Still, manufacturers are up to the challenge, and many are already developing AI-enabled controls and equipment that could result in numerous benefits for food retailers and contractors alike.

Current State

Unlike other markets — including the consumer product industry — AI in the commercial refrigeration industry is still in its infancy, said Michael Kellerman, director of electronic controllers and service business development at Danfoss. The trajectory is starting to change though, and the pace of AI development in commercial refrigeration is beginning to accelerate.

“Most manufacturers are advancing their equipment control intelligence but still defining how to leverage AI to provide value for their own business and for their customers,” he said. “Fundamentally, we are talking about gathering data from the equipment as well as external data sources, artificially analyzing that data, and then commanding the equipment to behave in a certain way.”

Finding that value for customers can be a challenge, as the most innovative or exciting AI case might not be the most practical for a food retailer, said Kellerman. Exchanging data is also continually evolving, and manufacturers have to decide what to do with all the data collected from building management systems, the cloud, and other equipment and sensors.

Data management is an issue, said Amrit Robbins, CEO and cofounder of Axiom Cloud, a software company that offers cloud-based analytics and AI refrigeration apps. In fact, he noted that one of the biggest barriers facing AI in the commercial refrigeration industry is just getting all of the data in a format and location where it can be acted upon.

“For decades, commercial refrigeration systems have been generating tons of data from thousands of sensors, which is used only for real-time operation and threshold alarming before simply being discarded without ever leaving the facility,” he said. “Because AI is already so widespread across adjacent industries, we believe it will only be a couple more years before we start seeing widespread use of AI in commercial refrigeration.”

Another reason that cloud analytics and AI haven’t caught on in commercial refrigeration, despite being proven technologies, is that some traditional technology companies may not understand the unique needs and constraints of the customers in this space, said Robbins. It has also historically been difficult to liberate the data locked in refrigeration controllers so that it can be used in cloud-based models.

While manufacturers recognize the potential of AI to deliver value in commercial refrigeration, food retailers and their servicing teams still have questions about AI’s role in their operations and are often hesitant to adopt it, said Charles Larkin, director of data and analytics, digital and connected technologies at Emerson Commercial and Residential Solutions. Proving the value of AI across a wide range of food retail applications will be necessary in order to remove these doubts.

“Many large food retail operations are already using AI in customer-focused areas of their businesses and have data science teams dedicated to personalizing their consumer loyalty programs,” he said. “However, very few have the refrigeration domain expertise or experience applying AI to critical facility systems, which can be significantly more complex and require a completely different knowledge base. At Emerson, one of the most important jobs we have is to provide this expertise and help demonstrate AI’s potential value to our customers.”

To that end, Emerson is engaging some interested customers in short-term, proof-of-concept trial periods, where the company demonstrates how its solution integrates with the operation and delivers the potential for long-term, continuous performance improvements. Once customers see for themselves how it works — where they’re making immediate gains and realize how quickly it offers a return on investment (ROI) — they’re much more interested in exploring a longer-term engagement, said Larkin.

Another reason for the slow uptake of AI is that commercial refrigeration manufacturing is largely fragmented and comprised mainly of small- and medium-sized, family-owned companies, said Dr. Krishna Vedala, IoT systems engineer at Minus Forty, a refrigeration equipment manufacturer. Many of these manufacturers lack the knowledge, resources, and IoT expertise to develop such complex systems.

“IoT systems development is a long process and typically quite expensive,” he said. “Additionally, such systems require ongoing monitoring, continual bug and glitch fixing, and updates and enhancements, as this technology is evolving rapidly. We expect within the next five years, all major players on the market will have some form of IoT systems with varying degrees of sophistication.”

On the Horizon

Vedala noted that AI and machine learning (ML) can be generally split into two categories: descriptive analytics and predictive/adaptive analytics. OEMs and users of large refrigeration systems have primarily adopted descriptive analytics through their control and monitoring systems as part of their operations.

“In terms of true AI and ML adoption and implementation through predictive and adaptive methods and analytics, the whole refrigeration industry seems to be behind the trend,” he said. “OEMs need to conduct better and comprehensive researches on actionable insights that come from AI/ML in the refrigeration industry and communicate such benefits to their customers.”

As for applying AI to commercial refrigeration systems, this can be broken down to three levels, said Kellerman. The first level is between the equipment and the retail customer; for example, when sensors detect customers approaching. The second level is between the equipment and the grocery store or restaurant, where the equipment is intelligently communicating with inputs via a BMS.

“The third level would be the cloud and external inputs via the internet, such as receiving equipment commands and information from a cloud platform, or inputs like weather and traffic data, electric utility signals, and store POS data,” he said. “All of these inputs could come together to better control how the equipment operates.”

Ultimately, the core of AI technology will reside in the system control devices, which are typically incorporated into the equipment itself, said Larkin. By capturing data from sensors, modern equipment controls can perform a variety of key system optimization functions, from system fault protection and diagnostics to performance management and event scheduling.

“It’s important to understand that in many cases, we can enable these capabilities without having to perform a significant retrofit,” he said. “Many of our existing customers already have a data-rich infrastructure — including sensors, controls and modems — that we can tap into and begin delivering ML insights. In some cases, we recommend adding sensors, but that is relatively inexpensive compared to a full retrofit solution.”

Benefits

As for the advantages that AI offers, not only does it deliver potential reliability and longevity benefits to commercial refrigeration equipment, it also addresses a diverse range of concerns for store operators and contractors alike, said Larkin.

“For operators, we’re building data models based on their unique requirements — such as case types and perishable food categories — to help them optimize food quality and food safety and reduce waste,” he said. “We’re also developing ML algorithms to detect asset health or condition issues; these allow retailers and their contractors to begin implementing more predictive maintenance programs. This will help to reduce energy costs by keeping assets running in optimum condition.”

Reducing energy costs is extremely important, said Robbins, so for an energy manager, the main benefits of AI will be reducing energy consumption with more intelligent operation; reducing expensive peak demand spikes by predictively shifting refrigeration loads in time; and unlocking the hidden flexibility of low-temperature systems to earn money from demand response programs.

“For facilities managers, the main benefits of AI will likely be reducing the headaches of high temperature alarms and unplanned outages; saving money on monthly maintenance bills; or substantially reducing refrigerant leak rates without installing more PPM sensors,” he said. “For contractors, AI can be used to provide higher levels of service through root cause analysis of issues in real-time, intelligent technician dispatching, and data validation that shows whether an issue was fixed correctly the first time around.”

Energy savings, increasing the longevity of the equipment, and minimizing downtime and repair costs are all significant benefits of AI, said Kellerman. “Another benefit will be a better experience with the retail customer, whether it is interactive marketing or more innovative presentation of high-quality food products.”

In addition to the benefits mentioned above, AI provides the missing tool required to improve efficiencies throughout the manufacturing process, said Vedala.

“The insights generated from AI can be utilized to optimize the efficiencies in procurement of raw materials, distribution of labor, targeted marketing, and sales strategies,” he said. “The cost involved in maintaining such a valuable resource is quite minimal when compared to the costs incurred through the industry. It promotes collaborations with sharing of knowledge and information and providing holistic solutions in micro market and vending industries. There is no doubt that AI will bring significant benefits to the commercial refrigeration industry.”

AI Solutions Available Today

Many manufacturers already offer AI-enabled products and/or smart solutions, some of which are listed below.

(Courtesy of Axiom Cloud)

Axiom Cloud currently offers “apps for refrigeration,” which utilize cloud-based analytics and AI to supercharge existing commercial refrigeration systems. Using proprietary read/write integrations with customers’ existing refrigeration controllers, the apps can ingest the data from every sensor and system component throughout the facility. The cloud-based software combines that data with utility rate schedules, weather predictions, grid signals, and other external data streams to create a highly accurate “digital twin” model of a facility’s refrigeration systems. This model allows the apps to do things like automatically reduce a building’s peak energy demand, predict equipment failures before they happen, identify more refrigerant leaks faster, and reduce energy consumption.

Danfoss offers solutions for more intelligent control of commercial refrigeration equipment, including BMS that intelligently control multiple pieces of refrigeration equipment in a retail store. Danfoss ALSENSE® is a cloud platform that connects to the supermarket retail store and equipment to exchange information and leverage AI and that data in the cloud. Since more and more AI is happening in the cloud, Danfoss is focused on getting equipment connected to other sensors, equipment, and the internet.

Emerson is working with many customers in the food retail, foodservice, marine shipping, transportation, and logistics sectors, delivering solutions that incorporate ML and AI to optimize key aspects of their operations. Their solutions utilize sensors that provide data to powerful control devices — such as the new Lumity™ E3 supervisory control — and integrate with advanced, cloud-based software that leverage AI and ML algorithms. Emerson relies on the deep domain expertise of their refrigeration engineers to create data models that optimize refrigeration performance and help to achieve a variety of customers’ operational objectives.

(Courtesy of Minus Forty)

Minus Forty QBD Corp. already has an Internet of Thing (IoT) platform comprised of SmartConnect freezers and refrigerators, gateway hardware, web-based portal for customers’ use with OEM’s backend monitoring and control, application programming interfaces (APIs) for machine-to machine communication (M2M), and related software. Among other features, their system offers: full operational monitoring of cabinet temperatures; remote cooler locking and unlocking; remote parameter changes such as set temperature, events logging, and alarms logging; and notifications via emails and/or Short Messaging Services (SMS). Further development efforts are concentrated on predictive analytics such as remaining time before compressor failure and other critical components’ remaining operational life.