Volume 17 | Number 1 | Spring 2005 | Automation Issue

Project Spotlight
Sensor-based Cutting System for Deboning Poultry

Top: Debao Zhou, postdoctoral fellow, makes an adjustment to the prototype device. Bottom: close-up view of cutting motion.

The Challenge
Automating poultry deboning operations has been a goal of the industry for more than 15 years. Early attempts initially appeared to be successful until problems began surfacing relative to sustaining acceptable yield performance (not leaving too much meat on the bone) without also yielding unacceptable levels of bone fragments in the product. The challenge proved too difficult to resolve with the technology of the time, and many operations reverted back to manual deboning techniques. While the equipment developers have never given up on refining their designs and in recent years have made strong strides in resolving the performance limitations of their earlier systems, many poultry processing operations continue to rely on manual deboning today.

The primary challenge faced is adjusting the cutting function to deal with the natural size and shape variability typical in many flocks. Some commercial attempts that are proving successful focus on controlling the variability that each machine must deal with. Researchers at Georgia Tech are taking a slightly different approach and are trying to develop a “smart” processing machine that uses vision and other sensors to recognize the size and shape differences of each carcass and then use that information to tailor the cut made on that carcass in a way that optimizes yield and reduces the risk of bone fragments. Initial study efforts have focused on the scapula cut, which is part of the wing cut, because it directly affects the yield of the breast meat (commonly the highest valued portion of the carcass).

The Mechanics of the Scapula Cut
At a chicken’s shoulder, three bones – the coracoid, clavicle, and scapula – come together to form the shoulder support. The moving part, known as the humerus, is connected through ligaments and muscle to form the shoulder joint. The scapula is a flat bone with a three-dimensional curvature. It starts at the back lower part of the shoulder, and runs along the back of the body and along the outside of the ribs. In order to achieve high yield, the cutting tool is required not only to follow the bone, but also to follow its orientation.

How the System Works
The sensor-based cutting system is comprised of a vision system, a cone line, and a cutting system, which consists of a two-axis manipulator with an attached knife, a force torque sensor, and a computer controller. The vision system identifies the correct starting position for the cut. Then the cone line moves the bird along a predefined circular path. The cutting system then moves the knife to adjust to the particular geometry of the bird to make the required cut.

Initial Test Results
Initial laboratory tests focused on the ability of a robot to cut through the joint and follow the scapula while using force feedback. The test was performed without the aid of the vision system, thus ensuring that the robot started at the correct position every time. This test resulted in the robot being able to successfully cut through the joint and along the scapula bone approximately 70% of the time. The research team ultimately determined that the greatest source of error in this system was the robot’s inability to be controlled in real-time. To solve this, a new approach to the control system and the cutting method was undertaken. This work has led the team to design a prototype cell where the knife has two degrees of rotation and the cone has three degrees of freedom (two linear degrees of freedom and a single rotational degree of freedom). By developing the hardware in-house, the team has introduced the ability to have a much higher update rate for the device (3 times a second for the ABB robot versus 200 times a second with the current system), and allows for a much higher responsiveness in the system. This is critical for the system to be able to identify and respond to typical variations in the bone structure of each chicken. By intelligently identifying the meat and the scapula bone and its orientation, the deboning knife is controlled to keep in contact with the bone with its surface parallel to the bone plane. The current cutting device achieves a speed of 0.6 seconds per cut without cutting the bone.

Current Research Focus
Currently, the team is focusing on studying the science of cutting and on developing a novel circular knife for shoulder joint cutting. Preliminary experiments using the circular knife design show that the cutting speed can be improved to one joint per second with a clean cut, which means that the muscles and the ligaments between the joint are cut and the socket and ball of the shoulder joint are exposed. “The cutting theory that we are developing will help us distinguish between the meat, ligament, soft bone, and hard bone. It will also help us design the optimal knife shape and provide the optimal cutting force and cutting angle, especially for the circular knife,” explains Debao Zhou, a postdoctoral fellow working with the research team.

Future Directions
Work will continue on fine-tuning the cutting technique, with the goal of using these testing results to develop specifications for the design of a stand-alone prototype device. Researchers anticipate that this device could be used on new machines or retrofitted to existing machines, adding more capability to adapt to product variability and thereby improving yield. Furthermore, the team plans to apply the general approaches and principles of the final design to other meat products, such as beef, fish, pork, and turkey, all of which would have inherently similar natural variability to poultry products.