September-October 2017 NPJ

36 NuclearPlantJournal.com Nuclear Plant Journal, September-October 2017 4. What are the measures in place with the help of Predix, which will help the plants know well in advance that my pump is going to be in trouble, my steam generator is going to shut down, or my turbine or hydraulic system is going to be in jeopardy? Predicting equipment failures is not necessarily new. For example, advanced pattern recognition (APR) software tools have been around for many years now. These APR scenarios look at groups of existing sensors to identify deviations from a normal pattern. This was a great improvement over threshold alarms that are standard in a digital control system, because with APR you get an alert of an abnormal condition typically before an alarm threshold is breached. Where a threshold alarm would give you minutes or hours to respond, APR tools can typically provide a few days of warning to respond. For us, within the Predix platform, what we’re enabling in our asset performance management layer is what we call analytics orchestration. This is the ability to stack analytics together, to then have more than one vote or more than one input about that system and about that alarm. For us, SmartSignal was our APR solution set and it will still be a base analytic. But then, we start adding the machine learning and the neural nets and all these new data science techniques onto that same asset, and now you start getting multiple items looking at the same system or even the same component, but in a different way, in a different format, looking for different types of failure signatures. This technique has shown that we can pull that alert time back from days with traditional APR scenarios to greater than a month in some cases. Additionally, Predix and the APM modules are allowing our customers to build and connect analytics based on their operating experience. So now with analytics orchestration you can start to have peripheral triggers. For example, upstream and downstream systems are now providing inputs to primary systems, where previously APR scenarios really only looked at a single asset and its own sensors. Predix also allows us to start using non-traditional data sets in our asset health monitoring concepts, such as volume of work orders on a specific asset, again, something traditional APR solutions could not achieve or leverage. This then leads us into some of the Intelligent Asset Strategy concepts and Plant Performance... ( Continued from page 35) models. Within the Predix APM solution we start creating risk strategies for each asset, based on the relative failure modes of that asset. Analytics as mentioned above mitigate some of the failure modes, but not all of them – or maybe a unit does not have all the sensors needed to use some of the analytics. The Predix APM Asset Strategy Optimization pillar helps system owners plan and assess what level of risk is acceptable, and again tying IT systems to OT strategies, we help these owners by pulling in the financial impact of their decisions. That becomes a really powerful tool, when you can see both the risk and financial impact of changing a preventative maintenance frequency or a specific overhaul with a single click. It is even more powerful when you link different sites, and teams can start to see what their counterparts are doing at a sister plant, or on a similar asset. Now you really can begin to see who runs the best strategy and in many cases, you can also see what they are doing different. With Predix and APM we’re trying to enable that whole picture to bemanaged by a single person, a single pane of glass view that looks at your whole intelligent asset strategy, and then allow rest of your organization to also see the strategy, to ensure continuity and transparency is maintained. That’s what the Predix toolset really enables, over and above just somebody identifying a new analytic. What specific failure modes does that new analytic mitigate? How does it affect the overall maintenance strategy of your asset? What is the long-term cost of deploying that analytic and changing your maintenance strategy because of it and then where should you leverage that new strategy into other assets within your plant or fleet? That’s the single pane of glass that we’re assembling, so those decisions can be made in a simplified and expeditious manner. Contact: Jonathan Allen, GE Hitachi Nuclear Energy, telephone: (910) 819- 2581, email: jonathan.allen1@ge.com . GE’s Predix platform connects machines, intelligence and people to drive operational and business outcomes.