Industrial service models are transforming, and service teams can no longer choose to sit on the sidelines. According to research firm IDC
, industrial service models are being transformed—away from traditional break/fix to proactive and predictive service. The business benefits are clear:
For the end-user: fixing equipment before it breaks increases uptime, productivity, and ROI.
For the supplier: reducing truck rolls by combining procedures on a schedule that’s convenient for the end-user and service team saves internal costs and enables better spare parts supply chain planning. Fixing equipment before it breaks also leads to reduced unscheduled downtime and increased customer satisfaction, resulting in higher renewal rates, lower churn, and better Net Promoter Scores (NPS).
Getting to the point where customer issues can reliably be predicted—and realizing those business benefits—is an achievable goal, particularly if a project uses all three components of predictive service simultaneously, not sequentially. While each of the three components is important and valid, they can be implemented together, or separately, to drive value, quickly and using all three ingredients together is what distinguishes a best in class practitioner.
Component #1: Traditional Analysis on Data from Connected Equipment in the Field
This is the approach that many companies use—and where they can get stuck, thinking it’s the only way to move forward. Companies with newly connected equipment often believe that the necessary data collection for proper analysis will require too much time and that a project ROI is too far away. Data from this process is incredibly useful and should be pursued, along with the other benefits of connectivity (alarming, alerting, remote service, etc.). However, connected data is just one part of an integrated approach; it does not have to be done first and does not have to slow implementation of the other two components.
The traditional approach to predictive analytics is well established: collect data, apply logic and achieve insight. It’s often that first step—collecting data—that creates a virtual obstacle to establishing a predictive service program and realizing the associated business benefits.
Why virtual? Companies assume that since they don’t have the requisite historical data on which to build models, predictive service is years away. The common beliefs:
- The connect/collect/analyze process is the only way to build a predictive program.
- All activities must follow that sequence.
- Since time is required to accumulate the necessary data, the program will not provide a timely return on investment
- These beliefs do not always reflect reality. Many suppliers have deep engineering foundations and that expertise can enable using the other components to build a predictive service program in a shorter timeframe.
Newly connected companies should absolutely continue the data collection process and build that data set. This strategy immediately supports the other two components, outlined below.
Component #2: Predict Using Existing Engineering Data
Don’t be put off by the term “model.” Technical product companies have already collected extensive data sets based on years of design and testing work. This accumulated experience can absolutely be used to build expected performance models and to set alarm/alert conditions. For example, your engineering team may already know that if the temperature or vibration levels of a specific bearing exceeds a certain threshold, a failure mode is pending. Based on design standards, testing and years of field experience, this is a credible “digital twin” model. Prediction based on this type of model is valid and can get the predictive service project going, quickly. The fidelity of that model can be improved over time by combining real-time data collected from the field with Machine Learning.
Component #3: Predict Based on Simulation
There is a wealth of new simulation technology available and accessible to engineers. Through PTC’s strategic alliance with ANYSYS, companies can now build simulation-based digital twins by combining ANSYS Twin Builder with PTC’s ThingWorx Analytics to predict how equipment operates and responds to its environment. The PTC/ANSYS framework facilitates creation of an accurate, physics-based digital twin, and as before, model fidelity can be tuned with a combination of real-time data and Machine Learning.
Consider the Uptime program created by Howden, a global engineering business. According to Maria Wilson, who is the Global Leader for Data-driven Advantage at Howden:
“Our Digital Twin builds on our core experience: over 160 years of designing and manufacturing rotating equipment that serves pretty much every air and gas handling application. This core experience is captured in the design data from our manuals, data books, selection software, drawings and models to create what we call a theoretical performance map. This theoretical data set is then compared with operational data from the sensors deployed in the field at the equipment level.”
Howden’s Uptime program is built on that OEM experience, combined with partnerships with PTC and Microsoft. According to Howden
, Uptime combines digital twin simulation and real-time data such that
“ . . . Data intelligence and visibility allow proactive and pre-emptive action to be taken to avoid failure.”