Until recently, technical limitations stifled the power of IoT to transform areas like marine technology, energy, and oil and gas exploration. Here's how a new approach, edge computing, is changing the game.
Think of a classic Internet of Things (IoT) setup and you'll probably picture thousands of sensors streaming data into a cloud. There, the data is processed, and some kind of result is produced: a warning to replace a component, or, a recommendation to adjust a machine for better operation. But what if running your data through a cloud isn't an option? What if internet connectivity is unavailable or is too slow, or if security is an issue?
Edge computing, the latest development trend to hit the world of IoT, solves this conundrum by moving the bulk of the processing action out of the cloud and putting it physically as close to the end devices as possible. A piece of hardware, called a gateway, collects information from the sensors, analyses it and generates insights. Then, if needed, it sends all or only a part of the data to the company's cloud.
Cutting IoT loose from the bonds of the cloud, of course, has huge implications for a number of industrial activities, from drilling for oil or gas in remote areas to creating autonomous vehicles. Some tech pundits note that the edge computing model might also gain traction as the amount of data from IoT devices swells, making bandwidth an issue.
Although it is still unclear how fast or widespread the migration will be, owners of the world's top public cloud platforms are taking note and making investments. Since 2017, Amazon and Microsoft have released their own edge computing platforms. Google followed suit this summer announcing that its Tensor Processing Unit circuit would be available for the edge.
Do these moves signal an edge computing revolution in IoT? Ergin Tuganay, Partner and Head of Industry 4.0 at the multinational strategic change and technology company Nortal, warns against over-hyping the phenomenon.
“In one sense, 'edge computing' is an evolutionary process we've seen for over a decade. Meaning, more and more intelligent field devices and gateways are performing data collection, aggregation, processing and controlling the system, sending limited amounts of data over the internet securely,” he says. “The real difference today is the amount of intelligence that's being deployed at the edge.”
“That intelligence typically comes in the form of a “digital twin” – a digital representation of the physical world – and predictive models to support and even automate decisions,” explains Tuganay. Helped by the rapid progress in AI, these models are developed in a cloud through a massive amount of historical data collected from separate systems and machine learning, and then “pushed out to the edge” to independently analyse new data in motion and draw conclusions on the fly.
Though it's an intriguing concept, Tuganay says, for typical industrial companies, edge computing at its full potential is, for now, still a rarity, with development focusing on breaking the data silos in order to proceed to the machine learning phase. However, development has sped ahead in some niche areas, typically asset intensive industries like oil and gas, energy, and pulp and paper, where demand is high.
Since both of Wärtsilä’s areas of speciality – marine and energy – are on that list of niches, it's no coincidence that the company has found itself on the forefront of edge computing solutions. Toby White, Vice President of Digital Engineering, details how Wärtsilä is leveraging the power of edge computing to improve ship performance and manage electrical grids.
“Marine is a particularly interesting case for edge computing because bandwidth is such a problem. There's just no way you're going to get the data off to do processing in the cloud,” he says.
Using the strategy outlined above, Wärtsilä’s marine technology company Eniram collects data from a ship for several weeks to build its digital twin, a bespoke model of that ship's behaviour. Back on the ship, an edge node runs sensor data through the model to create recommendations for optimising the match between power production and consumption over time – a basic operating procedure that is enormously important for fuel efficiency, but, is dependent upon too many variables for a human to work it out effectively.
Currently, Wärtsilä's edge computing setup “is limited in the range of advice it can provide for a ship's captain,” says White. “We're optimising a small number of parameters, such as trim. However, there are a lot of systems on board and there are still hundreds of decisions that the crew has to make. The more of these decisions we can take into account, the more advice the edge computing setup can offer.”
On the grid management side, where every millisecond is critical, it is latency rather than a lack of connectivity that edge computing solves.
As White explains, “a modern grid with renewable energy will have inputs from multiple energy sources, including solar and wind, which are notoriously unpredictable.” “They're all coming on and offline at different times and they all have different performance characteristics. Your demand is fluctuating as well,” he says.
Human operators are left to guess where the demand is going, and then switch inputs on and off at the exact right moment. “There are a lot of inefficiencies if you have too much power, but if you respond too slowly, you have a brownout.” Unlike blackout, which is a complete loss of power, a brownout is a reduction in or restriction on the availability of power in a particular area.
Greensmith, a Wärtsilä company, delivers grid management software that improves the process by taking the humans out of the loop. Only edge computing can allow it to perform at the required speed. “You need to be able to respond in milliseconds, so you need this processing happening as close as possible to the control software.”
Demand for such edge computing solutions in grid management has soared. White says, “it is set to grow as grids become more complex and renewable energy cheaper.” That growth, if it happens, would be in line with trends predicted for edge computing across the board. Gartner Research predicts that 40% of enterprises will have an edge computing strategy in place in 2021, up from about 1% in 2017. The cloud will always remain central, but IoT, it seems, has found a new place in the sun.