| Earthquake is a kind of special crustal movement that occurs frequently in nature and is one of the most serious geological disasters endangering people’s lives and possessions. For a long time, people hope to know the occurrence time, location, the regularity of magnitude and the probability of occurrence by the way of realizing various abnormal phenomena before the earthquake. Various kinds of macroscopic and microscopic anomalies can reflect the gestation, occurrence, development of earthquake. In recent years, the prediction of earthquake disaster has become one of the hot spots. So short-term and imminent prediction is full of challenge in earth sciences. As shown from the observation and research, the signal anomalies of electromagnetic fields are a good reference to predict earthquake which can reflect the temporay and spatical effects of earth activities. The signals carry a lot of valuable data to analyze seismogenic information.The analysis about seismogenic information based on the natural electromagnetic fields around the surface has been an important method to realize the the gestation, occurrence, development of earthquake. This paper focuses on discussing the characteristics of the seismogenic information based on the natural electromagnetic fields, analyzing the distributions of the electromagnetic signals in the time-frequency energy spectrum around the earthquake and studying the availability of using this kind of signal to conduct prediction. The signal can be defined as the superposition of the signals containing various noises which can be obtained from the surface and is aperiodic and nonstationary. The output of the data is the signal quantified by sensors and these data are stored as the format of time-amplitude and time-number of pulses (t-AH, t-NH).This paper can be divided into five parts:1. The first part analyzes the connection between Earth’s natural pulse electromagnetic fields (ENPEMF) and the occurrence, studies the characteristics of the changes about the natural electromagnetic fields and came up that in the VLF, the electromagnetic anomalies could be obtained before the earthquake which indicated that ENPEMF is reliable to predict the earthquake.2. For the sake of it that the signal is aperiodic and nonstationary, this part analyzed several typical methods to smooth the mean of massive data and illustrated that the average square method could reflect the tendency of NH and AH. So the average square method was used to smooth and compress the data to make the data become fluent and the size smaller. And then the joint time-frequency analysis method was used to process data.3. This part compared several typical time-frequency analysis methods, explained the limitation of Fourier transform and wavelet analysis, and studied HHT and WVD in details. From the comparison between marginal spectrum and Fourier spectrum, it is clear to see that marginal spectrum has overwhelming superiority on reflecting the real frequency of signals. As two effective time-frequency analysis methods, they have gradually been applied to various kinds of non-stationary signal analysis and have played good roles. At the same time, the paper also explained the weaknesses of them. In the process of decomposition of EMD, there exists end effect and mode mixing. When selecting the extreme point to form the envelope by spline interpolation, there might be some errors, and the more the number of screening is, the larger the errors will be. When something is serious, the decomposition might be meaningless. In order to solve the problem, the paper came up with some typical methods and took the mirror extension and extension extrema directly for example to analyze in details. Mirror extension is to set a mirror at the symmetric extreme point and then the signal will be twice the length of the former one. If taking one cycle to analyze, the end effect will be ignored. In fact, the result of mirror extension is better than extension extrema directly.As a result of the signal itself or the limitation of decomposition method, EMD can not decompose signals in the same scale correctly. This article made reference to the ideas of the EEMD, analyzed and simulated the process of decomposition. Afterwards, it drow the conclusion that EEMD could get rid of mode mixing to a certain extent, and changing SNR can restrain mode mixing in different degrees. EEMD is an improved version based on EMD, which is widely used in various fields such as pectorophony.4. Time-frequency analysis based on WVD has been widely used in all kinds of fault diagnosis fields. This method has good time-varying characteristic especially to define signal with accurate instantaneous frequency. It can well describe edge features, instantaneous frequency and localization, but the cross terms will become the bottleneck of its applications. Coss term makes it difficult to show each frequency component clearly. The interference generated in analysis greatly limits the application and development of WVD in reality.This article came up with some methods to avoid coss terms such as:PWVD, SPWVD, Cohen, EMD-WVD, EEMD-WVD and DEEMD-WVD. Taking three methods for example:(1) EMD-WVD:Firstly, decomposing the signal by EMD, analyzing the IMF components through WVD and piling them up to get the distribution of EMD-WVD. Then, the different components were separated for analysis respectively, the interference of different components and cross terms are reduced. Compared with decomposing directly by WVD, each IMF has better time-frequency clustering.(2) EEMD-WVD:Similar to EMD-WVD, the only difference is to decompose the signal by EEMD. Because EEMD is sometimes better than EMD, this kind of method can be regard as the improvement from EMD.(3) DEEMD-WVD:This method decomposes the IMF components again after obtaining them from EEMD, and then screens the single frequency components of high quality to analyze by WVD. SNR also plays an important part in the matter of getting rid of the cross term. Though EEMD-WVD sometimes is better than DEEMD-WVD in the case that SNR is small, on the whole, DEEME-WVD has preferable results for more general application.5. This part introduces the performance of ENPEMF in time-frequency analysis during the earthquake occurred in Lushan, and then analyzes the data about AH and NH from CN2 and CN3. Moreover, the article draw two-dimention and three-dimension time-frequency energy figures during the earthquake by MATLAB to find out the characteristics of seismogenic information. In order to get the valid time-frequency analysis diagram, two important screening conditions were came up with:judgment by correlation coefficient and convolution. Judgment by correlation coefficient is divided into two conditions:filtrate only in the first time and filtrate both in the first and second time. In conclusion, judging by correlation coefficient is better than judging by convolution, and filtrating only in the first time is better.In the research above, the last part summarizes the results, conclusions and major innovation of thesis. This paper also find some place need further improvement. |
PDF Full Text Request