Making food safety smart, the neural way

Challenge

How often do you see cases of foodborne illness or food recall? A February 2021 report by a U.S. Congress subcommittee stated that arsenic and toxic metals were found in baby food, including organic brands. This raises the question of how food safety issues occur, despite the presence of national and international food regulatory authorities. In the example above, we can deduce that the grain-based cereals contained arsenic as they were exposed to water containing the chemical. The World Health Organization (WHO) estimated that 600 million people fall ill every year due to the consumption of unsafe food.

Today, as the population rises, food wastage has emerged as a major cause that hampers food security. According to a US-based study of pre-harvest losses of vegetables, more than half of the crops were not harvested on time, indicating a huge source of food wastage. According to the Food Waste Index Report 2021, each year, 17% of global food production is wasted, 26% from food services, and 13% from retail. The issue is so acute that the United Nations has stated that by 2030, it intends to reduce food wastage as part of its Sustainable Development Goals. Countries and organizations are working to this end by incorporating innovative approaches and technologies such as predictive supply chain analytics or institute collaborations to develop a profitable product from food waste.

CHALLENGES IN FOOD SAFETY MANAGEMENT

The food value chain spans farm production, collection and storage, food processing, food packaging, and retail and food services. Food safety-related challenges are spread across all these touchpoints and are different at each stage, as illustrated in Figure 1.

_{Figure 1: Challenges across food value chain}

The lack of knowledge of soil health forces producers to use inadequate agricultural inputs such as fertilizers and pesticides. In some cases, the dearth of knowledge leads to an overdose of these inputs, which leave chemical traces, creating health hazards when consumed. According to the National Academy of Agricultural Sciences, 10% of every 1,000 lakh metric tons (MT) of wheat is wasted in India due to improper storage, and losses are associated with spillage, rodent attacks, pilferage, and more. In late 2019, approximately 1,941 pounds of raw chicken produced by a Santa Clara, California-based food processor were recalled due to mislabeling by the US government’s Agriculture’s the Food Safety and Inspection Service (FSIS), which is part of the United States Department of Agriculture (USDA).

Such challenges are spread across the food value chain. While regulatory bodies across regions, such as the GLOBALG.A.P. (Europe) and the FSIS, have their standard operating procedures to keep food processing practices in check, these are not uniformly implemented due to complex reasons. For instance, while food processing companies need to follow the standard guidelines set by authorities like the GLOBALG.A.P. and the USFDA, they still may face food safety issues since the raw materials they use may not be produced keeping in mind food safety guidelines.

Regulatory bodies such as the GLOBALG.A.P. typically offer recommendations about farming practices, but do not state food processing standard operations. Currently, no regulatory body traces these challenges across the entire food processing value chain. As a result, many cases of food recall in North America have occurred due to the wrong choice of raw materials, rather than any issues with food processing operations.

NEURAL NETWORKS FOR INNOVATIVE FOOD SAFETY INTERVENTIONS

With rising complexities in the food value chain, the food manufacturing and distribution business has become more customer centric. Companies with a resilient business structure reduce variability and earnings sensitivity to external shocks, such as supply chain disruptions due to war. Having end-to-end visibility in the food network helps food companies and their stakeholders synchronize all activities and offers them a comprehensive understanding of their operations at all levels.

Technology is the lynchpin in such cases where machine learning (ML), artificial intelligence (AI), big data, internet of things (IoT), digital twin, and smart packaging lend neural traits to the entire food value chain. Neural networks provide organizations with the right tools to predict potential problems and, in some cases, offer preventive steps, as described below (also see Figure 2):

• AI and ML for food supply chain risk management: The use of digital technologies such as AI and ML helps farm-producing and storage companies enhance their knowledge of food hazards. These technologies are used in food production systems, storage systems, and food safety-related risk management systems. Integrating AI in the food supply chain helps these firms sort food, monitor personal hygiene during manhandling, monitor the equipment cleaning process, and identify food adulteration.
• Precision agriculture and big data: Precision agriculture data, fertilization history, transport and storage temperature, fumigation schedules, and more are used for predictive analytics solutions. Historical data and predictive analytics help firms reduce the risk of food hazards due to unwanted chemicals. Storage data helps retailers decide which products should go on the shelves first and what can be stored for later, thereby reducing food losses.
• Food digital twin: IoT-enabled multimodal sensing technology in digital twins helps food companies check food quality in real-time across the supply chain. Simulation models of the supply chain, along with handheld food diagnostic devices predict the shelf life of food. Besides, enterprises can use visual analysis solutions to monitor food quality.
• Smart packaging and traceability: Embedding micro-sensors or radio frequency identification (RFID) systems improve food traceability as well as stores real-time storage and transport environment data. Active packaging is popular among fruit supply chains to remove oxygen, thereby increasing shelf life. Further, research on nano-coating of foods to improve shelf life is in progress.
• Augmented reality (AR) and virtual reality (VR): AR and VR improve remote monitoring of existing plants which are useful for audit purposes. Training on hygiene, sanitization, and safety maintenance are done without impacting plant operations.
• Predictive data analytics: Predictive analytics of machine health and capabilities help firms with maintenance planning, which minimize production outages, thereby improving cost-efficiency. Analytics are also useful to predict dynamic demand fluctuations of the final product, manufacturing rate, and inventory. Consumer data analytics help food companies understand market preferences and demand, which in turn help them introduce new products to their customers and reduce food wastage with optimized supply chain planning. 
• Distributed ledger technology and blockchain: These systems enhance the transmission of information of food materials across the value chain and also provide real-time data on customer attitude. The systems have the potential to write smart contracts during the trading of food materials.
• Whole genome sequencing (WGS): This identifies and provides characteristics of micro-organisms quickly, with a high degree of precision. WGS tracks pathogens by reading their gene combinations, which identifies the source of contamination in case of an outbreak. Such information allows food companies to take measures across various touchpoints of contamination. Computational genomics, multi-scale modeling, and simulation offer insights into the structural dynamics of biomolecules, which may be used for the analysis of nano-particle-based edible food coating to increase the shelf life of food and to also research new food additives, flavors, and more.

_{Figure 2: A neural network for the food safety ecosystem}