(79) What are the challenges in industrial device connectivity today?

It was projected that by 2020, more than 50 billion devices would be connected to the cloud. But that goal doesn’t seem to be achievable as we approach 2019. And the biggest reason for this lack of pace is the protocol incompatibilities of different kinds of industrial machines in the world — especially from legacy machines that do not support standard connection protocols and need complicated integration, workarounds and configurations to connect them to cloud and draw data out of them. If any IoT platform ever has to have a competitive edge over other it has to have a fast and easy way to connect, which my German friends would say, schnell and einfach.

I spoke to the CEO of Omnio, Mikkel Sørensen to understand the topic better.

Here are the key insights:

Why is it so challenging to connect industrial devices than it is to connect consumer appliances easy?

• In consumer appliances, you don’t have to know the exact model number and go online and find out how to write a driver and create a custom tool to write the driver. This is not necessary because the standards of communication were drafted in the 90s.
• Expensive to get communication standard: Industrial devices on the other hand were not built for connectivity in the first place. Manufacturers have used various kinds of protocols and there are many different kinds of standards. Therefore the connectivity features for every machine are very different from others. You can buy communication boards for bringing them to one standard but they are very very expensive.
• Unique data models: You can have two similar devices from the same vendor but you will have different data models. That means you need to understand each device data model in order to send it the right instructions to communicate with it and interpret the response. To extract information from all devices can take hours to days or weeks of effort. You then need to normalize the responses, which is painstaking exercise.
• Issues in data normalization: Different devices have different data models. Even though they may have similar data points, they will be different. For e.g, temperature can be measure in many different ways — Fahrenheit, Celsius, Kelvin etc . You need to normalize all of that before you can pass it to a IoT software application.
• IT tools are not compatible to OT tools: A lot of the IT tools don’t come from the automation world hence they don’t inherently understand the world of industrial automation. Hence, there are a lot of issues on usability and design.
• Cost on onboarding a device: There is a large cost of developing a driver for a device the first time and this effort can run into weeks for a single device. Also, the skill set needed for this is provided by an automation engineer.
• Hard to scale: Its hard to scale every time you connect to another device because you need to program a driver and need a few weeks to test it. This is why 70–80% of industrial IoT projects get stuck beyond the pilot stage.

What do you mean by data models?

• For example, Machines with Modbus protocols have a list of data points that you can get from a pump. For example, you can the flow, the speed and pressure. And average device might have a couple of hundred data points. These data points will have a name for example, Flow. They will also have units, for eg, cubic meters per hour. It will also have an address from where you need to ask for inputs. You need to send the address to communicate from the machines. You then need to get the data back and analyze it.