Automated 3D non-uniform volumetric primal box packing for Beef – Proof of concept and Prototype build

Summary

​Observations have shown the beef box packing process is very labour intensive as a result the packing rooms are often congested and disorganised. This often leads to incorrect products being packed or boxes being labelled mistakenly.  The process also requires staff to lift large pieces of meat over long periods of the day leading to health and safety risks. The use of automation in this process offers attractive returns for the processor.
An automated packing machine has the requirement to identify the correct primal items, pick up this product and pack into the correct box. Key to the effectiveness of this system will be to optimise how the products are packed to ensure the least number of boxes are consumed and distribution costs are minimised.

An investigation of previous work and visits to processing sites has allowed engineers to develop some raw concepts of what an automated packing machine could look like.  A set of primary components were identified; these include a vision system to classify the primal product, encoded conveying to allow for tracking of the product, a gripping tool to pick and place the product and a robotic arm to transfer the primal into the box.

In the initial phase of this project a site was chosen to trial the collection of vision data and observe the current packing process to understand the challenges for machine design and operation.

A requirement of this project, was to identify a single product that would be used for the proof of concept trial. The short-loin primal was chosen, as it is produced in significant production volume, relatively easy to recognise and is of a shape that was challenging to pick and pack using automation.

A gripping mechanism has been designed that is compatible with the environment and provides a robust mechanism for picking accentuated shapes, such as the short-loin product. The transfer from conveyor to box is conducted by a 6 axis robot that provides the required capacity and reach for this process.

Vision systems have been built with two key tasks. Firstly to capture the external surface geometrical information that will provide coordinates to allow tracking and robotic picking and optimised packing of the product.  Secondly, to capture images and subsequent vision analysis that can identify and classify the primal product so it can be packed into the correct box.