14 Arrowhead Drive
Laramie, WY 82070
Mon - Fri 8.00 - 18.00
Sat & Sun CLOSED
+1 (307) 761-3417
Mountain TIme

Big Data and Asset Management

Subtitle Text

Big Data and Asset Management

  • Posted by: Art Durnan
  • Category: Asset Management, Big Data

In my last submission I introduced myself and gave everyone an overview of my working life. Today I want to write about a really hot topic in industry and that is all this talk about “Big Data” especially as it relates to machinery reliability and health.

For the past few years I worked on a project for my former employer that culminated in an international assignment in Brisbane, Australia. The objective of the assignment was to build a Predictive Asset Health solution that could predict remaining useful economic life of a machine. Obviously, understanding the condition of a complex machine and its subsystems and many components requires information. Modern machines produce more data than any human can evaluate in the time a decision needs to be made. The adage, “Maintenance should be an information business,” is truer today than ever. I’ve always said, “Ignore the data and the machine becomes your boss.”

Data has been around for a long time but never in the volume the human race and machines are producing now. You would have heard statistics like more than 90% of the data that exists today was created in the last two years.

“In God we trust all others must bring data” -J.Edwards Deming.

Just as an aside, Deming lived in Powell Wyoming and graduated in Electrical Engineering from the University of Wyoming. As I write this I am less than three blocks from where Deming studied. That alone is inspirational!

IBM CEO Ginni Romety says “Big Data is the new natural resource”. IBM is bringing forth Watson Technology that is now becoming affordable. As part of a fact finding exercise on Asset Management systems, I joined a team that IBM facilitated. We spent 8 weeks on the road looking at high tech applications from rail to oil and gas, commercial aviation, marine, high tech manufacturing and spent a good amount of time in IBM laboratories. I even got to play against the original Watson Computer! What I learned really did open my eyes to the possibilities of using these techniques to better understand the machines that play such a vital role in our society.

You would all have heard about the Internet of Things a termed coined by British Entrepuner Kevin Ashton in 1999. Basically IoT is the collection of all connected devices.

So let’s talk about data as it relates to machinery health. First the 4 V’s of data.

Volume-big data is characterized by the amount of data present

Velocity-describes how fast the data is produced

Variability describes how different are the data sources, configuration etc.-numbers, words different units…


Veracity-how valid is the data. Data can lie to you just like a human does.

Volume and Velocity operate together to define the band width required to move the data, the amount of storage required and define how fast it needs to be processed. You would know about this if you have tried streaming real time vibration data for instance over a wireless system.

Variability is all about disparate data sources and they are for our purposes here, machine generated as from OEM systems, predictive technologies, production reporting, process parameters. Structured data from hand held device pick lists and defined fields in your CMMS. Unstructured data from comments on work orders, digitized images etc.

For years I worked under the assumption that more than one source of data could improve confidence when making a “call”.  We termed this an “Integrated Condition Assessment”. As far back as the early nineties predictive software started to allow the collection of two or more data sources such as oil and vibration but in practice, at least from my experience, not much was done to combine the data. In the early days network software was severely limited by band width and terribly unstable.

Even a couple of years ago we would knock a server over when we tried to push too much too fast.

What has happened now is a major step change in how data is handled. Cheap accessible server clusters in the cloud can expand and contract and offer fast computing. One major challenge is still working with 1980’s server architecture at our plants and mines. Eventually as applications migrate to the cloud seamless integration could occur. One challenge that I faced in my last job was actually accessing data that we already owned. Firewalls, unreasonable cyber security, fear of creating two versions of the truth all played into a very frustrating time. Even when our IS&T department wanted to help it was still difficult and in some cases took months to resolve a simple request.

However what were the learnings:

  1. Data is available and it is wasteful to ignore it. Operations will need to acquire and store more data than ever before. Those organizations that are successful will prosper.
  2. Sensors are cheap. We used to ask in an RCM exercise … Can it be monitored and is it worth doing? The answer today is more often than not absolutely yes and new machines come sensored anyway. You are paying for the data when you purchase the machine.
  3. Data can give insights we have never had before and will aid automated processes and provide decision support.

In summary to this section I just wanted to talk about how the world of maintenance and reliability is changing.  My next blog will talk in a little detail about prognosis-determining when a machine will fail, how to do it and my best understanding of the state of the art.

Here at XRT we are constantly striving to help you beat the iron. One sure way is to understand the condition of the machine.

‘Til Next Time – Good Luck!


Author: Art Durnan
With over 45 years of industry experience primarily in Asset Management, Art is an pioneer in Asset Management. Previously Art held positions in project management, engineering design, and process engineering in hydrometallurgical and pyrometallurgical processes.

Leave a Reply