Sign up for our newsletter and get the latest big data news and analysis.

Taking Machine Learning to the Edge

In this special guest feature, Matthew C. King, IIoT Solutions Expert, FogHorn Systems, discusses how edge machine learning combines two hot industry trends – moving industrial internet of things (IIoT) compute to the edge of the network and the ability to model new efficiencies in industrial assets. But what do these two trends mean and what can they do together?  Matt develops new innovative solutions in emerging technology spaces. He is responsible for evangelizing, designing and assuring success in new technologies, working closely with partners and flagship customers to define these categories. Matt brings over five years of technology experience to FogHorn Systems. Matt graduated with honors from the University of California, San Diego in 2012. He holds a BA in economics and a minor in political science.

Edge machine learning combines two hot industry trends – moving industrial internet of things (IIoT) compute to the edge of the network and the ability to model new efficiencies in industrial assets. But what do these two trends mean and what can they do together?

Many industry analysts believe the IIoT is the next billion-dollar industry. But while eliminating downtime, reducing scrap and creating safer working conditions represent staggering value in all industrial settings, the path to these types of optimization can be burdensome. Operations technology (OT) professionals on the floor are tasked with identifying the “known-knowns” every day, but what about the “known-unknowns” or, worse, the “unknown-unknowns?” And furthermore, if petabytes of industrial data are being produced on a regular basis and require real-time processing, the degree of difficulty for realizing value from IIoT could be prohibitive.

Machine learning represents a key component of IIoT, allowing companies to build models for condition monitoring, predictive maintenance and even real-time process optimization. Now the “known-unknowns” can be specified, and the “unknown-unknowns” can be isolated and corralled.  Once the algorithm model is built and applied on streaming data, new, previously untapped centers of business value are realized, including downtime that is never unplanned and higher yield counts. And when it’s combined with edge computing, this process becomes much more manageable and results in faster decisions, less ongoing bandwidth and storage cost, and no dependence on having persistent network connectivity.

Edge machine learning allows industrial companies to build complex algorithms that optimize machines, processes and even entire factories – all while eliminating crippling bandwidth and storage requirements. As an example, edge machine learning had to be utilized by a wind energy farm to build a predictive model around energy production – how much energy would the farm produce in a given week? This model is difficult to build, given all the live and historic variables, and exacerbated by the connectivity challenges a remote wind farm would likely face. Most importantly, there are significant costs to not being able to make an accurate energy prediction. Edge machine learning allows the wind farm operator to collect the relevant data from both streaming sensors and historical context to train this model on the edge compute node. Similarly, other companies applying edge machine learning might have a brand-new machine type and need to learn quickly how to prevent downtime, or want to build a specific algorithm for each individual machine to ensure the most tailored performance possible.

Edge machine learning actually trains these brand-new models and insights on the edge node itself. Advanced edge compute software requires a CEP (complex event processing) engine to handle these compute-intensive calculations. This CEP engine enables machine learning to be brought to the edge on any device, from a server even down to a computer that can fit in the palm of a hand. Additionally, this edge software must support common industrial machine learning model types to provide an appropriate solution for a wide variety of data.

The power of edge machine learning is in driving new efficiencies that a customer previously could not realize. Edge machine learning mines industrial data to create better outcomes, safer working conditions and significant cost savings. Industry experts know edge computing and machine learning are separately hot and important industry trends right now. Together, they can be doubly impactful.

 

Sign up for the free insideBIGDATA newsletter.

Leave a Comment

*

Resource Links: