Your browser is not supported. Please update it.

29 March 2018

Expert Systems, AI, and the IIC – how far have we come?

The Industrial Internet Consortium (IIC) recently announced a Deep Learning Facilities testbed, set up by Dell, Toshiba, and Wipro, which will look to find the value of using deep learning technologies in buildings and facilities management, as well as energy and utilities. The IIC, not a standards group itself, is looking to prove use cases, and bring a better focus on technologies that are needed in the Industrial IoT.

In a catch-up call following MWC, we learned that this project has already kitted out a Toshiba building in Japan with 35,000 sensors, generating data that can be used to prove the viability of the testbed technologies. The goal here is to take the known benefits of AI and ML in areas like automotive and financial deployments, and explore if they are possible in the Smart Facilities sector.

The IIC’s Executive Director, Richard Soley, said that the IIC strongly believes that you need the experience of actually doing something in the field in order to accurately contribute to any form of standards process, which is why the IIC runs its testbeds. This is recounted in a recent whitepaper from the IIC, available here.

When asked about the progress of the IoT so far, Soley said that he wouldn’t say it was now mainstream, but that there has been huge progress made. He said that AI was key, but stressed that he had been involved in artificial intelligence for around 40 years already.

Soley recalled working on Expert Systems for diagnosing faults in industrial Programmable Logic Controllers (PLCs), back in the 1980s, where there was a severe shortage of human diagnostic skills. Soley quipped that there were only about 3 people who could carry out these diagnoses, and at the time, they were all headed towards retirement – hence the drive to recreate their knowledge inside a machine.

When downtime from these PLCs was costing around $20,000 an hour (north of $60,000 today), there was an obvious need to solve both the outage costs and the imminent problem of a looming knowledge gap. That experience has followed Soley through his career, and now as the executive director of the IIC and the CEO of the Object Management Group (OMG), we were happy to pick Soley’s brain about the current state of the IoT.

He said that the IoT is largely about the price and availability of compute going to near zero cost, in tandem with the ubiquity of communication. However, the main issue facing the IoT, according to Soley, is the integrations of things that weren’t ever designed for it – such as those 1980s PLCs, descendants of which are still in use today, in a very different digital world.

But for all the well documented security concerns, there are huge opportunities in the IoT. Soley noted that the Track and Trace IIC testbed had achieved 50% time savings for manufacturing staff in an aerospace factory, using prompts to guide workers to tools in order to improve their productivity.

The tool and process triangulation is only possible in a connected factory, but many factories will house devices that were designed to be air-gapped – on the assumption that the outside internet will never reach them. Evidently, the air-gap approach can’t work in an IoT system, as it needs an internet connection at some point. Heavy firewalls and security software will be used to protect them, but bringing such systems into the same building that houses an air-gapped production line will be a cause for concern among the IT and OT teams.

Providing such security is a new market opportunity, but across industrial sectors, there is going to be a lot of discomfort as these air-gapped systems are replaced over time – moving into a world of hyperconnectivity, from the previously safe disconnected architecture. But as the IIC’s testbeds are finding, there are clear business reasons to invest in the IoT, and embrace those technologies.

But Soley warned that all this extra data actually requires some careful thinking about, when it comes to integrations. He recalled a conversation with a locomotive specialist that was able to capture nine million data points per second. Soley remembers asking how much of that was actually useful. “Three,” he replied, to which Soley said that 33% usage was actually quite impressive. The rail specialist had to clarify that he did not mean three million – just three data points.

When we asked Soley about the most surprising thing he had learned through the IIC’s work, he said it was the importance of semantic processing to environments like the connected factory. Having a way to define the terms used to let tools and other devices from different brands describe their status and actions in a common framework is exactly the sort of thing that the IIC would recommend, following its experience in the testbed.

On this point, Soley noted that the OMG requires that a standard is in use before it is published by the group, otherwise it is not worth the paper it is printed on. As just one of the 30+ liaisons that the IIC has with standards groups, the OMG is an advocate for public standards that can prove their value to the ecosystem. In theory, only the worthwhile standards get adopted, and this should give users more choice, and give vendors a bigger pie to slice between them.

As for his thoughts on the IoT in general, Soley said that it is an evolution, which is rather nuanced, but that it has been very badly over-hyped. Despite this, it is a huge opportunity, according to Soley, with manufacturing, agriculture, and water management being standout opportunities, based on what the IIC has seen.

When asked about future testbeds, Soley said the IIC had a pretty comprehensive portfolio of them now. He noted the current lack of a connected vehicle testbed, but said that there were two such projects in development.

An especially interesting upcoming IIC testbed is for the mining sector, which will explore how to create an underground network for worker safety. Because of the amount of copper ore in the mine, a broadcast WiFi network is not practical (due to the reflective surfaces interfering with the signals), and so the testbed will explore point-to-point communications devices, which include trackers on worker jackets – a big upgrade to today’s system of bells and horns.