Designing dropper data with artificial neural networks in railway electrification catenary systems

In this thesis, determining data of dropper lengths in rail systems by means of artificial neural networks is discussed. Electrification systems consisting of energy transmission lines, traction power substations and their command and control units are required to operate electrical railway systems. It should be ensured that the heights of contact wires which the pantograph constantly contacts should be the same at every point of the line. In order to keep the contact wires stable at a specific height, a messenger wire which runs in the same line but above of the contact wire is installed. The contact wire is hung on the messenger wire with intermediate connection elements called dropper. The heights and locations of droppers should be carefully determined according to the deflection of the contact wire and the messenger wire, tensions of the both wires etc. While previously, the dropper heights were determined manually, they can be recently calculated with a few special software developed by some companies. In this study, as a new approach, artificial neural networks were trained with Matlab® software by means of the dropper data from previous railway projects in order to be able to produce new dropper data. Finally, after the analysis with a test data, it has been observed that the dropper lengths can be calculated automatically with a high accuracy.