• M/W

Precision-Scheduled Maintenance

Written by William C. Vantuono, Editor-in-Chief
Linear asset management is a rapidly evolving area of m/w that increasingly relies on autonomous or semi-autonomous data collection systems, coupled with Big Data-driven analytics. (VisioStack)

Linear asset management is a rapidly evolving area of m/w that increasingly relies on autonomous or semi-autonomous data collection systems, coupled with Big Data-driven analytics. (VisioStack)

CSX is using advanced analytics and autonomous data collection systems in maintenance-of-way tasks. Here's how.

RAILWAY AGE MARCH 2021 ISSUE: Linear asset management is a high-tech term that defines what goes into railroad m/w. It’s a rapidly evolving area that increasingly relies on autonomous or semi-autonomous data collection systems, coupled with Big Data-driven analytics.

Among the many railroad/supplier collaborative efforts is one involving CSX and Greenville, S.C.-based VisioStack, a developer of condition monitoring, data analytics and decision support systems. The company’s rail-centric offerings are RAILLINKS® AI and RAILLINKS® PREDICT.

CSX Director of Track Testing Brad Spencer

As CSX’s Director of Track Testing, Brad Spencer has a lot under his umbrella. “There are many ways to conduct testing,” he says. “We use autonomous boxcars, our ATAC (Autonomous Track Assessment Cars) system, because they’re a bit more modular, and we’re not using a resource that’s quite as restrictive as a locomotive, because a lot of these platforms are still under development. We have not used machine vision until this year, but we conduct frequent testing, which is critical for doing predictive analytics and deep learning. The foundation of autonomous systems is data management, where we can synchronize all our different data platforms. There’s no better way than visualization to get information out in the field. We use various system providers and technologies. If there’s no way to marry and sync all that data together, it doesn’t matter how many times you collect it.”

“We started working in November 2018 with VisioStack,” explains Spencer. “We began with a small project, looking at curve alignment issues on one specific route, and we were in discussions with several companies about data management. We liked VisioStack’s platform, and wanted to see what they could do with it. We were pretty impressed with that initial project. It only took a few weeks, but it gave us some idea of what to look forward to and what their potential is.”

“We’ve been working on this technology for about 10 years, when we started seeing a trend toward autonomous data collection systems,” says VisioStack President and CEO Zachary G. Garner. “This meant there was going to be a lot more inundation of data. We set out to build a Cloud-based platform that can handle any type of railway asset or condition data, meaning track geometry, rail wear or linear sample data, as we call it, point cloud data—whether that’s LIDAR, or rail profile, or any type of imagery data as well as defects. Defects come from many different systems.

VisioStack President and CEO Zachary G. Garner

“We’ve really been working on workflows, which means whatever data we get, we try to establish an automated process by which decisions are made. So it’s very critical, if we’re going toward autonomous data collection systems, that we have business rules and automated processes in place that allow us to validate data quality and to perform various flexible tasks on the data, and then ultimately create actionable items off that data. Workflows are a key element of our RAILLINKS® platform. And we have to be flexible to what a particular client needs. CSX is a big organization, with many different requirements for the various types of data it collects.

“Data visualization is really important, especially when we’re collecting more data than we’ve ever collected before, not just in this industry but across industries. If we don’t understand what we’re working with, then we can’t make real step-change improvements on our business processes. We’re good at data visualization. And when you’re good at that part, you can start understanding data on a higher level than you could before. 

“If you’re swamped with the data you’re collecting, you’re still swimming, not knowing where you’re going. But if you can start to break it down and to visualize it at the macro and micro levels, then you really have a good understanding of what you need to do and where you’re going. That’s a key foundation of putting in place analytics that can work off of that data. Data always is going to tell a story. And until you can have a strong platform that allows you to visualize the various parts, you won’t be able to segue to the analytical phase. We’re starting to head over into the analytical side and be able to squeeze value from the data. A lot of money is spent on data collection systems.”

Degradation: This shows several track geometry runs, aligned in a single location, 
that are trending poorly.

Spot Maintenance to CAPEX Planning

As an example of improved spot maintenance, a track geometry profile can show a surface defect, like a dip or a hump. “Let’s say we are looking at a dip, running an autonomous test vehicle over this defect two to three times a week,” explains Spencer. “We establish a timeframe and look at all the runs across there. With the VisioStack software, we can quickly just click on that section of the curve, because every run is visualized with a different color. We can select a spot in the curve and hit ‘trending,’ and it’ll actually show you what the trend is over time. And you can see that profile and determine that the dip is getting bigger. We repair it, and then it goes right back to zero.

“If you really want to do some predictive things, you could say, ‘I don’t have a tamper freed up on the Columbia Subdivision, but I can see that this isn’t going to go very well.’ And I’ve got an idea how much time I’ve got with the current tonnage on that line. We’ll still send the inspectors out and have them focus on those areas, but we can predict when we can get our tamper to that area. So it’s very key for predictive analytics when you’re applying it to that type of situation. But on a bigger picture, on the entire subdivision, taking the data and actually putting real values to it, real data behind it, and being fact-based, we can decide how to prioritize the whole network. That is based on tonnage, speed, and all the risk characteristics. Without this kind of a data management system, we’re collecting data and not really using it to its full capacity. That’s what we’re finding most important with VisioStack.

Normalized Change Detection: This shows the predictive side of the platform, which finds the “needle in the haystack,” those areas of interest that should be followed even before they become defects. NCD is all about early detection and prevention.

“One of the really important drivers is finding repeat exceptions that drive permanent fixes. VisioStack has taken tabular reports and visualized them inside their platform. So when roadmasters look at strip charts from geometry testing, they can just click on a spot and determine very quickly where repeat sheets are, how they were addressed, did they come back within a certain amount of time, was the defect fixed correctly. It’s an excellent tool for the field. And having it in a visualization makes it so much clearer to those folks who are actually using the data on the ground. There are some big applications for this technology with the field people, helping them understand what’s going on.”

In terms of capital maintenance planning, “Cloud-based data management is having the technology where data flows go into the right places at the right time, enabling quick decisions,” says Spencer. “Our capital plan has become more dependent on data, vs. the old days when we had people making decisions that weren’t always the smartest when you’re looking at the whole network. Now, we’re able to make many more informed decisions, and lower our risk. If you look at our derailments, we are driving those numbers down every year. This technology is another means to take us to the next level.” 

For more, read “No Longer Atypical,” the Railway Age November 2021 feature that covers CN’s Autonomous Track Inspection Program (ATIP), using railcars equipped with sensors and artificial intelligence that automatically scan and analyze tracks while operating in revenue-service freight trains.

Tags: , ,