Sramana Mitra: I heard two broad categories. One is around information security. The other piece that I heard is around analytics of data that these machines are generating. Could we take a use case and look at what kinds of data are being collected in these processes?
Steve Pavlovsky: If we think about our architecture today, if you start at the machine level, you’ve got sensors that are collecting inputs off the machine. They might be position sensors of some sort. They might be vision sensors. They might be thermal sensors or thermocouples. You have a whole set of process variable data coming in, and then you have what the machine state is and the process state and what we in our industry call outputs. What motor speed was a motor turned to, or what other actions were taken? You’ve got an aggregation of data in terms of the current status of the machine, the work elements going on in the machine, and what the control system is telling that machine to do.
That exists on a machine-by-machine basis. For years companies have aggregated that data across cells of machines or even up into plant level systems. You now have all of this data from the actual machinery. You could be also starting to collect ambient temperature or humidity if you have processes that are sensitive to those kinds of elements. There are all kinds of data around the actual process that’s now possible. What’s important for customers to [understand is that they can] get the next level of productivity by improving not only their overall process but also by eliminating downtime or potential downtime. We’ve invested in technology that allows our customers to collect process data, whether it be machine data or other data from the environment around the process. We log the data in the historians. We provide a set of analytical tools that watches for and makes correlations between what the process variables look like when the machine or process is operating correctly and what state of variables can be correlated to when the machine is not operating correctly or the output of the process is not a high-quality product.
The building of those analytics and the monitoring of that data enables the customer to understand what are the specific process variables that influence the actual input of the process, and in many cases, we have customers who look at that data. Maybe the temperature, the speed, the vibration of a turbine or a pump or some other expensive asset they’re monitoring and understanding. When the machine is operating correctly, what do those process variables look like? [We are] building mathematical models that might predict when there will be vibration, and at the same time [indicating] vibrations when the machine is running at a particular speed that portend a potential breakdown. It’s much more cost effective for a customer to perform an orderly shutdown and perform a maintenance upgrade to that machine to solve a problem before there’s a catastrophic failure that might be an expensive and long-term repair.
We see all kinds of analytic capability based on the data. We have customers who are oil patch customers. They have oil wells, maybe hundreds or even thousands of oil wells, with our controls on them where they’re collecting data. All of that oil has to be collected. It is managed and brought to oil terminals or seaports via pipelines. Understanding the flow of oil coming out of particular areas of the oil field or an individual well helps clients understand the productivity of that well or the area of the oil field and what capacity they need from a transportation perspective.
Depending on the application, whether it’s an expensive piece of rotating equipment like a turbine or a compressor, an oil field application, a water treatment application, there are all kinds [of problems to solve]. We think about the correlation of water treatment facilities where part of water treatment is the moving of clean water and dirty water from place to place. Think about an integrated system where you’re taking weather data. There’s a storm approaching that’s producing lots of rain. You may need to pump dirty water out of a holding reservoir ahead of time so that you don’t have spills. There are all kinds of opportunities to co-mingle disparate data to build smarter systems that are either safer or more productive or have less downtime.
SM: We have companies in our 1M/1M program that are working in analytics. One is even working in manufacturing analytics. What would be the best way for them to plug into the GE innovation ecosystem?
SP: We’re absolutely interested in creating an ecosystem of partners that are building both control algorithms and analytic tools in our environment. I don’t know that we have a standard incubator for those kinds of folks, but we do have partner programs where we can engage them and give them access to our tools, infrastructure, and APIs and open up our environment to enable them to participate. We’re thrilled to have those conversations.
SM: OK. Thank you for taking the time to talk with me.
SP: I appreciate the opportunity.