How Machine Learning Understands Machines
Here in Massachusetts, we find ourselves surrounded by beautiful old mill buildings that act as a constant reminder of how our nation’s history is built on manufacturing. In the early days of the industrial revolution, water powered the mills that produced the products the young nation needed to move forward.
Even today, with tech companies all around us, we know that machines continue to power the world. They move us from place to place, whether that’s a car, train, airplane, or ship. They provide the power we need and till the soil to provide the food we eat.
The enemy of production is downtime, so to keep these machines running mechanics are constantly doing regular maintenance, often repairing and replacing parts long before they need it to simply avoid breakdowns.
In just the past few years, things have shifted dramatically. For decades machinery have been fitted with equipment that enable them to produce data, but only recently have we started to understand how to read that information. Today, with machine learning, deep learning, and other pieces of artificial intelligence, we can gain deep insight into the machines themselves.
Presenso is one company offering that kind of service and its CEO, Eitan Vesley recently spoke with TechRepublic about how this isn’t just about IoT sensors. In fact, this is about using the information that already exists and then applying the right algorithms to gain intelligence. Presenso works with companies running very large pieces of machinery, like wind turbines, power plants, oil fields and refineries, so the machines he’s referencing are not minor.
"Most of the data in these process industries, most of the sensors are already connected and the information is flowing to a centralized database known as a process historian. Though there are databases storing three, five, 10 years of historical data, which is just lying there, nobody really is doing anything with it except for post mortem analysis. We’re simply interfacing to that historian database and gathering thousands and tens of thousands of sensors and streaming them to our cloud,” he said.
All that data is incredibly important in understanding what happens within the machinery, but some of our customers have even more data than must be considered when looking at maintenance. While it’s not difficult to analyze machine-created data, but there are other sources of information that can help you be more predictive about what to do with that information.
In the case of aviation customers like one of the leading aircraft manufacturers we work with, there is often an iPad-toting human who does an inspection and takes notes what they see. That inspector may note a chip on the exterior or a loose bolt, or even some peeling paint. Something that may be small on an early inspection could turn into a larger issue later, or it could be a result of a certain use-pattern. But if the system were to only rely on information produced by the aircraft itself, all this additional data would be lost.
Farm machinery manufacturers also collect unstructured human-generated information, but in this case it often comes from customers themselves. By including CRM notes and recordings of calls with customers, the company can better understand what its customers need and what is happening with its farming equipment.
This not only leads to proactive maintenance, but also can help in designing better products in the future.
Of course, this doesn’t mean that the robots are taking our jobs. In fact, while the Attivio system relies on human input to be truly successful, Vesley also sees opportunity for growth here. For now, he sees people still getting their hands dirty in replacing parts, but beyond that he also envisions increased roles for data scientists, application developers, and others who will help better understand how to use all the data machines produce.
Want to learn more about how machine learning helps drive preventative maintenance in manufacturing? Contact us.