Mining Publication: Technology News 456 - A Fault Detection Neural Network for DC Trolley System Protection
DC trolley haulage systems move personnel supplies, and coal in approximately 50 U.S. mines. A suspended trolley line energized at 300 or 600 V dc provides electrical power, and a system of steel track serves as the return path. When roof falls and other events force the trolley line down near the ground return rail, inductance inherent in the trolley system facilitates continued current flow along an ionized path between the line and rail. This releases a significant amount of energy in the arc and may damage and/or ignite surrounding material. Conventional circuit breaker systems cannot prevent this continued arcing because the magnitudes of the currents involved may be significantly less that typical breaker trip settings. In 1980, the former U.S. Bureau of Mines demonstrated research to detect arcing and other types of trolley faults. The system required an oscillator to superimpose a 3-kHz signal in the trolley line, a signal wire suspended parallel to the trolley line fore circuit breaker coordination, and a filtering system on locomotives larger than 25 tons. Although the system functioned satisfactorily, the coal industry did not adopt it because of the complexity and cost of the hardware. Today, the threat of trolley fault-induced fires still exists. The National Institute for Occupational Safety and Health (NIOSH) has sought solutions that would require minimal hardware maintenance and be cost-effective. Using an artificial neural network (ANN) based system to detect trolley faults would require no modification of the trolley rectifier, line, feeder, or its vehicles, lessening maintenance concerns and costs. Further, all hardware used in the development of this system is commercially available.