Applying data science to specific aspects of the rail industry can provide an opportunity to optimize maintenance-of-way operations. When it comes to detecting rail flaws or identifying worn-out components, track inspection is increasingly being automated. Traditional track geometry and rail-profile data is being augmented with video images and machine vision systems that can learn from and analyze data to produce predictive maintenance models.
NLP Logix LLC has been building and hosting machine learning predictive models for a variety of industries, and the Jacksonville, Fla., company recently released a video showcasing an application of “deep learning” in the railroad industry.
“The video was in response to a contest put on by NVIDIA, a manufacturer of graphic processing units,” explained Theodore (Ted) Willich, chief executive officer of NLP Logix. “We have been working with a Class I railroad operator to automate car-brake and track inspections using video, and using deep learning techniques to train the computers to identify inspection items. The railroad gave us a subject matter expert to work with, and he said that 90 percent of Class I inspections on inbound trains could be done with vision systems.”
For this application, the railroad was looking for a partner company with experience in both computer vision and tabular data, someone to create computer models that could identify points that needed further human inspection.
“The railroad wanted us to create a system that could look at machine-vision-created images, and use other data sources to enhance understanding,” Willich said.
Such a process takes a team full of specialists, which is what NPL Logix provided:
• Machine-learning experts set up algorithms to train the models to understand what they’re “seeing.”
• Statisticians interpret the output of the algorithms.
• Business intelligence experts to collect and merge the data.
By building the data models, “there’s a huge opportunity to optimize [MOW] operations,” said Willich, adding that it’s already happening in other industries. He cited the example of energy modeling in the facilities market.
Energy modeling lets customers buy and sell megawatts of energy from and to the energy grid, Willich said. NLP Logix creates algorithms that model the effects of weather, congestion and load on the use and cost of energy. Using such models, customers can be much more proactive about what they spend on energy (how much and at what rate), when they schedule machines to go offline, when they can sell energy back to the utilities, etc.
“That gives them a competitive advantage,” said Willich.
The same is potentially true for railroads. Data science is enabling better scheduling of maintenance tasks, optimization of human and machine resources, and a lot more—as evidenced by the presenters at the recent Big Data in Rail Maintenance Conference.
“This stuff is exploding,” said Willich. If you’re not doing this stuff, and starting to do it well, the competitive advantage is going to evaporate.”
His advice: “Take little bites. Build on successes. You just don’t click your heels and get it done. You’ve to have leadership, sponsorship at the executive level, and some good people helping you. Data science is a team sport.”
To see the original article please click here: