Study on DC Bias Vibration Signal Separation and Experiment of Transformer Based on FAST-ICA Algorithm

Abstract

Transformer vibration research is an effective diagnosis of DC bias fault. The vibration signal acquired is actually the mix signal of vibration from both the core and windings, and this makes it difficult to analyze the above two separately. Therefore the paper presents a new approach to separate vibration signal of core and windings by Fast-ICA algorithm. The simulated results of core and windings vibration are given and compared with experiment test. Comparative analysis verify that transformer vibration in no-load is core vibration and vibration in short circuit is windings vibration. Transformer vibration signals under different DC bias are tested by experiment. Vibration signals are separated by Fast-ICA algorithm. Core vibration signal and windings vibration signal are respectively obtained. The research lays the groundwork for further research of transformer DC bias vibration.