Sensor processing platforms add performance, cut SWaP for the Industrial Internet
As the Industrial Internet demands more out of resource-constrained industrial data acquisition systems, sensor processing platforms are required to add performance while reducing size, weight, and power (SWaP). Rubin Dhillon, Marketing Director of the Embedded Systems division at GE Intelligent Platforms discusses his company’s approach to developing “brilliant machines,” and how off-the-shelf hardware is enabling these sensor platforms for the Industrial Internet.
For GE, where do sensors fit into the Industrial Internet?
DHILLON: The fundamental building block of the Internet of Things (IoT)/Industrial Internet is, of course, the communications infrastructure. Attached to that are increasingly smart, brilliant machines with substantial data acquisition, processing, and storage capability built into them. For the Industrial Internet and for the Connected Battlefield, those very smart machines need to be incredibly rugged, able to operate in the harshest, most challenging environments.
Sensors are a critical element within the IoT, Industrial Internet, and Connected Battlefield because they acquire the data necessary to effect huge improvements in efficiency and reliability – with minimal human intervention. GE has a long history of dealing with sensor-acquired data through our work with the military – radar, sonar, video, and so on. We’re now finding that that expertise and experience plays out well in the Industrial Internet space.
Given the resource constraints of industrial devices, what innovation is occurring at the sensor level to realize the full potential of the Industrial Internet?
DHILLON: The key here is that the proliferation of sensors is producing a fire hose of data, and that situation is exacerbated by the resolution of the data that is being captured. Video sensors, for example, now commonly acquire and transmit high-definition data rather than standard definition, with huge implications for the numbers of pixels being captured, processed, transmitted, and stored. That in turn has huge implications for both processing power and network bandwidth. The trick, of course, is to be able to process that data in real time and extract only the meaningful data to deliver to the network.
Here, GE believes that one answer is GPGPU technology – using the graphics processors designed for high-end video gaming and so on, and leveraging their hugely parallel architectures to process vast amounts of data at incredible speeds. We’re working closely with NVIDIA, and an outcome of that relationship has been products like the recently announced rugged mCOM10-K1, which is based on NVIDIA’s Tegra K1 technology (Figure 1). That’s an incredibly appropriate solution for Industrial Internet applications. First, it’s based on the COM Express architecture, which means that it is very small and can be deployed in very tight spaces. Second, it delivers 326 GFLOPS of processing power. And third, it consumes less than 10 watts of power, meaning that it can be used in environments where power availability is at a premium.
We’re also working very closely with AMD, as a result of which we’ve just announced the bCOM6-L1700 (Figure 2). That also is a COM Express architecture product, featuring AMD’s latest SoC technology, and, like the mCOM10-K1, it is small, rugged, and consumes minimal power.
It’s going to be devices of that level of processing performance, ruggedness, size, and power consumption that we’ll see increasingly drive the Industrial Internet because of their ability to help make sense of the huge amounts of data that sensors are capable of collecting.
What other trends do you see arising in Industrial Internet sensor networks, and how is GE addressing them?
DHILLON: As I noted previously, if we’re to avoid dragging the communications network to a halt with the sheer burden of sensor-acquired data, that data needs to be processed locally in the machine attached to the network. That has important implications not just for processing power, but for size, weight, and power – so-called SWaP. That processing power has to be built into small spaces in harsh environments that are subject to extremes of temperature, shock, vibration, contaminant ingress, and so on. That means it also has to be rugged. That’s a real strength for GE. It’s what our customers know and value us for.
The other key area where the Industrial Internet can make a real, valuable difference is in improving efficiency, asset utilization, reliability, and availability. It’s one thing to have all those machines generating huge amounts of sensor-derived data about themselves – but what to do with all that data?
GE is investing hugely in software, especially the software that will drive the Industrial Internet by making everything hang together in a way that will deliver benefits to businesses. An example of that, and a rapidly growing business for GE, is our predictive analytics business. We at GE believe that the basis for real transformation here lies in the shift from reacting to equipment failures or current condition indicators to becoming truly predictive and proactive. Answers about the future health of equipment or locomotives or whatever are already there in data that is being collected from sensors onboard those platforms today. It’s a matter of leveraging an analytics solution with the proven ability to find the signal in the noise. SmartSignal from GE is that proven solution, with over ten years of proven experience across a broad range of asset-intensive industries.
What predictions do you have for the IoT?
DHILLON: For many companies and organizations, the Industrial Internet is already a functioning reality – it’s real, it’s here, it’s today. Many people don’t realize how much progress it’s already made. There’s no doubt in our minds at GE that the Industrial Internet will become pervasive across all industries, and the day is not too far in the future when we’ll wonder how industry ever managed without it.
GE Intelligent Platforms