Artificial Intelligence in Food Industry

Application of Artificial Intelligence in Food Safety and Quality Management

Modern-day food safety management involves collecting and analyzing data from various sources, including temperature sensors, quality control devices, supply chain partners, etc., where integration allows the software to gather data automatically from these sources and centralize it in one location by seamlessly connecting with inventory management systems, point-of-sale systems, or other related platforms. Integration capabilities can streamline data sharing, reduce duplication of efforts, and enhance total efficiency, which reduces manual data entry errors and ensures that all valid data is available in real-time, where most of the AI-based modern software solutions offer real-time data exchange, thus providing up-to-date information across the management. Nonetheless, food safety regulations and standards are also continually evolving, where integration features with regulatory databases and authorities permit the automatic updation of compliance requirements so that the company remains in adherence to the latest regulations, reducing compliance-related risks.

 

Nevertheless, leveraging technology can significantly streamline food safety management processes, whereas automation tools and software solutions can assist in data collection, analysis, and real-time monitoring, facilitating quick and informed decision-making. Moreover, advanced technologies such as blockchain and IoT (Internet of Things) enable enhanced traceability, transparency, and accountability across the food supply chain, reducing the risk of fraud and ensuring the integrity of the products. Further, regular auditing and certification play a pivotal role in maintaining and continuously improving food safety management systems. Thus, independent third-party audits provide an objective assessment of a company's adherence to food safety standards, highlighting areas for improvement and ensuring compliance with regulatory requirements, where implementing a recognized food safety certification such as ISO 22000 can demonstrate the management’s commitment to food safety, which also can instill confidence in consumers, suppliers, and business partners.

 

The contemporary food industry is based on a holistic approach to organizing, analyzing, integrating, and generating conclusions that can help benchmark and track key performance indicators (KPIs) related to food safety and quality, whereas the application of machine learning and artificial intelligence (AI) can provide the conceptual tool to transform food safety and quality data management, driving it from a parallel and repetitive control base model toward a value-based food safety and quality system. Hence, AI can deliver from the need to have a feedback loop for existing food safety and quality programs as well as whether they are meeting the needs and expectations of the organization/regulatory quality assurance management and generate frequent analytical information that can be summarized in reports for the company senior management teams. Further, AI may provide a number-based system to justify investment in a company’s food safety and quality programs, redirect resources to where they have the most impact, and potentially lower food safety and quality costs while enhancing the delivery of superior food safety and quality control and management.

 

AI emphasizes the creation of intelligent machines that work and react similarly to humans, where the early adaptation of AI included speech recognition, facial recognition, biometrics, planning, and computer-based problem-solving. The AI algorithms can learn and improve their effectiveness similar to the learning process for human food safety and quality professionals, where structured data are fed into the computer systems and identified with a label or annotation to be recognizable to the algorithm’s data point during the AI learning process. Nevertheless, the algorithm starts to examine the input data and compare it with known data that the algorithm already has analyzed, as well as modify the results by receiving more data inputs or solving equations for the human operator, where the algorithms can learn from data that are either numerical, i.e., such as colony count or pH, or a statement, such as an auditor note creating an output such as a simple grouping of data with means, modes, etc., or more valuable outputs such as statistical probability or classification or categorization. Hence, the more data fed into the algorithm, the more learning, data outputs, probabilities, and classifications can be derived. In addition, AI’s reach and applicability are expanding rapidly, whereas AI is enabling computers to think and learn in a manner to similar humans. Thus, the use of AI as a tool for food safety and quality is in its infancy, but with extended effort and investment, it has the potential to emerge as a game changer, forming a structural foundation to incorporate facility, industry, and government data to form a complete picture of risks, vulnerabilities, and opportunities for improvement.

 

Further, the successful application of AI can be used to analyze vast amounts of data on previous violations related to food safety, including time since the last inspection, operation period, nearby garbage and sanitation complaints, 3-day high-temperature readings, nearby burglaries, and tobacco or alcohol licenses issued by Public Health agencies to help identity at-risk restaurants to form a collective picture of the potential for food safety or quality risks and violations. The algorithm output is also can designate high- and low-risk restaurants and public places using the AI system and following the indicated KPIs designated by the food safety expert for the AI system where the AI results and the software design may publicly available, which requires the input of a significant amount of data to allow the algorithm to learn and optimize its performance for an AI solution to be successful.

 

There are two categories of food safety and quality data, which are internal and external data, where external data such as product recalls, foodborne outbreaks, relevant electronic health records, and finished product testing help to form a more complete picture and are being actively collected by numerous nongovernmental and governmental organizations. On the other hand, internal data include sanitation verifications, pest control programs, internal audits, supplier verification, consumer complaints, hazard analysis critical control point and preventive control data, good manufacturing practice and supplier verification data, foodborne illnesses, and company-generated internal and external lab records.

 

The data collection and analysis need to consider both external and internal risk-related data on various areas of industrial and public activities. Regardless of all the critical controls the world has overtaken over the years, still the most commonly reported foodborne diseases are listeriosis, salmonellosis, campylobacteriosis, and illnesses triggered by Shiga toxin-producing strains of Escherichia coli throughout the developed as well as developing world, while other zoonotic (transmitted by animal) foodborne diseases such as brucellosis are a significant public health issue in developing countries, where Trichinellosis and echinococcosis are diseases caused by animal parasites in humans. Antimicrobial resistance, caused by increasing usage of antibiotics in animal feed, are human-created condition that is gaining importance, while persistent organic pollutants, acrylamide, pesticides, and dioxin represent public health risks that all are chemical contaminants and hazards in food due to human activities.

 

Determination of all possible indicators that might affect food safety and quality as well as determination of the best approaches to monitor those parameters and variables are the key to a better analysis. Thus, food safety and quality data in a manufacturing plant can be collected using automated sensors that feed data to computers, as well as by individuals, i.e., ATP readings, pH, temperature, and composition of the in-process product; metal detectors; and optical scanners as well as outside contract laboratory testing can provide data on pH, microbial load (environmental and possibly finished-product pathogen testing), allergen residues, and labeling compliance, which are also important sources of food safety data. In addition, the design of experiments and/or predictive modeling using AI and machine learning approaches can provide systematic approaches that can optimize the impact of input variables from food safety and quality data measurement with the results of the food safety system implementations (outcomes), where already available software applications can be used to conduct such experimental designs that incorporate practical and easy-to-understand and approaches. Further, the conversion of the data into actionable information is the most important part of the food safety and quality database implementation system, where the overall structure of the food safety and quality database incorporates all elements discussed to build the holistic approach to AI-based food safety. On the other hand, the challenges of using AI in food safety and quality programs are also very important as food safety data have tremendous diversity in format, type, and context, whereas merging big food safety and quality data into conventional databases is challenging and hard to implement.

 

Therefore, one of the major challenges for food safety professionals is that the people creating algorithms for food safety and quality purposes are not the food safety professionals, where both food safety professionals and computer specialists need to learn more about each other’s profession or employ a translator that understands the interests and professions of both. However, there are a number of other challenges such as industrial data release, data interpretation, and data inaccuracy are very common in the food industry. Access to data is challenging because, accessing confidential industrial food safety and quality records is strictly protected, where issues such as data ownership and the right to access or sell data to other AI and non-AI companies include potential security breach consequences related to the food safety data. Although all data are valuable, the data interpretation is further challenging as internal and external inspector and auditor notes are unstructured, making them difficult to interpret and process, whereas the variability and inconsistencies in the format and location of the data require additional effort to organize the data into standard structures, which is a waste of resources and drives up the operations costs to use the AI technology effectively. Data inaccuracy has a major impact on results, where all data must undergo some type of cleaning or verification testing to ensure they are accurate and representative of what they are measuring.

 

Regulatory requirements are another major area, where relatively new food safety laws and regulations in many countries allow government inspectors to have access to all food safety and quality data, including that generated as the result of using AI. On the other hand, liability is another major factor, where the liability of the AI-processed data is important. Hence, the use of AI technology creates a larger liability burden for a company related to its food safety and quality programs, and the impact related to liability insurance must also be included when companies agree to insure food manufacturing companies using AI as a tool. Further, trust another very important factor where the food manufacturer’s senior management should be willing to trust a food safety and quality-related prediction from a software algorithm instead of a human, which is also can be a very challenging issue.   

 

Considering the current trends and ongoing cost-inefficient factors, hybrid models of AI applications can be the best tools for improving food safety and quality programs for food manufacturers by using internal and external data points collected from many sources and then integrating and analyzing them to predict the likelihood of unfavorable food safety events. The application of AI with human instinct and experience from internal and external inspectors, auditors, and food safety and quality professionals will support the detection, preventive action, and identification of risk factors, recognizing that food safety and quality professionals and company senior management retain the ultimate responsibility for making the right food safety and quality decisions.

Reference:

https://www.foodprotection.org/files/food-protection-trends/nov-dec-20-tajkarimi.pdf
https://smartfoodsafe.com/essential-guide-to-food-safety-management-software/
https://www.digicomply.com/blog/food-safety-management
https://www.rapidmicrobiology.com/news/leveraging-data-analytics-for-enhanced-food-safety-and-quality-control-download-white-paper
https://www.carlisletechnology.com/blog/improved-food-safety-through-data-collection
https://www.specpage.com/food-safety-lims/