KANSAS CITY, MISSOURI, US — Few things are more profoundly human than harvesting wheat and milling it into flour to make food that feeds millions of people every day. It has been this way for thousands of years, even as ever-advancing technology, a product of the creative human mind, has helped make the process more efficient, productive and healthy in recent decades.

Now, artificial intelligence (AI) is generating excitement as another step in the evolution of technology, and by extension, flour milling. Industry suppliers and observers see a great opportunity to put an exciting new tool in the hands of humans in a number of specific applications that, while not altering the best-practices and basics of milling, can continue to give rise to a more effective process that remains challenged by climate change and the nutritional needs of a growing population.

“I believe AI is a disruptive technology that will impact all aspects of our lives, such as the internet and cellphones (have),” said William Gonzalez, solutions architect for Bühler, Inc., Plymouth, Minnesota, US. “The milling industry should look forward to adopting AI and machine learning to make better use of its resources, become more profitable, and remain relevant in the market.” 

Customers of Italy-based Omas Industries have been expressing interest in the possibilities of AI, said Pietro Barbalarga, commercial director, though reactions to the emerging technology have varied.

“Many (customers) are excited about the possibilities AI offers, viewing it as a way to improve productivity and reduce costs,” Barbalarga said. “However, there are also concerns about initial costs, the complexity of implementation, and the need for specialized skills to manage these new technologies.”

Turkey is the world’s largest exporter of wheat flour, and the country’s Association of Milling Machinery Manufacturers (DESMÜD) has recognized its membership’s interest in AI, which it described as a prominent topic during January’s World Economic Forum meeting in Davos, Switzerland. 

“Today, research and development scientists are advancing at an astonishing pace in imitating learning, questioning and perception so that artificial intelligence technology is becoming more and more advanced,” said Zeki Demirtaşoglu, president of DESMÜD, in a statement on the topic he issued in January. “While some people think that these innovators will soon be able to develop systems that surpass the human capacity to learn or evaluate any subject, others remain skeptical because all cognitive activity is embedded in value judgments that are subject to human experience.”

Jeff Gwirtz is president of JAG Services Inc., a milling industry consultant based in Lawrence, Kansas, US, and a former milling science professor at Kansas State University. With decades of experience, Gwirtz has witnessed the increased use of technology in a variety of mill settings. 

He said he sees opportunity for AI, but some aspects, he thinks, are currently beyond the technology’s ability to fully comprehend the human interaction and wheat behavior within the milling process. As a professor, he took students into the old mill to show them manual processes so they could better appreciate what happens during the milling process, no matter how automated and cutting-edge the technology has become.

“I feel like there is a tremendous opportunity for AI, but it’s a matter of taking advantage of what’s being done with the automation and monitoring of machine information to looking at what’s happening actually to the product as it moves through the system, and until we can do both, I’m not sure how beneficial AI can be in terms of operating the milling process,” he said.

The nature of AI

While AI likely can provide an answer to the question of its own origin, it still requires information and data created by humans from which to draw its responses. 

According to international technology firm IBM: “AI in its simplest form is a field that combines computer science and robust datasets to enable problem-solving. It also encompasses subfields such as machine learning and deep learning. Generative AI refers to deep-learning models that can take raw data from multi-data sets and ‘learn’ to generate statistically probable outputs when prompted. At a high level, generative models encode a simplified representation of their training data and draw from it to create a new design or approach that is similar, but not identical, to the original data.”

The recent attention given to AI likely can be traced to the 2022 emergence of OpenAI’s ChatGPT, an AI system that uses a large-language model to engage users in humanlike conversation and churn out topical written documents, which seemed to jump-start global excitement and questions about its potential.

The concept of AI goes back several decades to the earliest computers, when programmers pondered the very idea that a machine could replicate human thought patterns and intuitively learn new skills or generate unique ideas. As the processing power of computers grew, so did expectations for AI technology.

AI takes an interdisciplinary approach to machine design and production, drawing on mathematics, computer science, linguistics, psychology, and more, Demirtaşoglu said.

“Artificial intelligence is an intuitive imitation of human intelligence managed by computer software, and today these computer software codes, such as artificial intelligence, are cloud-based in every field from business applications to consumer applications,” Demirtaşoğlu said. “However, even ChatGPT, the most well-known AI application today, benefits from only a very limited portion of AI.”

Artificial intelligence can be described as a technology that enables machines to simulate human capabilities, Barbalarga said. 

“AI can analyze data, recognize patterns, make predictions, and make decisions autonomously or semi-autonomously,” he said. “AI differs from traditional predictive systems because it goes beyond making forecasts based on predefined statistical models. While predictive systems can anticipate future trends and behaviors based on historical data, AI can adapt and improve its capabilities over time by learning from new data and experiences. This makes AI particularly valuable in complex and constantly changing environments where parameters can shift rapidly.”

Gonzalez described AI as the next evolutionary step in digital systems and computational aids that help us improve daily business operations.

“These systems process large amounts of data and learn from the patterns over time by employing algorithms and computational models that have been trained to help us make educated decisions to achieve higher productivity, lower operational costs, and better-quality control, leading to increased market competitiveness based on these insights,” he said.

He said the advantage of AI is its capacity to evolve on the fly as daily production faces multiple, even unrelated, challenges, while also fine-tuning long-term operations.

“AI is not exclusively tied to automation or intelligent systems; it can also benefit process improvements by analyzing the data collected in any form,” Gonzalez said.

Italy-based Ocrim, a supplier of milling plants, feed mills, and grain processing systems and turnkey solutions, scheduled a session on the topic of innovations and applications of AI during its annual event Wheat, Flour and … gathering in Cremona in September as it continues to hear more inquiries from customers.

“During the open days event, Ocrim will unveil some innovative application in this specific field. Our customers are very interested and it is a topic that we - at Ocrim - have been exploring for some time now, in order to enhance the application of the AI solution in the milling industry,” said Fabrizio Baccinelli, sales director at Ocrim.

In addition to the milling process, preventive maintenance and energy optimization within the mill are other areas to consider for AI, he said.

Credit: ©PARILOV – STOCK.ADOBE.COM

Rolling with AI

The machines are one thing, but the applications of those machines in the use of different raw products is another level. The reactions of raw materials at each step toward the desired end products require millers to make precise adjustments in the process. With so many inputs at so many places, the tiniest change can have profound consequences on the flour quality characteristics.

Omas’ Barbalarga said AI can offer significant advantages.

• Process optimization: enhancing energy efficiency, reducing waste, and optimizing the use of raw materials through continuous analysis of production data.
• Predictive maintenance: predicting and preventing machine breakdowns, reducing downtime and maintenance costs.
• Quality control: detecting real-time anomalies in product quality, ensuring consistent and standard-compliant production.

“There is a growing interest in better understanding AI and its applications, and many customers are seeking information and training to fully grasp how to leverage this technology,” he said. “In my opinion, it’s important to highlight that the role of the miller is becoming increasingly valuable yet scarce, with an emerging need to digitize processes. This shift is particularly important in attracting new generations to the field, as digital tools and automation can make the profession more accessible and appealing.”

One initiative of Omas has been the Omas Milling Pilot, a system designed to explore and implement cutting-edge technologies in the milling process. This system represents a step toward AI in mill control, automating the regulation of rollers and temperature, according to Omas.

“This system helps bridge the gap between traditional milling practices and the digital future, ensuring that customers benefit from the latest advancements in the industry,” Barbalarga said.

Ideally, AI would be specialized to perform a specific operation, Demirtaşoğlu of DESMÜD said. A subset of AI, machine learning in which computer programs spontaneously learn and apply new data without any human guidance. Such deep learning techniques enable the automatic learning of large amounts of ungrouped data such as text, images, graphics or video. 

“As technology advances, previous criteria for defining artificial intelligence are becoming obsolete,” he said. “For example, the work done by machines that compute basic functions or recognize text through optical character recognition is no longer considered a function of artificial intelligence, but a normal computer function.”

The stability of the wheat milling process over thousands of years, despite its complexity, tends to provide a comfort level that is appealing to millers and their customers, even as technology has added to its efficiency and ability to refine outcomes and nutritional values. AI might allow the process to more quickly adapt to unforeseen challenges.

Automation has helped milling operations become more effective and streamlined, said Gonzalez of Bühler, which has its global headquarters in Uzwil, Switzerland. However, automation doesn’t provide analytics, and the limited solutions in the market are ineffective in providing recommendations, benchmarking, comparisons, etc. This type of intelligence is critical to improve operations. Many points during production gather data but are not effective at analyzing this data, or there is no analytics at all.

“With the challenges that the market is facing and the small margins from the industry, it is critical to have effective analytic solutions that help our customers improve and remain competitive in the market,” Gonzalez said. “I believe AI will bring the edge needed to the operations and aid in improving production.”

Gonzalez said most of Bühler’s work with AI is for specific cases from customers trying to identify an issue or improve a process, and the clearest example is with its Sortex machines.

“The image processing models have been trained to make a more effective selection of the produce that needs to be discarded, like grain affected by pests, non-wheat grains, rocks, etc. Bühler’s latest Sortex machines have leapt into better product selection thanks to AI and machine learning,” he said. 

Gwirtz, of JAG services, emphasized the human factor within milling, even as technology has evolved because so many conditions are present within the entire value chain: hard versus soft wheat, geography, climate, cultivars, etc. Wheat has its own personality that must be understood as much as the machine in terms of the final product. Combining machine and wheat optimization should be the goal. 

“Wheat from one state isn’t going to mill the same as wheat from another state,” Gwirtz said. “You can find lots of things that will tell you what the objective is, but maybe not in terms that you could turn into data in an automation program or AI program so it could provide a solution. How do you explain the feel (of a particular flour)? While we can take a lot of measures, we (millers) still have to look at and feel to get that end result, and there’s not yet sensors out there that can do that.”

Focus on machine performance and data is important, Gwirtz said, but how the machine impacts the wheat as well as characteristics of the wheat and how it behaves throughout the process are equally valuable.

“So, it’s at an information deficit (focusing on machine performance), and that’s where I think AI, in terms of running the milling process, is right now at this point — an information deficit,” Gwirtz said. “And the only information that be obtained, that a number of people have done with their automation, is machine learning. The machine temperature, the machine pressure, the machine amps, the height of the product, the rpm of the roller mill, the rpm of the feeder — those are very discreet things that can easily be measured, but we don’t tell you what happened to the product when it ran through the roll at this roll gap.”

He said the physical conditions and design of the mill itself also will necessarily impact the intellectual decisions made by any AI technology.

“That’s the thing about flour mills: The technology is there, but we can’t use the technology in all the different places in the mill,” he said.

Looking ahead

Demirtaşoğlu expects AI will continue to evolve and make itself more useful to milling applications, including crops other than wheat.

“Artificial intelligence is being used more and more today,” he said. “High-quality products from milling can be achieved through intelligent monitoring, diagnostics, real-time evaluation, production optimization, raw material management, machine learning and the use of powerful data tools. By incorporating artificial intelligence-based algorithms, the complex operations of the milling process can be performed with excellent results.”

Omas recognizes AI as a significant opportunity for the milling industry. For Omas’ customers, Barbalarga said, this means relying on systems that enhance efficiency, optimize processes and reduce operational error.

“Omas sees AI not only as a future means to optimize existing processes but also as a way to innovate and create solutions that can address future challenges,” Barbalarga said. “The company is committed to supporting its customers in the transition to an increasingly digital industry, providing the knowledge and tools necessary to fully harness the potential of this technology.”

Gwirtz thinks a particular area of the value chain could show immediate benefit from AI-powered solutions.

“Where I think AI may come into play is making a difference in flour sales, doing a better job of analyzing which of my mills is best suited to serve this customer,” he said. “Take advantage of learning all of what I call the upper-level data that’s available from regular accounting systems and things like that. For example, how do you take all of these plants that you own and evaluate them for providing flour into a specific bakery and at a particular location? And that’s a real industry analysis kind of thing, because you’ve got to know what wheat you need at what mill.”

While AI is a great technology with much potential, other aspects need to be in place before this, such as the network and electrical infrastructure, automation systems, and the right milling equipment to get the maximum benefit from AI, Gonzalez said. AI is not exclusively tied to automation or intelligent systems; it can also benefit process improvements by analyzing the data collected in any form. 

Process improvement, packaging, raw materials, cleaning, issue identification, and machine idling are potential areas of AI impact. However, since each customer’s production varies widely, customization and partnership are the key points to making the right implementation and investments. Gonzalez said Bühler has a team of engineers dedicated to AI and machine learning. 

“I believe it is not a question of if but when,” Gonzalez said. “However, proper adoption is critical to avoid wasting resources with ineffective solutions, so it is critical to seek trusted providers with the right focus and expertise.”