New Analytics Approaches for the Industrial Internet of Things

I attended my first ARC forum a few weeks ago.  Having been immersed in presentations and panels for three days though, there was something almost Schrödinger’s cat-like in the air about the Industrial Internet of Things (IIoT).  The two apparently contradictory states are:

1. IIoT is real!!!
2. Oh. I thought there was more to IIoT than that…?

So I definitely get the sense that people are less skeptical than last year about the Industrial Internet of Things.  And yet simultaneously, some people seem disappointed by the progress so far.

Let me suggest an answer to this conundrum:  While the Industrial Internet of Things is a revolutionary concept, it will be incremental in execution.  Or, as I often advise clients about big data:  Big data isn’t a thing, it’s a journey.

The session I hosted, New Analytics Approaches for the Industrial Internet of Things provides a prime example.  Three strong case studies – all with solid business benefits – taking those early steps analyzing data from the Industrial Internet of Things:

​Hedi Ago from the Orlando Utilities Commission presented on the OUC’s smart meter project.  With 190,000 electricity customers and 110,000 water customers, smart meters are bringing a number of benefits – to both OUC and its customers.  Consumer benefits include rapid re-connection of service – so rapid that if a customer uses a payment station to settle a bill, the power is often back on before they get home (7-12 minutes).  Customers can also monitor their usage via dashboards and visualizations accessible from mobile devices (although not in real-time).  For OUC, the most obvious benefit is avoiding the cost of reading meters, ad infinitum.  However, analysis can also reveal which customers are watering the yard on days that they shouldn’t be (during a shortage), and also detect and eliminate fraud.  You can see full details of the cost and time savings in the slide below.

Click to Enlarge

Adam South spoke about Kennametal’s use of machine tool data and complex event processing to increase productivity in discrete manufacturing.  Kennametal was faced with talent erosion, as well as the increased complexity of both tools and manufacturing processes.  And, the expectations of the Facebook generation – that is, data at their fingertips.  The overall goal though was to reduce the time from art-to-part.  Adam noted that the traditional approach to increasing productivity is to reduce downtime.  The new approach was focused on wringing out more productivity by reducing cycle time.  The solution uses complex event processing software to gather and analyze production data in real-time (see diagram below).

Click to Enlarge

This helps Kennametal to understand which operators are out-performing the production plan – and crucially, to guide less experienced operators to improve.  For example, in one machining operation, it was discovered that taking a faster, shallower cut reduced cycle time by 16%, compared to the slower deeper cut that the production plan called for.  Overall, best practices have been shown to reduce cycle time by 20-40%.

Scott Abramson presented Duke Energy‘s use of analytics with their wind genera​​tion project.  The wind power generation business is centered on long term contracts to sell power to the grid.  With that in mind, Duke Energy purchases wind turbines and expects to amortize them over more than 20 years of productive life.  Ensuring that turbines live a long and productive life is therefore critical to the financial performance of the business.  A significant event in the life of the turbines is when their warranty period expires.  Since the original equipment manufacturer covered the warranty, Duke had no visibility into the part of a turbines history.  However, Duke did have strong data from assets in the field and instigated an analytics program to reduce unplanned downtime and optimize asset productivity.  This involved integrating structured and unstructured data and matching data from an historian to work orders.

Three different use cases, all aggregating detailed, low-level data – Industrial Internet data.  So, yes the Industrial Internet of Things is kicking into life.  And nobody should be disappointed by the scale of these projects.  The arrival of the Industrial Internet was always going to be incremental.

However, it’s clear from several conversation with clients that more details on how IIoT analytics projects actually work are desired.  Presently, it’s a bit of a mystery how sensor data is collected, managed and turned into information.  And that’s no criticism of the three case studies presented – each presenter only had 20 minutes and were instructed to focus on business challenges and benefits, not technology or implementation.

So upcoming research from ARC will take a deeper look at how Industrial Internet analytics projects are implemented, from raw data through to insight.  I expect to highlight many challenges that are similar to those that have bedeviled business intelligence and analytics projects for decades:

• How to decide which metrics and key performance indicators to present to management;
• How to build-in the flexibility and agility necessary to deal with future changes – and unexpected questions;
• How to deal with dirty and incomplete data;
• How to integrate and aggregate data from a large number – and wide range – of data sources

Other challenges – such as how to manage large data volumes and high velocity data – are big data problems, but we’ll cover those over the course of 2015 too.​

David White is ARC’s expert on Big Data, analytics, and Business Intelligence.