Talk of Internet of Things / Industrial Internet / Connected Enterprise is growing. Consumer applications are partly to blame. Wearables for health and fitness and home automation are major consumer applications attracting attention and investment dollars.
The industrial case dates from at least 1999 when the cellular carriers were looking for new markets to sell data transmission (not foreseeing the smartphone boom). There was something called SCADA that used dial-up modems, FHSS radios and maybe T1 lines to send data from remote locations to a central command. Cellular could carry the data just as well — well, almost, because back in that day cell phone quality and dropped calls were a problem. Not like today. This was called M2M–Machine-to-Machine or Machine-to-Mobile.
The Internet and TCP/IP won. Now we see many companies jumping on the IoT bandwagen. We’re still defining what all the benefits will be, but we are certain that there will be many. GE calls this the Industrial Internet.
The GE Intelligent Platforms Business user conference, an occasional gathering, pops up on the radar in a couple of weeks. Unfortunately, I will not be there. A one-man shop can’t hit every conference. Especially a bootstrapping startup like this one. An interview with Rich Carpenter, Chief Technology Officer of GE IP became the next best thing to being there.
Rich described a progression that I’m beginning to see more of: collect data–>store data–>analyze (say for predictive maintenance)–>diagnose why things are not 100%–>recommendations that we’ve seen this before and here’s what you need to fix it.
Carpenter told me, “We still feel there’s a lot to be learned. We’re good at collecting data. Good at storing. We even can analyze for future failures. In some equipment, we can be 99.9%. But, we’re at best 30%-50% in diagnosing why. And we are really at 0%-10% on giving a recommendation on we’ve seen this before and here’s what you need to fix it.
He continues, “Our Smart Signal product that does predictive diagnostics has proven to be good. What’s happening these days is we’re changing from pushing it and doing pilots to where customers are saying ‘Wow this works, we have to have it.’ The reason is simple, every company if left on its own goes through this maturity curve. First, a guy goes out to feel asset. But that doesn’t scale. So we begin to remotely monitor, but still have to go out. Then we analyze and set limits for alarms. But those have problems. Then we add more intelligence like Smart Signal looking at the relationship among 20 variables and now you can predict better. Now you can start planning downtime–not wasting time and dollars due to unstable operations.”
Regarding what competencies are required, “First is platforms. We are trying to master cloud development paradigm. With the evolution of hardware power, we have to master massive parallel compute figuring out what it means to have infinite disk space. Second is around big data. That is an overused term. We look at it as volume of data. It also is a data variety problem–written notes, logs, genealogies, need to correlate and at ecosystem level. Third, expertise is needed user experience (not interface). How do you understand what problem customer is trying to solve and not just add new features. We hired 60-70 people to work on that. Then data science. How can we tease out insights from all the data and identify algorithms. Then cyber security–secure by design. Finally machine learning, model based control, sensor analytics.”
Then at last we need to think about the entirety of manufacturing. Carpenter, “We think these systems–ERP, PLM,MES–can no longer operate in silos. We have a close relationship with PTC. We are working to close the loop from design to manufacture. Then we have to look at the supply chain as just a step in the manufacturing process. Have the MES look at the whole chain and avoid things like incoming inspection.