Data in the Industrial World

Originally published at the Maana blog

According to IDC, more than 50 billion assets will connect to the Internet by 2020. The data these assets generate will lead to vital operational insights and untold business value—but only if the data can be effectively captured, analyzed, and acted upon. GE Digital is a pioneer in transforming industries with analytics and the Industrial Internet of Things (IIoT). 

Starting with the basics, what is the Industrial Internet of Things (IIoT)?

It is a collective network of intelligent machines working smartly together. Similar to the consumer Internet of Things, GE’s customers want to use data to learn everything about their machine in order to optimize performance and reduce downtime. Collecting data from the machines allows us to create equipment such as an aircraft engine that is able to manage its own maintenance. An airline can avoid a multitude of issues from flight performance to flight delays by gathering insights from the data in the engines, and proactively managing when it might fail, – all of which effect consumers and business performance.

At the end of the day, the Industrial Internet is about optimizing machines, reducing downtime and how it all translates into business value and increasing profitability.

 What does data look like in the Industrial World?

Industrial big data is different from what most people think when they hear the term Big Data. The key focus is on machine operations – optimization, unplanned downtime, etc. In the consumer domain, Big Data is primarily regarded as the digital footprint of a business and/or consumer that is left behind in the form of web logs, images and videos.

Let me explain some of the characteristics of Industrial Big Data:

• It is huge. While we think of Big Data as meaning “volume”, the amount of data produced in central networks by industrial equipment is huge. At GE, it used to be that we would only get three snapshots of a jet engine performance during flight. It didn’t matter if it was a 1 hour or 12-hour flight, we were only able to get data from takeoff, climb and during cruise. Now, we can capture up to 3000 data snapshots at 1/16 of a second intervals. So you can imagine a 12-hour flight generates an enormous volume of data we now can analyze – more than a terabyte just from the engines. Compared to the entirety of the consumer internet which generates terabytes of data per day, the Industrial Internet of Things generates petabytes daily. GE specifically has millions of assets worldwide; from jet engines to locomotives to gas turbines producing data we can tap into for business value.
• It varies. Besides volume, there is also a wide variety of data coming in about each machine. Since GE takes machines from crib to grave, we have a whole spectrum of data to leverage, including design images, manufacturing, time series, and quality control.

It is older than you think. Before the data-driven companies like Google, Facebook and LinkedIn were even conceived, the industrial world, through machines built and operated by companies like GE, was producing data – more than half a century of data gathered from millions of machines and sensors. 

What is a good data strategy?

From an industrial company perspective, it starts with data integration: reduce or, ideally, eliminate the data silo. It is important to understand how disparate information and data still exists today for many industrial companies that have focused on manufacturing and servicing of big equipment. In most cases, they have thousands of data silos operating across a single business unit and multiple ERPs feeding into the same piece of equipment. None of the systems talk to each other, so a good data strategy must start with leveraging data from across silos and lead to data integration.

Second, ensuring data quality. When I joined GE I came from the financial services space and I had a preconceived notion that data quality would be “near perfect” because data was being generated directly by machines with little outside interference from human input. I was quickly proven wrong, but the challenges were not directly related to human interference. For instance, data gaps are present in many industrial applications. Consider the jet engine I mentioned earlier. It transmits information from a remote location as it is flying to a secure destination. However, not all the data will make it through. There will be gaps as a result of network latency or poor connections. Those gaps are crucial for us to have impactful analytics to build our algorithms. Combine these quality issues with the volume of data transmitted from the machines and sensors, it is important to collate, cleanse and prep the data before the analytics are built.

Finally, when it is time to build the analytics, the focus must be on what business value can be gained: how analytics are going to solve a customer pain point like unplanned downtime. It’s ultimately about increasing asset performance and efficiency, which improves the business profitability. Approaching it this way, many unexpected and beneficial discoveries can and will be made. In the case of monitoring a machine’s performance and applying analytics, we can learn when and why parts fail and not only fix the problem, but influence the next generation with improved design. It is huge advantage.

Analytics focused on business value can also help get the right sensors put in place to build better algorithms. Sometimes we don’t see correlations in the data so we learn where to place the sensors to gather the right data resulting in us building better algorithms. In doing this, we are able to offer improvements in our products that we can share with our customers who can benefit from these insights.

Read more at the Maana blog.

-------------------------------------------------------------------------------------------------------

Note: Opinions are my own and not related to any of the organizations or institutions that I am or have been associated with.