Most people can agree that using data to improve manufacturing operations makes sense, yet I find it interesting that many still suggest companies should start by moving data to the cloud. This may stem from a prevalent misunderstanding that plants are as consistent as the products they manufacture. While using the cloud to analyze data can add value, I feel this focus causes people to miss the chance to use the data to enable continuous process improvements.
Today, the convergence of information and operations technology has led to new capabilities at the edge that can help accelerate insights by shifting analytics from the data center to a location on the shop floor closer to the “action.” We call this new capability edge computing.
The vertical integration and shop floor flexibility enabled by this new class of edge computing solutions opens up new operational effectiveness opportunities. For example, a diaper manufacturer improved energy efficiency by analyzing the peaks and valleys of energy consumed by cells in the production line. The information helped the manufacturer significantly reduce its energy footprint for the site. Additionally, leveling out consumption let the manufacturer negotiate reduced rates based on lower consumption patterns.
Background on edge computing
Edge computing has been around for several years, but interest has increased over the past year with the rise of internet-of-things (IoT) devices and sensors in manufacturing plants. Edge computing helps manufacturers turn vast datasets generated by machines and sensors into insightful and actionable data. It accomplishes this by using analytics at the source of the data, whether that is temperature sensors, alarms or motor drives. Improved analytics increase the ability to visualize the data and gain new insights, while machine learning and artificial intelligence help manufacturers apply these new insights in the production environment.
“Edge” refers to the computing infrastructure that resides closest to the sources of data, such as robotic arms and automated heavy-lifting machinery. These are considered to be at the edge because they tend to exist farthest from the heart of the computing infrastructure, which today often resides in the cloud. In many of these situations, only summarized data gets sent to the cloud. In addition, in environments where connectivity remains intermittent, data gets collected and saved locally, ensuring that all the data gets captured.
The impact of edge computing
Edge computing offers some powerful results. In an oft-cited example, a self-driving car going 55 miles per hour on a crowded road can analyze data and turn it into a machine learning command right on the vehicle, which then tells the car to avoid a pedestrian, preventing a bad accident or even a fatality.
And while the autonomous car example gets discussed in the consumer press, it misses the point of what’s really important in commercial, industrial settings. Even if a manufacturing company has multiple plants around the country or world, each plant has its own parameters and dimensions and set of machinery, so it’s difficult to maintain a centralized data analysis capability in the cloud or at a corporate data center that can analyze the data and respond in a timely manner.
Front-line managers at plants don’t have time to wait several days for a centralized capability to analyze the data and send a machine learning automation function back. With the automation tools available today, plant managers can use data at the edge to create a machine learning function they can deploy in a matter of hours. By taking advantage of these capabilities, plant managers can identify an issue on the shop floor and take corrective action well before a machine breaks down and stops production.
For industrial plant managers, edge computing offers a unique opportunity to use analytics to run predictive maintenance, potentially avoiding a costly plant shutdown long before a part or a sensor breaks down.
In many ways, the industry has gotten ahead of itself. We are several years away from the kind of edge computing in autonomous cars where edge devices make snap life and death decisions on a mass level. But edge devices for IoT environments in plants are available today, and with the right training and expertise, manufacturers can realize the promise of predictive maintenance, reducing plant accidents and increasing uptime.