The Research Of LMT Time Domain Data Analysis And Processing

Long Period Magnetotellurics (LMT) is based on the development ofMagnetotelluric sounding (MT). Since there is no restriction of low-frequency, thedetection period of Long Period Magnetotelluric can be up to several hundredsthousand seconds and the effective detection depth can reach several hundredkilometers. With the enhancementof the sensitivity of instrument and the increaseinthe accuracy and speedof data processing in recent years, the validity and reliabilityof LMT have been greatly improved, so LMT has become an important means in non-seismic detecting areas. Signal processing is the fastest growing discipline ininformation scienceand is widely used in all scientific and technological fields.thetime-domain signal processing of Long Period Magnetotelluricsis to analyze thefield observed time-domain signal in order to obtain reliable spectralinformation.Therefore, we must ensure the integrity of the original time-domain data,as well as the purity of the data. For methods to ensure data integrity, At present, onlyto ensure the integrity of the data collected in the collection process as much aspossible. Toensure the purity of the data, there are a variety of de-noising methods:remote reference method, digital filters, wavelet de-noising,neural networks,Hilbert-Huang transform and generalized S transform, these treatments can be carriedout for different noise repression. But the noise filtering has always been a hot issue,so explore more suitable method for de-noising is an important issue of the LongPeriod Magnetotelluricstime-domain data processing. For these two aspects of theproblem the signal characteristic analysis, forecasting of missing data, noisesuppression of time domain signal are analyzed and studied, and a visual time-domainsignal processing package is achieved in this paper:(1) First, the theory of signal classification and the basic methods of signalprocessing are introduced, and the analyzing of the characteristics of Long-PeriodMagnetotelluric signals in time domain and frequency domain indicates that the timefor collection of LongPeriod Magnetotelluric (LMT) is long, the amount of dataishuge, its spectrum is wide and the energy amplitude is small, so the Long-PeriodMagnetotelluric signal is vulnerable to interfered.(2)A problems in the long-term field data collection process is proposed in thispaper: in the Long Period Magnetotelluric collection process, As the effect oftemperature and humidity to the instrument or GPS is not normal, the collected data is sometimes skipping. To solve this problem, no incentive AR (p) model isbeintroduced. According to the known sequences to determine the order and modelparameters of AR (p) modelto establish the correct prediction model and to predict themissing data, thento compare the spectral data through the prediction and the actualsample data and to show that AR (p) prediction model can be solved the problem ofdiscontinuity raw data.(3)For noise presented in the LongPeriod Magnetotelluric time-domain signal,the common type of noise is analyzed. After the convolution filtering forLongPeriodMagnetotelluric data, the effect of de-noising and the problems of convolution filter issummed up.And based on this, combined with wavelet analysis theory and improvedthreshold function, improved threshold function of LMT signal wavelet de-noising isintroduced, and the good results is demonstrated.(4)To the invisible problem for LongPeriod Magnetotelluric data processing,thevisualization system interface of time data processing is finished, making every stepof the data process can be viewed, the efficiency of data processing is Improved.