Artificial intelligence (AI) is quickly impacting just about every industry and transforming the world. This is true for those involved in the dairy business, as well.

For instance, AI-based quality inferential, model-based advanced control, and real-time optimization, along with vision and predictive maintenance model-based predictive insights, are used today in dairy processing equipment.

The use of AI in the food and beverage market is currently estimated to be worth $9.68 billion in 2024, and is expected to reach $48.99 billion by 2029, growing at a CAGR of 38.3%, according to Mordor Intelligence.

Jorge Izquierdo, vice president of market development for PMMI, The Association for Packaging and Processing Technologies, noted other uses of AI include speeding up the programming of robots, PLCs, vision inspection systems and camera training for quality defects in facilities.

“The most common benefits AI can offer are in the integration of disparate systems, reducing complexity, enhancing productivity and reliability, and empowering workforces with training and support,” Izquierdo said.

The benefits

Michael Tay, analytics platform manager for Milwaukee, Wis.-based Rockwell Automation, noted the benefits of AI include increased processing throughput/capacity (up to 10%), improved quality (reduced variability and give-away by up to 65%) and improved energy efficiency. Additionally, problem detection using vision AI can be 50-75% faster and maintenance spend per mass product produced is reduced up to 35%.

“A broad set of exploration and drying operations have increased throughput over 10% on average across tens of lines by different processors,” Tay said. “These same customers have reduced over-drying by moving their powder or concentrate averages 50% toward their specification with equivalent or negligible off-spec, and as a result of both of these improvements, reduced energy spend per ton produced by between 5-10%.”

These achievements, he noted, were made with industry-validated model-predictive control driving performance continuously and managing equipment and product constraints in real-time.

“Most of these applications are supported by real-time quality prediction models to control manual laboratory specifications with real-time estimates,” Tay said. “And follow on projects leverage the reduced product moisture variability to improve fat and protein giveaway by halfway to the product specification, with other components leveraged in nutraceutical products.”

In the same way cream cheese production (from composition blending, maturation scheduling and end product filtration) is driven to reduce feedstock costs per mass produced, and simultaneously reduce quality variability by 50% on average, the use of AI based machine sensor information is being used to monitor and track equipment stability to provide rapid predictive information about developing failures — such as centrifuge operations — to support managed maintenance and early detection of issues before they become critical or harmful to essential, moving equipment.

“One customer is leveraging high speed predictive models to forecast and adjust filling volumes on varying viscous jar filling so that mayonnaise and cheese spreads are reducing overfills by over 50% with negligible underfilling,” Tay said. “Other products are under investigation.”

Sushil Verma, co-founder of Ever.Ag’s Austin Data Labs, noted AI integration has significantly advanced beyond just equipment to encompass entire dairy farming processes.

“Notably, AI-driven solutions at the farm level include optimizing feed compositions, tracking animal health and assessing readiness for fertilization using sophisticated sensor data,” he said. “These applications not only streamline operations but also enhance animal welfare and milk quality, directly impacting farm productivity and efficiency.”

Some of the key areas where AI has improved efficiency are cheese yield optimization, transportation optimization, and sales and operations planning.

“By applying AI to optimize milk hauling, grain and livestock hauling routes and delivery schedules, we’ve managed to reduce costs and improve service delivery,” Verma said. “In the realm of production, AI tools have been instrumental in optimizing cheese yield by better managing variables such as starter cultures and milk quality from different farm inputs.”

Mikael Bengtsson, Infor’s director of strategy, food and beverage industry, has seen AI make big inroads in the dairy industry.

“When we talk about AI, we need to consider there are different flavors of AI — the machine learning, the deep learning, the generative AI, and then there’s robotic process automation and there might be an AI component to that,” Bengtsson said. “Today, most customers are thinking of AI in terms of machine learning, and that could be on the farming level, manufacturing or other parts of the supply chain.”

One of the big aspects of machine learning is having sensors in different places, such as the manufacturing floor or on pieces of equipment, and measuring humidity, temperature, vibrations and numerous other factors important to the dairy products.

“Those sensors measure and gather tons of data, almost too much data, especially for humans to go through,” Bengtsson said. “That’s the beauty with the machines and AI, the technology can easily digest, consume and make sense out of that data. It can see how the equipment is working and then recommend the best ways to make improvements.”

On the farm level, Bengtsson said it’s a similar concept, with AI sensors measuring and analyzing important data in the fields and discovering trends and abnormalities and getting recommendations on improvements.

“That could be something reactive on the maintenance side, where something is broke and I go fix it,” Bengtsson said. “Now it can be predictive and you might see what’s going to happen before a problem arises.”

Improving productivity

One of Infor’s dairy customers, Amalthea, a goat cheese manufacturer in the Netherlands, has always collected data on its manufacturing process, but historically it took a while to analyze and make any substantial change

“We have implemented AI for them so they can now implement things in almost real-time,” Bengtsson said. “They can see what’s going on in the yield of the ingredients and adjust to improve the yield instantly, helping them get more, and it provides a more consistent product.”

Using AI, Bengtsson noted, helps improve efficiency for anyone involved in the dairy industry.

“When we talk about monitoring equipment and fixing things before they break, it obviously helps you become more efficient,” he said. “But also what happens is if you have a production line go down, you might waste half a tank, and you’re losing profit and adding to food waste, so it makes a big difference. Amalthea has seen that firsthand.”

In the dairy production process, many quality parameters are captured by manual operator sampling and transported to quality laboratories that provide results which are available periodically for plant operators.

“AI leverages a history of laboratory qualities from moisture, composition, color, solids, texture and viscosity to map models from real-time available temperature, pressures and flows, among other useful measurements, to provide a real-time predictive virtual application,” Tay said. “This supports real-time, closed loop quality control of previously intermittently available operator feedback or, at least, real-time predictive feedback. Off-specification production is reduced to negligible levels, and give-away is reduced by faster, immediate feedback designed to learn and adapt to equipment-produced quality.”

Making a difference

Rockwell has a lengthy history of success using AI in the dairy industry, which includes everything from improving raw milk processing, processing powdered or finished dairy products from milk, butter and cheeses, and derived products, including yogurt.

“AI is predominantly focused on no- to low-code solutions, targeting support for our regular OT clients and driving value in easier to apply packages, from our Virtual Online Analyzer applications to broader Soft Sensor solutions,” Tay said. “We also have our intelligent device-based AutoML predictive maintenance solution through more advanced and directed measurement-based predictive maintenance applications, and model-predictive control that’s been driving dairy performance broadly across the industry since 1996, through new applications just released.”

The company’s vision- or camera-based AI is one of its newer technologies in an area where high value, expectations and strong needs for simpler usability drives value and customer interest.

“Rockwell is focused on adaptive, easy to implement and maintain applications, including an intelligent device monitoring leveraging RA drives and known anomalies where the equipment drive is leveraged as moving or rotating equipment sensor,” Tay said. “RA maps its fit for purpose AI to intelligent modern drives, and [is capable of providing] pre-mapped failure alerts in hours.”

This also provides a simple to label, local or unique alert instance, so technicians can investigate an unidentified, new problem, make repairs and label alerted patterns for the next time.

Beyond maintenance, AI’s role extends to optimizing plant level operations. For instance, AI algorithms are used to link farm level data with product yields, which is crucial in managing the variables that influence cheese production efficiency and quality.

“AI technologies are also extensively used in monitoring and ensuring the quality of dairy products from milking to packaging,” Verma said. “Particularly, AI models link milk quality directly to the feeds and medications given to the animals, ensuring that only the highest quality milk is used in products like cheese, where yield optimization is critical.”

Data drives information, and leveraging process historian and laboratory information management systems as a data model of any integrated processing operation drives AI systems.

“For predictive maintenance applications, prebuilt applications that can be adapted online are simplest, but for more traditional measurement based predictive analytics, having useful history of data around equipment targets, including time to failure or repair and identified classified repairs needed, is extremely helpful,” Tay said. “Most AI is going to leverage available measurement captured in process historians or similar data archives or data lakes, but a data model and some process knowledge helps users filter information to relevant signals and avoid or eliminate spurious correlations, confounders and other data/AI challenges.”

Looking ahead

Many dairy experts think vision systems are going to expand into easier to use and reliable solutions as AI becomes more capable.

“Then the integration of a range of technologies (vision, with measurement, with intelligent device data) is going to be integrated into useful data lakes with a data model to enable targeted, new problems to be rapidly modeled and directed to improve human/AI collaboration and drive higher value by solving more complex challenges,” Tay said.

The area where AI will really make a difference for the dairy industry, Bengtsson noted, is with decision support.

“It’s not that AI does it for you, but it’s about collecting the data and making sense of it all, and providing help,” he said. “At the end of the day, that support can do a great deal in improving life for anyone involved with the dairy industry.”

Emerging AI technologies that promise to further revolutionize the dairy industry include more advanced machine learning models for real-time data processing and decision-making at the edge.

“These technologies will allow even more personalized and immediate adjustments in dairy farming and processing,” Verma said. “The challenges will likely revolve around scaling these technologies across varied farming environments and maintaining stringent data security measures to protect sensitive farm and production data.”