The Challenge
Hard rock quarries process material through large crushers to create aggregates for road and construction use. This crushing process involves multiple sequential stages, each significantly energy-intensive and highly sensitive to variations in input rock size and crusher configuration. Incorrect crusher settings and mismatched input rock sizes cause inefficiencies and substantial energy waste.
Continuously measuring rock sizes at high production speeds and promptly detecting processing downtimes are persistent operational challenges. When crushers run but aren't actually processing material, energy is wasted entirely. These inefficiencies result in significant financial losses—potentially up to £200,000 annually—and substantial unnecessary CO2 emissions. Without real-time visibility into what's happening on the conveyor belts, operators can't respond quickly enough to optimise the process.
Our Approach
Working with Tarmac as part of their Emerald Challenge-winning initiative, we developed a lightweight deep neural network that analyses high-speed images captured by cameras placed above crusher input and output conveyors. This solution accurately determines rock size distributions and detects processing downtimes instantly.
The computer vision system uses instance segmentation models to identify and measure individual rocks as they pass on the conveyor belt at production speeds. By processing imagery in real-time, the system provides continuous monitoring rather than periodic sampling—giving operators immediate visibility into what's actually flowing through the crushing stages.
Initially deployed via cloud computing, the solution has already enabled the identification and elimination of downtime representing 5% to 10% of total operational hours—periods during which crushers were running but not processing material. We are now transitioning the deep neural network onto edge devices to achieve comprehensive real-time monitoring, empowering operators to instantly identify inefficiencies and dynamically adjust crusher settings based on live rock-size distributions.
The Outcome
Eliminating previously invisible downtimes has significantly increased production throughput while substantially reducing wasted energy costs and associated CO2 emissions. This optimisation translates directly into enhanced productivity, substantial cost savings, and improved environmental performance.
The transition to edge deployment will further enhance capabilities, enabling operators to respond immediately to changing conditions and continuously optimise operational efficiency. Real-time rock size analysis allows dynamic adjustment of crusher settings—feeding smaller rocks allows for more aggressive crushing, while larger input requires different configurations.
This work formed part of Tarmac's Emerald Challenge-winning project, demonstrating how AI-driven process monitoring can deliver both economic and environmental benefits in heavy industry. The approach establishes a template for deploying computer vision in industrial settings where real-time feedback and autonomous monitoring are essential requirements.