Why will 2018 be the year that Predictive Analytics moves from theory to practice? Talk is cheap, as the saying goes, but data infrastructure and sensor installations aren’t.
Compare 2018 to 2008, and the differences are clear. Today we have those two crucial elements – sensors that produce data, and the analytics tools that make sense of it — more widely distributed across the industrial spectrum:
- More sensors that are built into rotating equipment that have the ability to output high-frequency telemetry data (6 million samples per second!) Industrial equipment manufacturing cycles are slow, so it has taken a decade before new systems with high-frequency sensors could be installed
- Better data access platforms that can aggregate and apply algorithms to that data. When data from disparate sources is organized into a single virtual “analytics bench,” incredible business insights can be achieved.
The race is on to see which business will get to those insights first and gain a competitive advantage. These predictive analytics initiatives are most likely to be found in several key vertical markets, including power generation (gas turbine power plant or wind farm operators), aviation and aerospace, and large EPCs.
Let’s cover a few use cases that might apply in each market:
There are many applications for predictive analytics in the complex power gen cycle, but let’s focus on a digital twin selling tool. When a customer that wants to build a new power plant engages a large firm like GE and asks for a bid, there may be hundreds of different types of configuration for that plant. The optimal configuration for that plant depends on balancing models (through a model orchestration process) that touch on cost, performance, lifing and other environmental factors. Running models for these bid proposals took hundreds of engineer hours in the past, but an efficient digital twin tool can produce several recommended configurations within an hour.
Wind farms are only as good as the health of their turbines. Wind farm operators can save millions of dollars by improving their condition-based and predictive maintenance regimes. In other words, saving even half a day of downtime (multiplied by hundreds or thousands of turbines over a year) could significantly improve the bottom line. Anomaly detection algorithms that compare test cell data with actual field data can trigger alerts that tell maintenance crews where they are needed in real time.
Gearbox maintenance has bedeviled airplane makers for decades. A part that does well on a test cell might cause catastrophic failure in the field under the wrong conditions. Predicting gearbox failure is now possible if every gearbox has sensors that relay real-time data back to the manufacturer. Predictive analytics solutions can pair that field data with what was observed on the test cell and see metallurgical weakness before it becomes a serious problem.
It’s the Year of the Dog, according to the Chinese Zodiac (happy New Year’s). At Peaxy we’re calling it the Year of the Sensor.