Online Fatigue Estimation and Prediction of Switching Device in Urban Railway Traction Converter Based on Current Recognition and Gray Model

• Published in: IEEE access Vol. 7; pp. 123307 - 123319
• Main Authors: Wang, Lei, Xu, Weihua, Liu, Shenyi, Qiu, Ruichang, Xu, Chunmei
• Format: Journal Article
• Language: English
• Published: Piscataway IEEE 2019; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Accelerated tests; Converters; Electric railroads; Equivalence; Estimation; Fatigue; fatigue estimation; Fatigue tests; Insulated gate bipolar transistors; Junctions; lifetime prediction; Railcars; Semiconductor devices; Service life; Switches; Switchgear; Switching; Switching circuits; Switching device; Thermal simulation; Traction; traction converter; Traction motors
• ISSN: 2169-3536; 2169-3536
• DOI: 10.1109/ACCESS.2019.2936259

It is difficult to estimate and predict the fatigue state of switchgear in traction converter of urban Electric Multiple Unit (EMU). At the same time, this is just the purpose of this paper. Firstly, the power loss and junction temperature of the device are analyzed by using the results of electro-thermal simulation and actual working conditions. Then, a novel Bi-directional accelerated fatigue test (BAFT) is proposed, and the data obtained from BAFT are used to establish a fitting fatigue model for a specific switch device. BAFT data enable the fatigue interaction between Insulated Gate Bipolar Transistor (IGBT) and Free Wheeling Diode (FWD) to be displayed in the common switching device module, and the interaction is represented by an acceleration factor. In the fatigue model, the service life and fatigue state of the equipment is related to the equivalent fatigue current which is used to generate the fatigue. The equivalent fatigue current is observed by the proposed analysis identification model. The historical data of fatigue values are subjected to Gray Model (GM) (2, 1) for trend prediction, and then the service time of the switch device is calculated using the predicted fatigue values from GM (2, 1). The examples presented in this paper are based on recorded field data.