The meat industry uses tools ranging from simple hand-held knives to complex mechanized systems, up to adaptive, sensor-driven automation.
Mechanization and automation bring operational, economic and social benefits to the meat sector. They help companies to cover the shortage of skilled labour, to improve the safety of operators engaged in objectively dangerous activities, resulting in high insurance costs for the employer.
The supply chain
Private and public companies and entities contributing to the meat supply chain belong to the agri-food and agro-industrial supply chains, i.e. to the set of activities, technologies and resources contributing to the creation, processing, distribution and marketing of a finished product. The production process in the strict sense comprises: Breeding, slaughtering, processing, transport, logistics, distribution and sale. Breeding methods vary according to animal species, the characteristics of the farm and the territory in which it operates.
Intensive (or battery) breeding has an industrial approach, aimed at maximising productivity. Animals are fed high nutritional value to enable them to grow rapidly; they are reared in confined spaces and constantly monitored to ensure high yields. More structured farms use technology to track production, costs and increase efficiency. Extensive breeding involves a limited number of animals per hectare, open areas, grazing animals except during the overnight stay in the stable and during the winter season, mainly natural feed consisting of grass and hay.
In both cases, the public service veterinarians check the health status of the herds and verify the provenance and health records of the individuals entering and leaving the herds. Once the animals have reached an established age and weight, they are sent to slaughter, which is also supervised by the veterinary services who check their state of health before slaughter and the hygienic-sanitary quality of the carcasses obtained. Undernourished or sick cattle are excluded from the food chain.
Slaughter
Slaughter is defined as the slaughtering of animals intended for human consumption and the operations related to it: Handling, stabling, restraining, stunning and bleeding. Stunning is the process that rapidly leads to a state of unconsciousness of the animal and lasts until death; killing induces death. Stunning can be done by electricity, mechanical systems or gas. Electronics can be used with all species; electric water baths are the preferred technique for poultry.
Mechanical stunning is achieved by penetrating captive projectile, a technique valid for all species; non-penetrating captive projectile used for the slaughter of ruminants up to 10 kilograms; head percussion used for piglets, kids, calves, rabbits, hares, large poultry; cervical displacement for poultry up to 5 kilograms. Stunning with gas (carbon dioxide or inert gas) is used for pigs and poultry. Stunning, stucking and bleeding must take place as soon as possible. Depending on the animal species, peeling/skinning/plucking/scalding-flaming is then carried out for removing the bristles.
Large animals undergo then to the removal of the extremities of the limbs and head, abdominal evisceration, opening of the sternum, thoracic evisceration, division into halves, health inspection, toileting, subdivision and classification of the carcasses. They are followed by cooling to quickly reach an internal temperature of 4°C to 7°C depending on the type of meat, tenderisation, quartering, boning and processing of quarters, cold storage, packaging, labelling for identification and traceability, shipping and delivery.
Mechanization and automation
While drones, sensors, artificial intelligence, robots, and 3D printing are now part of the poultry industry, implementing them in the red meat industry is more difficult. Due to the extreme variability of size and characteristics of slaughtered animals, most of the operations described above still involve human intervention, even if they are routine, repetitive, and rules-based activities. More structured slaughterhouses have recently identified some processes that can be automated, focusing on the most dangerous and strenuous ones and those subject to compliance checks.
In these cases, automation reduces the risk of injury and makes the steps taken transparent. The most promising applications are skinning and boning. In the beef sector, skinning is a very challenging phase due to the length of the animals (over 2 meters) and the need to cut the skin exactly along the center line of the carcass. Using an infrared laser distance sensor, the robotic skin stripper acquires information about the size and profile of the carcass to be processed. The processing of the data thus collected allows the cutting tool to follow a path as regular as possible.
After cutting, the skin is pulled down, separated from the dorsal fat using a blunt knife, and then pulled over the head and away from the carcass. Researchers are also experimenting with sensor-driven robotic systems for boning. The most promising studies involve removing the bones of the forelimbs of beef. The first step is visual evaluation of where the meat adheres to the bone. The image is compared to other images stored in a database; if the comparison is positive, boning occurs with the blade force parameters stored on previous occasions. If no image matches what is stored, the system acquires the processed image, sets new force and blade path parameters, and the data is stored for future reuse.
Assessing the meat quality
Farmers, processors and consumers judge the quality of meat in different ways. For the former, the quality is related to the percentage of musculature on which to base the request for payment; for the second, the suitability for processing and all the characteristics that influence the purchasing decisions of the consumer are important. The latter evaluates the appearance of the product (colour, lean/fat ratio), its other sensory characteristics (tenderness, succulence, aroma, colour, texture, elasticity, fibrosity, adhesiveness, density, firmness, cohesiveness, chewability, and farinosity), its nutritional value (macro and micro-nutrients), food safety (presence/absence of toxic compounds, drugs, pathogenic or deteriorating microflora), breeding and slaughtering techniques (animal welfare, sustainability).
Beef has a mild taste, but it varies according to breed, type of cut, age of the animal. Pigmeat has a more intense taste. Sheep meat has a stronger aftertaste. Better fed animals develop a fatter mass which in turn affects the taste of the meat. Beef is more fibrous than pork, which has a greater tendency to retain water. New technologies help to objectively evaluate the mix of these features to support pricing. Today, the most tested quality control technologies are electronic nose, tongue and eye, taken individually or interconnected via ANN (Artificial Neural Network), or CNN (Convolutional Neural Network). Near infrared spectroscopy and the latest biosensors also contribute to the evolution of the meat industry.
Technologies for non-destructive testing
The experiment focuses on the integration of innovative techniques for monitoring the processes and characteristics of meat cuts. The aim of the projects is to optimize the use of various digital instruments put into the system so that together they can assess compliance with quality and product standards defined by companies and contribute to the construction of databases that can be exploited at different stages of the meat supply chain. Instrumental investigations used techniques applicable directly online or in company laboratories. They allow information on the status of the product being processed to be acquired digitally in real time and with quantitative accuracy.
The most commonly used instruments are nose, tongue and electronic eye. The electronic nose comprises a sampler and a detector consisting of sensors that react to a wide range of odorous substances. When the sensors detect the presence of these chemicals, they determine the variation of one or more system parameters, for example, mass, oscillation frequency or electrical resistance. A device receives the data generated by the sensor transforming them into an electrical signal, which is analysed by the odour processing and recognition system, whereby the incoming compound is associated with a similar one stored in the device database.
The more known samples stored, the more likely it is to detect a particular substance accurately. In the meat sector, the electronic nose is theoretically usable to evaluate the state of preservation of the product, without interferences due to variations in characteristic odours of the meat itself, potentially different depending on age, breed, diet, state of health, and sex of the animal of origin. The challenge lies in being able to map the nature of the odorous compounds alone due to product alteration and in developing reliable validation systems. Electronic tongue is a multi-channel sensor that goes well beyond the five basic tastes. It consists of chemical detectors with a partial specificity to different components present in a liquid sample; it has an method of recognition of stored patterns and a multivariate calibration system for data processing.
Chemical detectors can be thin synthetic lipid/polymer membranes superimposed on a multi-sensory chip. They are not specific and each reacts, with different degrees, to all the components of the flavour. The results obtained from the individual detectors are cross-referenced, and an algorithm creates the gustatory profile of the analysed product. The profile is compared to a sample with known characteristics, and a selective qualitative and quantitative description of the substances contributing to the taste is obtained without distinguishing between its individual components. The progress of this technology is due to the development of biosensors with high selectivity and specificity. We then enter the field of research into biosensors, analytical devices, which convert a biological response into an electrical signal.
They work with two components: A bioreceptor that recognizes the target analyte and a transducer that converts the recognition into a measurable and quantifiable electrical signal. At the heart of the development of the electronic eye is computer vision technology that converts optical images into digital images. The unit that performs the function of a human eye is a sensor that collects images and uses computer simulation criteria to identify them. The process includes capturing images, processing them, extracting their characteristics, recognizing a pre-acquired pattern, and making decisions. The system can be interfaced with mechanical instrumentation, optics, electromagnetic sensing, colorimetry, spectrophotometry, digital video technology and image processing.
The use of the electronic eye does not involve product sampling, does not limit the size of the evaluated sample and the number of samples that can be evaluated in one step. Under standardized conditions, the instrument reliably and reproducibly measures the colour and shape of the objects being controlled, stores the results, and guarantees traceability. In the meat industry, the electronic eye is used for monitoring product deterioration, detecting foreign bodies, colour changes expected as a result of certain processes such as freezing, cooking, frying. On the slaughter lines for large animals, carcasses are classified according to the amount to be paid to the farmer.
The assessment is now carried out by experts with all the uncertainties of the case that arise due to errors, fatigue or bad faith. A vision system supported by ANN or CNN neural networks can easily ascertain the quality of the carcass and classify it according to established standards. The set of data from the electronic eye, tongue and nose processed by a single network provides a comprehensive response on the quality of the product.
Another technique for assessing the freshness of meat is near infrared spectroscopy. The main meat chromophores (myoglobin – red/purple, oxy-myoglobin – bright red, methamyoglobin – brown, nomyoglobin – bright red/ fuchsia, sulfomyoglobin and microbial contamination colemyoglobin – green, colemyoglobin, heat myochromogenic globin and denaturation – dead red), as well as fat, water, collagen are characterized by a typical absorption in the visible-near infrared (NIR) spectral region. Therefore, structural and compositional changes in meat can lead to differences in light absorption.
The main components of the resulting data can be analysed (PCA) to precisely define the degree of freshness of the product. The AI-based sorter of packaged meat uses X-rays, NIR spectroscopy, lasers, cameras and a machine learning algorithm to analyse different aspects of the packages. It is able to ascertain the quality and quantity of packaged meat and packaging anomalies. Out-of-standard packages are automatically rejected.
Plant Hygiene and Predictive Maintenance
Sanitation of environments and equipment is a priority for the meat industry. New automatic systems include sensors that detect non-removed particles. They operate with optical fluorescence imaging and ultrasound detection technologies, providing data to the machine learning algorithm that helps identify the presence of micro-organisms and even minimal food residues in the equipment.
The plants used in the meat industry are large and complex, and more and more companies are planning to adopt predictive maintenance systems. Timely repairs reduce maintenance time and costs. Companies are now accustomed to regularly carrying out scheduled preventive maintenance based on the machine’s life cycle, considering main parameter the time interval elapsed since the last maintenance performed. However, preventive maintenance has the limitation of using a very simple data model to provide a very broad recommendation that takes little account of the actual operating conditions of the equipment.
Predictive maintenance, on the other hand, continuously analyses the condition of the machine and connected equipment, gathering data useful for building models that reflect the state of the machine and its use. Data analysis leads to more accurate maintenance recommendations. In some cases, remote monitoring of complicated mechanisms takes place by creating a digital twin of the machine, deriving operating data and performance; thanks to machine learning it is possible to identify the factors that influence the quality of the production process, analyse the main causes of machine downtime, and eliminate the problem at source.
