Research On HHT Feature Extraction Of Contaminated Oil Signal

Transformer oil is easily mixed with pollutants such as particles and moisture during production,transportation and usege,which affects its physical and chemical properties,leading to the equipment breakdown and reducing the efficiency of the equipment.Therefore,grasping the dynamic motion characteristics of the contaminated oil in the pipeline is of great significance for extending the service life of the oil and ensuring the safe operation of the oil equipment.The dynamic characteristics of contaminated oil can be characterized by signals.Because it is a typical non-stationary and non-linear signal,the current methods for processing such signals include short-time Fourier transform?STFT?,wavelet transform?WT?,Hilbert-Huang Transform?HHT?and so on.However,the STFT and WT have difficulties such as the selection of window functions and wavelet bases,the HHT is not restricted by the basis functions.Therefore,this study proposed to use the HHT method to process and analyze the signal of the contaminated oil,in order to explore the internal relationship between oil signals with different degrees of pollution and the degree of oil pollution,so as to reveal the dynamic characteristics of pollutants in oil.First,a experimental system that can test the dynamic characteristic of oil was constructed.Meanwhile,16 groups of mixed oil samples containing Cu and SiO₂ particles and with different pollution levels were prepared and tested for pollution degree.Under the pressure conditions of 0.1-0.17MPa,the dynamic motion signals of oils with different pollution degrees were obtained.Secondly,the HHT was used to process the collected oil signals of 0.1-0.17MPa.The IMF component,amplitude-frequency diagram and Hilbert spectrum of the oil signal were analyzed.It was found that each IMF component had obvious frequency modulation characteristics,and the maximum peak frequency of the oil increased with increasing pressure.At the same time,the IMF components,amplitude-frequency diagrams and Hilbert spectrum of oil signals with different pollution degrees were analyzed,and the characteristic moments and frequencies of five groups of oil signals with pollution degrees were extracted.It was found that with the increase of the pollution degree,the oil frequency of the IMF2-4component shifted toward the middle of the interval with the increase of the pollution degree,and the oil frequency of the IMF5-7 component shifted toward the 5Hz,3Hz and 1.6Hz directions,respectively.The transient velocity vector streamline diagram of the contaminated oil obtained through the PIV experiment verified the reliability of 0.1s as a characteristic moment of the first group of pollution degree.Finally,a T-S fuzzy model was established between the amplitude of the IMF 2-5component frequencies and the degree of oil pollution,and the internal relationship between the amplitude of each component frequency and the degree of oil pollution was analyzed.The results showed that the prediction error range of the established T-S fuzzy model was basically 0-0.21%,with good prediction accuracy,and could effectively reveal the internal relationship between the oil dynamic characteristic signal extracted by HHT and the degree of oil pollution.This research had laid a solid technical support for the online monitoring of oil.