Making sure plant floor machines remain up and running is a priority for manufacturers. As such, OEMs are beginning to bundle in remote management software from technology suppliers so that they can monitor machines, troubleshoot and even repair issues from afar before they impact productivity.
While avoiding downtime is the overall goal of remote management, not everyone is sold on the idea. Specifically, the information technology (IT) department, which is responsible for keeping the enterprise secure. Having an access point that connects a machine to the cloud to apply analytics is a potential “open door” that could inadvertently allow a bad actor onto the network.
The alternative to remote access technology and predictive analytics in the cloud is to have a dedicated team of data scientists and maintenance technicians onsite to keep tabs on overall equipment performance (OEE) in order to proactively fix problems. But, between a skills shortage and budget constraints, many manufacturers don’t have the option of having extra expertise on hand. And, in fact, it could be an inefficient and uneconomical use of their professional time on a narrowly-focused processing or packaging machine.
According to Omron Automation Americas, all that’s really needed is anomaly detection inside the machine. And that’s exactly what the company’s newest controller does. The Sysmac Artificial Intelligence (AI) Controller integrates machine learning functionality into an edge-level industrial controller. The controller learns the data patterns of nominal machine behavior without the need to be programmed.
“It’s in the machine and it’s a small scope of analysis,” said Mike Chen, director of Omron’s Automation Center Americas. “It doesn’t need an enormous amount of data, but it is still very valuable to production.”
It’s value is twofold: First, because the AI controller is collecting raw data right at the edge of the machine, it is ensuring high data fidelity, consistency and security. Second, because it is localized and data does not have to be sent to the cloud, anomalies can be accurately detected within milliseconds, and acted on in real-time.
Leveraging Omron’s machine learning model and AI predictive maintenance library, the controller automatically creates data models from correlation analysis, and monitors machine status based on that model to make data-driven decisions related to specific equipment issues.
“When you have a packaging line there is wear and tear that is unique to each machine,” Chen said. “That’s why you need machine learning inside the machine, to learn the unique wear and tear of that machine. The controller learns the profile of what normal is and if it is out of the designated spectrum, it tells the programmer that something is wrong.” This is done through an AI visualization tool that shows an anomaly score on a scale that “empowers the programmer to make key decisions, like slow down the machine or exit material off,” Chen said.
This is a platform that can be quickly utilized by programmable logic controller (PLC) programmers who speak the language of ladder logic. “There’s no need for a data scientist who may take three months to solve one narrowly scoped problem using machine learning,” Chen said. Instead, a programmer can be trained on this within in a few days.
Rolled out to Japan and Europe first, the AI controller is currently being tested at customer sites in the U.S., Chen said. And, while it might seem like a small step in the direction of machine learning, it is a risk-free way for manufacturers to get acquainted with AI.
And, it is an important stepping stone to the factory of the future, as it is another way that people and machines can work in harmony.
Leaders relevant to this article: