Gary Wolf, principal at Wolf Railway Consulting, tells Trains that New York Air Brake’s Friday announcement that it started a stopped freight train, operated it on a 48-mile-run and then slowed and stopped it — all from a computer, is another step to efficiency and safety.
New York Air Brake achieved the feat using its LEADER train handling system on a test track with an unknown version of positive train control installed. LEADER’s prime competitor in North America is Wabtec’s Trip Optimizer.
Wolf says that automation improves the performance of less-skilled engineers, saving the railroad fuel and increasing safety by reducing the potential for human error. Whether it’s a human or a computer in charge, running a three-mile train over an undulating grade has many variables, including how much of the train is going up, how much is going down, and any upcoming changes in speed due to curves, speed restrictions, or signal indications.
Wolf says he once worked with a western railroad which studied the fuel efficiency of a group of bulk commodity trains, testing different engineers running the same consist as it cycled back and forth with the same locomotives, cars, and tonnage. Over the roster of engineers, he discovered a 15-20% variance in how much fuel each used to make the same run with the same train. Driver assist systems such as LEADER bring each engineer up to the same level of performance, which Wolf says is better than most engineers.
Asked if some engineers could beat the computer’s performance, Wolf says that that is possible, but that averaged over an entire crew roster, the computer would win every time. He related an experience working on computer-aided dispatching, where a railroad ran tests with a computer in charge one day and a human the next. The humans performed better a few days where they knew things the computer didn’t, but over the length of the test, the computer easily won.
“It all depends on what you feed the algorithm,” Wolf says.
Wolf acknowledged that railroads are a harsh operating environment for a computer, subject to transient voltages, arcing, temperature and humidity, oil, vibration, impact forces from rolling over track; and that when a computer crashes, the train will have to stop while it reboots or waits for a human to diagnose the problem.
He doesn’t see this as a deal-breaker, though, saying that solid-state computers are only getting more reliable. If a train is stopped by a broken air hose or knuckle, tripped relay, or computer crash, Wolf predicts the response will be the same as when a crew today can’t solve a mechanical problem: send out a mobile unit to wherever the train stopped. He noted that this will be easier on western railroads where service roads often parallel the track, while tracks are often less accessible in the East.
According to New York Air Brake, the current LEADER system is used on over 5,000 locomotives worldwide, controlling locomotives under an engineer’s supervision to reduce in-train forces and improve fuel consumption. The company claims that it improves fuel consumption between 6 and 17 percent, and states that fuel amounts to 14.7 percent of Class 1 operating costs.
To Greg Spindler – It’s amazing how fast “fanciful rubbish” turns into reality. Not this year or next year, but in ten or twenty years it’s almost guaranteed.
To Ian Narita – You ask what will a computer do when it comes across an unexpected situation – it will do just what a train engineer would do – stop the train.
“It all depends on what you feed the algorithm,” In other words, “garbage in, garbage out”. Why is sorting through the garbage?
Automation can be very helpful in certain instances, but the absence of Level 5 self-driving cars speaks volumes about the hype surrounding automation. As does the Boeing 737 Max debacle.
The notions of having robot trains in wide application in “the real world” outside of limited transit or ore runs is fanciful rubbish.
Copies of this report will be available for all at the next management/BLE renewal contract negotiation….
Gary Wolf, principal at Wolf Railway Consulting has made some very astute observations concerning the breakthrough of controlling trains by a computer. Where I differ with Mr. Wolf are his concerns about how the system would handle, a three-mile train, or a train going up a grade, or down a grade, and any changes in speed due to curves, speed restrictions, or signal indications. Let us start with a three-mile train. The length of a train is a variable. The length of a train is determined when the train is made up. This length will become part of the operating rules that would be loaded into the computer on the locomotive before it ever starts its run. These values will be checked by sensors that will determine the length of the train, as well as the weight loaded on every axle as the train leaves the yard. That data will be checked against the data loaded into the operating rules to confirm it is correct. The length and the weight of the train will be a known variable. The other concern is going up or down a grade. The topography of a railroad never changes, there for it is a constant. With Positive Train Control and GPS the location of the train is a known variable. It is also a known constant as to where the curves are. Speed restrictions are part of the operating rules. Signal indications are a known variable with the data coming from PTC. Add to this, the system will also use AI. The results of every trip will be recorded and the data will be added to the data base. The system will continue to learn.
“Wolf acknowledged that railroads are a harsh operating environment.” Modern locomotives already have computers on board. The “LEADER” will just be another computer. Computers that are used in harsh environments have been built for years. Let us consider a computer that controls a launch vehicle putting a satellite into orbit. The computer can be subjected to up to 9g and high amounts of vibration, and continue to function. As far as reliability is concerned, consider the flight control system of an Airbus A320. The flight control system uses seven separate computers to control flight. Two are dedicated as backup for the pilot and copilot computers. The system is redundant and does not fail. When LEADER is implemented, it will use a similar architecture.
Wasn’t Muskingum Electric RR in Ohio automated in 1968? It ran between a mine and a power plant until the coal played out.
All that comes to mind is to quote Darth Vader “this technological terror that you have created”…………..
The problem with computer automation is this you can program the computer to do this or do that when this or that happens.
How do you program a computer to handle the unexpected, like Harry Potter ‘s car hovering over the tracks or any of a million things that Murphy’s law predicts can go wrong.
Two words: Flying cars.