Ever since the Germans defined the term ‘Industry 4.0’ back in 2011, companies across the manufacturing sector have been puzzling over how and why they should go about deploying connectivity and big data in production operations. Some companies start small with pilot projects and plan to scale up. Others rip-and-replace entire lines of equipment. Like many emergent manufacturing technologies (robotics or 3D printing for example) it can be a challenge to see past the technology and get a straight answer on the bottom line. Setting aside the ‘cool factor,’ how will the IoT help you make money?
To help answer that question, engineering.com sat down with Ron Victor, founder and CEO of IoTium
, a Silicon Valley-based network infrastructure company. Because his company is focused on the IoT architecture itself rather than whatever AI, machine learning (ML), analytics or monitoring solution the network supports, Victor has a good perspective on these business-level drivers.
What are the factors that slow digital transformation and adoption of industrial IoT in manufacturing?
Two things: Deployment complexity and security. Everybody can do a nice little proof of concept in a lab but trying to deploy at scale in a cost-effective manner and making sure it is secure as possible given all the concerns related to hacks, et cetera are the two fundamental drivers that are hindering it. The third driver that is hindering it is business value. In other words, can they make more money because of it? Can they reduce costs because of it? Can they eliminate truckloads because of it? Can they provide an additional service level agreement (SLA) with the service so they can charge more for it? These are the business insights that drive digital transformation. Here are a few examples:
- I am trying to connect my machines to an application so that I can charge my customer more for minimizing downtime.
- I am trying to connect my machine so I can now remotely configure that machine versus having to send a person out there and reduce my costs.
- I am trying to connect my machines and run a machine learning application that can predict that the bearing is going to fail after x time or when the oil is low. So, before the machine breaks, I can take preventive action.
You have to look at IoT completely from the business angle. Ask: what are we trying to achieve that we cannot achieve today with an unconnected asset? If my asset was connected:
- Could I provide my customer with an SLA and say, ‘I am guaranteeing a certain uptime to you, therefore pay me more?’
- Could I do some work remotely and eliminate flying in an engineer?
- Could I prevent the machine from going down through continuous monitoring?
- Can I run machine learning on this asset so that I can build a model that tells me that typically after 36 days of operation an oil change is required, and give an informed decision to my customer saying, "You have not performed the required maintenance, please change the oil or we’re going to take the warranty off.”
These are the problem statements that need to be addressed. IoT is just a means to an end for those problem statements. Companies think they need to transform digitally. Well, the question is why? Once you identify the business driver, then it comes down to deployment complexity and the security element.
Do manufacturing companies understand what’s involved in building an IoT network?
Some do. Companies don't need to add the overhead costs internally of building IT infrastructure for OT. Building IT infrastructure is not their core competency as manufacturers. Instead, most manufacturers are better off focusing on data science, machine learning, AI—on the insights they want to gather. Smart manufacturers put their dollars to work related directly to their business expertise, and just like they outsourced their IT infrastructure in the IT world, they outsource IT infrastructure for their OT world. Some people get it right off the bat. Some people have to plan an experiment for whatever reason and then they will get it.
Image courtesy of IoTium
What is software-defined networking (SDN) and why is it ideal for IIoT applications?
Software-defined network infrastructure means it is completely defined and managed to an orchestration layer in the cloud, where you can, for example, connect point A to point B with one click of a button. And tomorrow you say, "no, I'm gonna disconnect that connection and connect it in another way." Whether this orchestration layer lies on-prem or off-prem or in the cloud doesn't matter.
Imagine I am using Azure and taking my data to Azure. Tomorrow I want to take it to AWS Greengrass, and the day after I want to think about IBM Bluemix to be able to leverage analytics applications that Bluemix has done. And then the day after that, I want to take it the Mindsphere because there's this beautiful predictive maintenance that it is on Mindsphere. If this was not software defined, you would have to physically go to that edge gateway and manually configure it. All of this being software defined from an orchestration layer allows me to sit in my office in Chicago and move data wherever I want it to go. I can implement security applications at the edge, saying, “I want this encryption for this tunnel and that encryption for that tunnel and this machine learning for this sensor and that machine learning for that sensor.” All of this being completely provisioned and deployed all from the cloud is software defined infrastructure.
How has today’s IIoT technology changed industrial analytics?
The difference between 15 years ago and now is the internet and the cloud, that's all. Nothing else is a different type. Whatever you could do on-prem in a closed environment people are doing today in a walled garden environment. With the onset of the Internet as well as the cloud, what happens now is it democratization of applications. 10,000 people can write all kinds of applications to do many things.
So, today’s manufacturers can leverage more and more applications because of democratization of the applications. Today there are plenty of people in Silicon Valley writing machine learning applications for machines that they don't know what the machine does. Why is that so? Because they basically say, give me a pattern. I will acquire that data. I will build a model based on that and based on the model I will be able to predict. Compare this to the legacy application which was built 15 years ago. It is a great application. It does what it does, but it's an on-prem, closed application. Now you can leverage 10,000 applications from other people democratizing that application, making sure you're not vendor locked into one vendor. Today, third parties can predict the failure of a Siemens machine earlier than Siemens can. That’s the difference.
To learn more about industrial internet of things in manufacturing, check out The Connected Factory and More: 5 Examples of How IIoT is Changing Manufacturing.