sensing technologies. Since Ultralytics YOLO models support computer vision tasks like object detection, the system can identify the components and conditions that matter most on a conveyor line, making YOLO one core piece of a broader, multimodal inspection solution.
That means picking out individual rollers and flagging the ones that are missing or no longer turning, spotting material building up beneath the belt, and catching anything that falls outside the normal state of the line, such as a fence left open or an object or vehicle sitting where it shouldn’t be.
To keep detections reliable in these demanding conditions, SK Godelius began with Ultralytics YOLOv8 and has since moved to Ultralytics YOLO11. They leverage the medium-sized model variant, which offers a strong balance of speed and accuracy, and run it on the robot's onboard hardware at the edge rather than in the cloud. This means detection happens right there on the line.
Why choose Ultralytics YOLO models?#
For SK Godelius, the main advantage of Ultralytics YOLO models is accuracy. On a conveyor line, the system has to identify exactly which component is failing and where it is located, and YOLO provides the precise detections needed to pinpoint a specific part as the inspection moves along the line.
Fig 2. Ultralytics YOLO in deployment on mining conveyors. Image source: Godelius
Just as important is consistency. A single worker walking kilometers of belt can't reliably catch and remember every fault, but a YOLO-based vision model assesses the line the same way every time. That consistent detection feeds into SK Godelius's broader inspection platform, which turns it into standardized reporting, traceability, and maintenance workflows that support more dependable decisions.
The models also run efficiently at the edge, directly on the robot rather than in the cloud. This keeps detection fast and reliable at remote mine sites, where connectivity is limited and conditions are demanding, with heat, dust, and acid in the air.
Maintaining over 95% average weekly coverage with Ultralytics YOLO#
By automating conveyor inspections with an autonomous inspection solution powered by Ultralytics YOLO, SK Godelius has made them faster, safer, and far more consistent. When the system detects an issue, it ties the finding to a precise GPS location and sends the maintenance team a report that pinpoints where the problem sits, often within a meter, along with supporting data.
Replacing manual walks of the line also reduces how often personnel are exposed to hazardous environments, and standardized, repeatable detection means teams can act on problems before they escalate rather than responding to failures after they happen, helping avoid the costly downtime a stopped conveyor can cause in a large mine.
Fig 3. Monitoring a mining conveyor belt for defects and anomalies (Source)
These benefits are clear in one of SK Godelius's mining operations, where the autonomous conveyor belt inspection system has been deployed at more than 10 mine sites, both open-pit and underground, and has assessed more than 16 conveyors across operational campaigns, maintaining over 95% average weekly coverage.
Over that period, it has detected more than 3,400 anomalies, of which over 900 were classified as critical, giving teams early warning of the problems most likely to lead to failures. Overall, the result is an inspection process that is traceable, repeatable, and data-driven, helping mines improve maintenance prioritization, identify risks earlier, and keep critical assets running.
Building safer, smarter mining inspections#
As SK Godelius expands, the company is working to broaden what its systems can detect and the conditions they can handle. It is exploring ideas such as using 3D imaging and vibration analysis to gauge belt wear, X-rays to inspect belt components, and robots to clean beneath the belts, and it is keen to use the latest Ultralytics YOLO26 models to push detection accuracy and efficiency further.
By combining robotics, multimodal sensing, and Ultralytics YOLO, SK Godelius is turning conveyor inspection from a hazardous manual task into a standardized, predictive process that keeps critical assets running and people out of harm's way.
Interested in how computer vision can reshape your business? Explore our GitHub repository to see how Ultralytics YOLO models are transforming innovations like AI in manufacturing and computer vision in robotics. Find out more about our licensing options and start your journey toward smarter, more efficient automation today.
