“Troubleshooting is a problem that has always existed in manufacturing,” says Jonathan Wickstrom. He is the manufacturing digitalization leader at Husqvarna. Husqvarna, headquartered in Stockholm, Sweden, has been around since 1689, when it began manufacturing muskets. It produced its final firearm 300 years after that and today produces outdoor power products like robotic lawn mowers and chainsaws. It also makes equipment and diamond tools in the construction and stone industries.

The company, in collaboration with Microsoft, is using gen AI today to reduce unplanned shutdowns on the factory floor, when technicians and operators are troubleshooting equipment. The AI Factory Companion is a gen AI copilot designed to help technicians and operators diagnose and solve issues with machinery. The companion can help by describing the symptoms that a machine is experiencing and then suggest tests to validate and solve the problem.

He says that a machine can break for many different reasons. “Historically, this was solved by documentation.”

The technicians would read through user manuals and maintenance history to try to understand the problem. The manuals are often in paper form or PDF format, and the maintenance history can be logged either in operations where Wickstrom is located, or in another silo. Wickstrom says, “We have connected machines.”

“Sometimes, our machinery stops due to an IT outage. All of these outages are documented on our IT incident system. “A newer employee who enters the shop floor would not have that depth of experience,” he says.

Moreover, Husqvarna’s shops run 24/7. During the day, highly-skilled technicians who spend the majority of their time developing new solutions are available, but not at night, which could result in extended periods of downtime.

Powered Azure OpenAI, AI Factory Companion can access data generated by machines on the factory floor and a knowledgebase of manuals and documentation for maintenance history, regardless of which department it is in. The companion can access the knowledgebase using

RAG (retrieval augmented generation)

Not only can users ask for help from the companion, Wickstrom and team also use the capabilities of the companion to make machinery in the factory proactive. If, for example, a piece of machinery emits an alarm that something’s wrong, the companion can take the initiative and query possible diagnostics and solutions which it can then provide to an operator proactively.

Simplifying complexity

Husqvarna, which has been on a journey of digital transformation for years to improve its efficiency, is now launching the AI Factory Companion as part of that journey. The company, like many manufacturers, faces the challenge of legacy industrial equipment that has been in place for decades. Integrating data from this machinery with modern systems can be a difficult task. In order to achieve low latency, some data infrastructure must be localized at the edge. It also needs the resource capacity of cloud.

Husqvarna has leveraged Azure IoT Operations to serve as a central hub for collecting data from industrial equipment. IoT Operations is designed to help organizations transform physical operations through Microsoft’s adaptive approach. This unifies siloed sites, teams, and systems in a single data, security, and application model.

Microsoft’s adaptive cloud is about bringing cloud service to places where they haven’t gone before,” says Bernardo Caldas CVP of Azure edge Product Management at Microsoft. “Implementing AI into edge environments such as factories can be difficult, because many companies struggle with the complexity of data management and integration with existing systems.

Manufacturing data is traditionally layered:

• level 0: physical process from devices like sensors
• level 1: local control systems like programmable-logic controls (PLCs), and remote terminal units, that manage and control physical processes
• level 2: supervisory data acquisition (SCADA), systems, which monitor and control) systems, like PLCs and RTUs and processes.
• Wickstrom says Husqvarna has made significant progress in its transformation, including a move to a point where it can measure and control downtime. However, it is difficult to quantify how much more efficient it is with the AI Factory Companion.

  Daniel Johansson is the manager of manufacturing digitalization at Husqvarna. He says that technicians and maintenance personnel spend up to two hours fixing stoppages. “In some cases we will reduce the time by a small amount, but in others we will reduce it by 50-60%.”

  Work in progress

  Wickstrom says that the companion’s knowledge base is still in its infancy and he expects it to become more useful as the database grows.

He says, “We can’t yet answer every question, but we can address simple ones like 20% of a full bucket and in some instances, we can accelerate the time to resolution.” “If we don’t have the data, then we can’t do anything. And it will be that way for a while.” Now we can improve the knowledge base, instead of filling out another PDF that nobody will read.

The real magic lies in good search – finding the documents that are most relevant to answer a user’s questions. This requires combining traditional technologies with LLMs and mastering the technology. He says that if the user asks, “What’s the latest thing to happen to this machine?” it is difficult for gen AI because “latest” can be a complex concept. A SQL query can sort by date to provide the top results.