Regulations and Leaks

For more than a decade, Robbins has been helping grocery chains and large-scale commercial refrigeration facilities optimize their refrigeration equipment through software technology, not hardware. He believes refrigeration is a large, overlooked, and underserved sector that can significantly be improved through the use of software and automation, which can yield substantial benefits for both the environment and the end user.

“The elephant in the room is refrigeration,” said Robbins. “On the energy side, refrigeration is consuming 60% to 70% of all the energy in retail grocery stores and even more for cold storage facilities, all while refrigerant costs continue to rise faster than inflation. On the maintenance side, it's estimated that up to 90% of store service calls are driven by refrigeration. In addition, refrigerant leaks are of growing importance, not only because of the cost implications of a leak, but also the business interruption that occurs when there's a big leak. Then there's also the compliance element, which is becoming a forefront issue for our industry.”

On the issue of compliance, Robbins noted that 15 states already have regulations (or proposed regulations) to lower refrigerant leak rates, and there are federal regulations on the horizon as well. Through the AIM Act, the Environmental Protection Agency (EPA) will soon require food retailers to reduce leak rates nationwide, as well as install automatic leak detection systems in many of their stores.

This is where AI can really shine, as the technology can help food retailers meet regulations by swiftly identifying leaks, thus enabling prompt repairs. According to Robbins, leaks are frequently unnoticed until the store has virtually run out of refrigerant entirely, leading to a crisis scenario. Technicians rush to the store to add refrigerant, but often lack the time to address the root cause of the leak.

To solve this problem, Axiom Cloud developed the Early Leak Detection (ELD) module, which uses AI technology to detect refrigerant leaks before they become catastrophic. The company recently closed on $5 million in funding, which is intended to accelerate the deployment of the ELD module. According to Robbins, the modules do not require any new hardware or sensors, as they use data from existing controllers and can be installed in hundreds of facilities quickly without any site visits. The company also offers its services under a simple subscription model, with modules available on an a la carte basis for a monthly fee per site.

“The ELD module utilizes state-of-the-art AI technology to continuously monitor refrigeration systems in real-time, analyzing data to detect slight deviations from normal operation,” said Robbins. “When the AI detects combinations of performance trends across eight families of indicators (including superheat, valve positions, component runtimes, enthalpy values, etc.), it flags a potential leak. A series of actions are then triggered to determine root cause, location, leak rate, and other information, and a U.S.-based refrigeration expert reviews each and every leak before it is released to the customer.”

After leaks are detected, they can quickly be repaired, as the ELD alert includes information such as leak rate; location of leak (e.g., display case, evaporator, rack); refrigerant type; and whether leaks are new or recurring. Catching leaks early can pay dividends, as Robbins noted that in 2023, a California-based grocery chain implemented the ELD module across 115 stores. “Axiom detected 43 leaks/year early, with all leaks exceeding 1 lb./day identified within an average of 9 days. This enabled the customer to lower its leak rates by 40.1% in 2023 (11,071 lbs. of refrigerant/year).”

Tech Shortage

While using AI for leak detection and preventive maintenance can result in significant cost savings for food retailers, the technology can also help alleviate the burden on technicians, who are already a scarce commodity in the commercial refrigeration industry. As Robbins noted in a recent webinar [below] about AI, “There's a shortage of qualified technicians, which is getting worse, not better every year. And that's getting exacerbated by work-life balance issues, which are leading to turnover.”

Maintaining a balance between work and personal life proves challenging in an industry where late-night or extended service calls are frequent, and technicians lack sufficient data to enable success, said Robbins. Robert Eidson, vice president of CoolSys in Gainesville, Georgia, echoed this sentiment during the webinar, noting that technicians do not have a work-life balance and that predictive maintenance would help solve this problem.

“Our technicians have a lot of overtime during the summer months, so they're raking in some serious dough, but they don’t have much of a life,” said Eidson. “When we get our first heat wave, the number of service calls goes through the roof, and they don't stop until we get a break in the weather. We talk about how we don't have enough qualified technicians, and then we are basically just killing them during the summer. Any problems that we can find and flag during April before we have major heat waves, are one less thing that we've got to send the guys out during the summer to try to fix.”

Repairs can also be complicated, said Eidson, so it may not be a simple 15-minute service call to make the fix. Meanwhile, there are 10 more service calls backing up, he said, stressing them out, and leading to an increase in technician burnout.

Again, AI is already being used to solve this problem by predicting and identifying problems before they become emergencies, said Robbins. Once the problems are identified, they can be batched and triaged in order to reduce the number and duration of these service calls, he said. Calls can also be reduced by empowering store employees to solve simple problems, such as closing a door.

“Technicians will also be empowered to solve the root cause during the first site visit, because they will be provided with this rich dataset. If it’s a refrigerant leak, for example, they can see how fast it’s leaking, so they won’t just tighten up a Schrader valve and say it’s fixed. They can find the real root cause, instead of using Band-Aid fixes, which are a key driver of unnecessary callbacks. So again, a key theme here is working smarter, not harder, using data and insights. And AI can help with that.”