System Description

Oversize Detection

Minealytics utilises state-of-the-art object detection neural network models to accurately detect the size and location of oversize material real-time and feed it back to the control system as well as through a web dashboard to operations so blasting practices and crusher CSS settings can be optimised. In harsh, dusty environment (eg. above the jaw) specialised laser camera can be utilised to augment the vision data for processing.

Prevention and Control

Mass flow-based control is used uniformly for primary control of product discharge flow in a crushing circuit. While a number of feed forward strategies have been trialled (bin level, AF hydraulic system pressure, AF current, grizzly / crusher current ratio, crusher current etc.), they have all suffered from the deficiencies of the underlying model. Minealytics can provide more actionable insight about the load conditions than any of the listed process signals, which then can enhance the existing apron feeder throughput model and its control scheme.

Minealytics can integrate with the following subsystems within the PCS to improve Crusher utilisation:

1. Operator:

   • Reduce change of missed bridges

   • Reduce time to respond to bridges

   • Remove load of operators by monitoring the camera feeds

   • Provide preventive feedback about the chances of bridge formation

2. Mine / Blasting:

   • Oversize feedback to the mine for blasting optimisation

3. Control System:

   • Vibrating grizzly oversize loading feedback control

   • Crusher blocked interlock to apron feeder

   • Enhanced apron feeder control scheme

   • Closed loop control of the apron feed can be enhanced via the feedback of vibrating grizzly load, which would substantially alleviate the chance of puzzling and overloading bridges in the primary crusher.

Minealytics offers real-time, machine learning models that can predict the chance of a bridge forming based on the combination of visual and PCS data (e.g. crusher current, apron feeder hydraulic system pressure, ROM bin level, etc.).

Recovery

Minealytics offers an integrated, vendor-agnostic Autonomous Rock Breaking (ARB) algorithm that can position the rock breaker above detected bridges autonomously and has the capabilities to autonomously initiate rock breaking processes.

It has been demonstrated in a purpose-built rock breaker simulation environment that the autonomous positioning, training, and deployment of state-of-the-art rock breaker control algorithms are within the capabilities of the  Minealytics ARB. The Virtual Rock Breaking environment can model the dynamics of the machine and the environmental obstacles. The Minealytics ARB control algorithm detects obstacles and collisions in this 3D space that is then fed back to the training algorithms for devising inverse kinematic models for the rock breaker autonomously. The result is an adaptable, robust rock breaker algorithm that is specific to the environment and the machine and optimises itself over time.

Technology

Leveraging YOLO's real-time object detection capabilities, it's possible to develop a system that can rapidly and accurately identify instances of oversize in either primary feed or primary crushed ore. With its robust balance of speed and accuracy, YOLO can operate on conveyor belts or over crushers, hoppers or overlooking truck trays and loader buckets. The algorithm can localise oversize in real-time, allowing for prompt removal and reducing the likelihood of process downtime and downstream issues. Its versatility and adaptability mean our team could train the network to recognise oversize, making it a valuable tool for enhancing the efficiency and safety of mining operations.

The YOLO framework, noted for its effective balance of speed and accuracy, has seen multiple iterations, each building upon the previous ones to address limitations and enhance performance. This paper explores the key innovations, differences, and improvements across each version, from the original YOLOv1 to the latest YOLOv11. These improvements span various aspects including network design, loss function modifications, anchor box adaptations, and input resolution scaling. The paper also discusses the trade-offs between speed and accuracy that have marked the framework's development, underlining the importance of contextual and application-specific requirements when choosing the most suitable model.

DETRv2 benefits from being an end-to-end transformer-based model. It eliminates the need for anchor boxes and non-maximum suppression (NMS). This simplifies the pipeline and reduces error sources. DETRv2 leverages transformers to encode global relationships between objects in the image, which helps with more accurate detection of objects in complex scenes. Transformers allow the model to focus on both local and global interactions simultaneously. DETRv2 with deformable attention mechanisms can handle small objects and occlusions better, as it dynamically focuses on the most relevant parts of the feature maps, leading to better handling of challenging objects or cluttered environments. It generally handles larger images and resolutions better due to the attention mechanism’s ability to scale to larger contexts. DETRv2's transformer architecture is inherently more flexible and extendable for other tasks like segmentation and panoptic tasks. DETR models are well-suited to multi-task setups without significant architectural changes.