Our company was the prime contractor on a railroad equipment contract that used small gas turbine engines for propulsion in a rail car. During predelivery testing on the third train set, we encountered an intermittent issue where one engine would shut down for over speed during the start cycle, typically at 60 percent of rated speed. It was always the secondary (redundant) over speed circuit, and we thought we had it solved with the replacement of a defective component.
Shortly after delivery, the problem resurfaced. Obviously, it had not been corrected, and it was still intermittent. After several weeks of further testing, we were still stumped. Whenever we captured a failure on the scope, we would see the frequency on the speed input suddenly jump but quickly return to the proper frequency for the magnetic pickup.
One lucky afternoon, it happened several times, so we had the opportunity to test a variety of conditions. We found it would not happen when 480-volt auxiliary power was off, but we could often repeat it when the aux power was on. Since the only relationship between aux power and the engine was the battery chargers, we tried a few starts with various battery chargers on or off. Sure enough, we had a source in the charger for the 28-volt starting batteries, but a series of prior modifications had already isolated the 28-volt starting circuit from the regulated 24-volt engine controls at every level, including separate batteries.
Eventually, we disconnected a fuel usage display that used a second coil in the same magnetic speed pickup, and we were able to make the problem go away. Again, the meter was connected to the regulated 24-volt system and had no connection to the starting batteries.
Going up the line, we started tracing the power supply wires for the fuel meter, looking for cross connections or damage. All we could see was a section about two feet long where some wires for the 24-volt system ran on the same tie bar as some sensing wires from the starting circuit. With nothing else to check, we separated the starting system wires from the fuel meter power wires, and we never had the problem again. As a precaution, we did the same separation on all the other units, though none of them showed the same problem.
In a post-mortem analysis, we found the battery charger for the 28-volt starting system used a 20kHz switching frequency and was designed to shut down during the start cycle. Once the starters de-energized, the battery charger would come back online at full current. The timing of this cycle meant the charger was always coming back online as the engine was accelerating past 60 percent.
The EMI from the 28-volt starter sensing wires to the regulated 24-volt system was being transmitted through the fuel metering unit to its own engine speed input. This was then passed through the isolated coils of the magnetic speed pickup like a 1:1 transformer. The other coil on the pickup was connected to the redundant over speed detector, which saw the now nicely filtered 20kHz as 120 percent of rated speed.
No other source was ever found, further testing validated all the components, and the problem never returned. So all we could determine is that the specific lengths of wire and natural variances between components provided the right conditions on this unit to amplify and filter the 20kHz switching noise into the over speed detector.
This entry was submitted by Ryan Flinton and edited by Rob Spiegel.
Ryan Flinton spent 12 years working for a contract manufacturer of locomotives and other rail equipment as both a test engineer and a program manager. Since 2009, he has been with a leading alternative energy-generating company as a regional manager and currently a field service engineer. He lives in upstate New York, where he and his wife own and operate a commercial dairy goat farm.
Tell us your experience in solving a knotty engineering problem. Send to Rob Spiegel for Sherlock Ohms.
Tweet This
4 
