A proof-of-concept system has been created by the Southwest Research Institute (SwRI) to autonomously identify compressed air leaks on trains and communicate their position to maintenance staff. The automated method may decrease the amount of time, money, and manpower required to locate and fix air leaks as well as the overall fuel and exhaust emissions produced by the locomotive industry.
Compressed air is used by trains for several purposes, such as air brakes, valve actuation, radiator shutters, horns, and bells. According to estimates, air leaks in trains that happen at various locations cost the rail sector between 2 and 3 percent of its annual vehicle efficiency. Additionally, the operability and safety of trains may suffer as a result of these leaks.
According to SwRI Lead Engineer Christopher Stoos, “air leaks significantly increase fuel consumption and decrease the effectiveness of a locomotive’s automatic engine stop-start (AESS) systems, which causes locomotives to run more frequently, burn more fuel, and shorten the lifespan of parts like starters, air compressors, and batteries. We may possibly save millions of gallons of fuel while lowering emissions of carbon dioxide, nitrous oxide, and particulate matter.
Currently, identifying air leaks requires railroad workers to do so manually. They frequently climb on, underneath, or between train cars to feel or listen for leaks. The procedure is ineffective, time-consuming, and puts mechanical staff members at unnecessary risk. In light of this, allowable air leak rates for trains have been set by the Federal Railroad Administration and railways.
SwRI has developed a system that autonomously detects, identifies, and reports air leaks, even on moving trains, using auditory detection technologies, cameras, and machine learning.
Leading the study are Stoos, Senior Research Engineer Heath Spidle, and Research Engineer Jake A. Janssen. It is financed by the TRB’s Rail Safety IDEA initiative.
The system makes use of a compact Fluke SV600 fixed acoustic imager, which is commercially available. It has a 64-microphone array and a camera tuned to detect frequencies between 30 and 45 kHz, which are the frequencies at which compressed air leaks stand out the best from the majority of background noisethe frequencies at which compressed air leaks stand out the best from the majority of background noise. Together, this device and a secondary visual-spectrum camera operate. The team trained and used machine learning algorithms to identify air leaks from the sensor outputs while disregarding non-leak-related outputs to automate the detection procedure.
With a false-positive rate of just 0.03%, the prototype system effectively identified a variety of air leaks at various locations on locomotives during testing. On a moving train, the system discovered 11 out of every 13 leaks on average. Once an air leak was discovered, the necessary employees were sent an alert with an image that showed the region that required inspection and repair.
According to Stoos, “the technology should lessen the workload for mechanical staff and boost the effectiveness of the compressed air system.” “If implemented properly, this system could possibly save the locomotive industry millions of dollars in fuel savings and maintenance costs.” However, further field research and testing are still required. “By increasing locomotive fuel efficiency, this technology may also significantly lower greenhouse gas emissions.”
