Ever wonder what it’s like to stop a problem dead in its tracks? Some people do it for a living.
At Uptake, we like to think of them as Digital MacGyvers — talented individuals who are incredibly resourceful, and who are quick to adapt to anything that comes their way. For them, no problem is unsolvable. There are only chances for creativity. They believe there’s an art to conquering industrial-scale challenges.
But don’t just take our word for it. Be sure to check out the first installment of Model Behavior — Predicting and Preventing Failure on the Front Lines of Heavy Industry — for an inside look at how Uptake’s industrial data scientists are powering a brighter future for wind energy.
Solving Industrial-Scale Problems Takes Industrial-Grade Expertise
You’re not alone if buzzwords make your head spin. And when it comes to AI, there are plenty of them out there. To help cut through the noise, here are a few baseline definitions to get on board with:
A machine learning model is a codified structure and digital representation of an industrial asset that physically exists in the real world. It organizes the data generated by the asset and standardizes how that information relates to each other — and, importantly, how all of it relates to the physical properties of the actual asset. A model targets a specific component of a specific asset, growing increasingly accurate over time with each new data point.
At Uptake, our machine learning models are the output of our AI engines, which are a tool that automates the development and deployment of our models. Our engines use valuable insights gained from different industrial use cases to come up with approaches that solve common sets of problems.
A catch is when one of our models — running in a live production environment for our customers — generates an alert that correctly warns of a failure or performance issue that’s bound to happen. Or, in other words: A catch means our model did what it set out to do.
A failure happens when an asset or one of its components shuts down unexpectedly in its operating environment, causing problems like unplanned downtime, lost production, service disruptions, unsafe working conditions and more. Failures are what models try to predict and prevent from happening.
Now Here’s the Catch
Uptake industrial data scientist Stephanie Kirmer works hard to help keep locomotives on track. As a member of our Rail Team, Stephanie makes catches for our customers’ locomotives that, if left untreated, can result in unplanned downtime, outages and delays. The goal is to:
- On the track: Improve mission success.
- In the shop: Complete the right repair fast, the first time.
- At the yard: Ensure mission readiness.
One of the top challenges Stephanie spends her time solving is improving the fuel efficiency of locomotives. In this example, Stephanie’s model identified a failure in the Exhaust Gas Recirculator (EGR) of a customer’s diesel locomotive.
It probably comes as no surprise that diesel locomotives produce a lot of exhaust. But what you might not know is that if it weren’t for systems like the EGR, that exhaust would otherwise be emitted into the environment. Thankfully, the EGR is specifically designed to recirculate the gas back into the engine and run it through again.
Unfortunately, there’s a big issue with EGRs. They’re susceptible to cracking. It’s due to the fact that extremely hot air first passes out of the engine and is then cooled down very quickly. That rapid change in temperature — combined with the high pressure levels that build and build — can cause cracking to happen.
But here’s the good news: This is precisely where our model comes in. Using data from multiple sensors along the engine path that the gas follows, our model is able to accomplish the following:
- Identify the expected pressure versus the actual pressure that’s being read off of sensors at different points of the path. This makes it possible to understand how gas is moving through the system.
- Use temperature sensor data throughout the process of gas moving through the system. This is how our model knows where the temperature is expected to get very hot and where it’s expected to cool down. Our model is able to distinguish between those points.
- Identify when a crack is likely to occur by looking at coolant levels. When a crack starts to happen, coolant begins to pass into the exhaust system and join with the hot gas, which increases the temperature of the coolant.
Using the above data, our model was able to identify the failure signature — or the point at which the failure is imminent, including the corresponding signals — of the EGR breakdown on our customer’s diesel locomotives.
Because our model successfully predicted those failures days and even weeks in advance, we gave our customer the ability to perform the right preventive maintenance tasks to address those issues early on and stop catastrophes from happening.
What does that look like on the front lines?
Our model helped prevent locomotive engines from failing and suffering drastic damage on the tracks. Not only do catastrophic failures compromise safety and cause backups in rail traffic, they can even require the engine to be replaced entirely — a significant cost that can skyrocket to as much as $500,000 per engine when including the cost of labor and downtime.
Today, our model informs our customer about what is determined to be the physical manifestations of the failures it detects. That saves the business time, money and resources while upholding its commitment to safe and reliable operations. Furthermore, our customer uses information from our model to perform more cost-effective maintenance tasks, and also to inform the engineering of better future systems so that the risk of repeat issues is mitigated.
Ask Me About My Failures
Check out this video to hear Stephanie explain the catch in her own words:
