Detecting bone much earlier in the process - ideally while it is still whole muscle - presents challenges.
Millions of pounds of meat are recalled every year due to foreign object contamination. For the poultry industry, there's the added complication of bone fragments as an additional foreign body on top of the standard metal and glass contaminant risks. Since poultry products take many forms - from whole turkeys to chicken nuggets and everything in between - there are a variety of inspection options during the production process. So how do you decide where to install contaminant detection equipment to provide the greatest security with minimum wastage?
For bone-in products, the approach is straightforward. Inspection should be done as late in the process as possible - ideally once the product is sealed, to prevent any further contaminants being introduced. By designating a high-risk, unpackaged area and a low-risk, packaged area, X-ray can act as a 'gatekeeper' to only allow compliant products through. Even in whole birds, metal contaminants on the scale of a single millimetre can be detected by X-ray and removed from the product flow.
For products expected to be bone-free, however, inspection decisions become more complicated. Inspection at the end of the production line, in a sealed container, is beneficial as it ensures no new contaminants can be introduced. However, if value has been added, then it might not be possible to rework rejected products.
It is better to detect bone much earlier in the process - ideally while it is still whole muscle. But this also presents challenges. The natural variation between birds, and the automated preparation of chicken breasts, means the resulting muscle is not uniform. The varying thickness and uneven presentation can make it difficult to detect bones, which are low density and hollow.
The fact that poultry is 'designed' to fly means the bones are light and hollow - and therefore difficult to detect. And different poultry products can provide different challenges. Fresh chicken breast fillets, for example, are likely to be from young birds that may lack full calcification of the bones. Meanwhile, older birds used for processed products such as pies and soups are likely to have been egg layers, with calcium removed from the bird's skeleton to form the eggshells. This often makes the bones porous, and they can appear as sharp shards - very different from the bone fragments sometimes found in fresh fillets.
The development of dual-energy X-ray inspection technology has addressed some of these issues. Dual-energy systems generate two images - one produced by high-energy X-rays and the other from low-energy X-rays. Since bone and muscle are made from different atomic constituents, they respond differently to the two energies. This allows the system to differentiate between variation in thickness and the presence of bone - leading to more sensitive detection. Low-energy X-rays are used for tissue detection, while higher-energy X-rays can identify regions of higher atomic number elements, such as the calcium in the bones and other contaminants. The technology can detect even tiny wishbones or rib and fan bones of around 0.2 inches.
If any unwanted bones are identified in a product, the item can be reworked at an early stage to minimise wastage. If the product is to be sold as whole chicken breasts, items given the all-clear can then be packaged and move from a high-risk to a low-risk area, via a second X-ray system to confirm the absence of metal, glass or other contaminants in the sealed product.
For processed products, is the meat is being passed through a pipeline, the predictable cross-section of the pipeline can be used to give a uniform X-ray inspection - to save having to use dual-energy technology. This uniform presentation allows for excellent sensitivity against bone and sub-millimetre sensitivity to metal contaminants. Additionally, the rejected product can be diverted into a mechanical separator, removing the bone fragment but preserving the protein - ensuring there is minimal waste.
If the product has passed through a grinder, the X-ray system can be configured to ignore bones below a certain size while still rejecting larger bones. This can provide useful feedback for whether the grinder is operating as expected. Further in the process, this material may be formed into nuggets or other shapes. Wide-format X-ray inspection lends itself to positioning directly after this, where the product is separated into individual items.
Inspection at this stage can verify individual item shape and mass - ensuring every nugget is exactly the same weight, for example - and check there has been clean removal from the mold, at the same time as detecting contaminants such as metal, glass and bone fragments. A multilane reject or air curtain allows for the removal of a single item. Contrast this with inspection once bagged, where a single reject results in the removal of dozens of items.
As with unprocessed products, a final inspection after packaging is also valuable to ensure no foreign objects have been introduced during the packaging process. This final step also provides the opportunity for a final count of the number of items leaving the production line. Unlike other foreign object detection systems, X-ray technology provides a product count. By comparing the product count at each step along the process, it is possible to determine where in the process losses are occurring. Batch reporting and other KPI data can be made available to ensure uptime, detection performance and production rates are meeting targets.
Until now, automated inspection systems have generally used rule-based algorithms, which require skilled software resources and substantial testing. However, with the advent of artificial intelligence (AI) and machine learning (ML), X-ray imaging can be combined with AI-based algorithms so that manufacturers can automate the training and defect detection processes with unparalleled accuracy.
AI algorithms, particularly those based on ML and computer vision, are trained to recognize patterns and anomalies within X-ray images. Once trained, these algorithms can rapidly analyze large volumes of images and flag potential defects, such as foreign bodies, locations, voids, misalignments, or other irregularities, which might otherwise go unnoticed. For example, the wish bone is generally near the top of the breast fillet, and is relatively dense, whereas the fan bone is lower down the fillet, is tapered and can be attached to cartilaginous material. This automation not only reduces human error but also speeds up the inspection process, resulting in higher throughput and lower production costs.
The effects of foreign object contamination can be far-reaching, with the associated disruption to operations and the costs of managing a poultry product recall often running into millions of dollars - not to mention the potential damage to a company's reputation. Balancing performance and quality assurance with capital costs and maximizing the ability to rework rejected material is key to ensuring efficient operations. It means choosing where to inspect in the process is just as important as choosing what equipment to use.