| Middle and lower reaches of the Yangtze River region is an important metalog- enic belt in our country. Exploring its fine structure and deep ore-controlling structure are of great significance to the search for mineral resources in the deep area. MT studies in the noise measurement data to identify, process technology and to improve the quality of MT data are of practical significance to detect the fine structure of the crust and seek the deep ore-controlling structure.This paper is based on the"MT measurement in the ore concentration area of Jiurui"subject, assumed by Jilin University. We studied the characteristics of the strong MT noise interference and its main sources of interference, and through the combination of theoretical study and analysis of examples'effect, we took methods to inhibit and eliminate noise interference, then achieved the ultimate aim of improving data quality.In fact, MT sounding method is the kind of Cagniard resistivity sounding method whose field sources are natural electromagnetic signals. These signals are the result of the combined effects of different attributes, different intensity and different distances from field sources on a wide range. This kind of signals is weak and has frequency bandwidth. So it is inevitable that the original data will be subject to a variety of noise pollution. When L≥3δ(L: distance between field sources and observation points;δ: the skin depth), the electromagnetic field meets the requirements of the plane wave field, so we call it far-field zone; when L≤3δ, the electromagnetic field is non-plane-wave field, so we call it near-field zone. In actual MT data-processing work, we generally use Cagniard resistivity formula to calculate the impedance. The formula applies only to plane electromagnetic wave. When it is used to calculate near-field wave formation, we can obtain apparent resistivity curve 45°asymptotic upward trend line, resulting in near-source effect. The ore concentration area of Jiurui has many sources of interference and complex human environment. MT original data collected in the wild generally are subject to different noise. According to the characteristics of noise'waveform, we divided them into 6 categories: impulse noise, periodic noise, triangle-wave noise, square-wave noise, step noise and like the triangular charge-discharge mode noise. According to the research on time series, we found that the characteristics of the six kinds of noise are obvious, and they impact a wide frequency range. The width of the noise waveform data can affect the low-frequency data. The height can affect the high-frequency data, and in general the energy of noise is strong. In more serious cases, it can completely destroy useful signals.In the ore concentration of Jiurui, mining industry is developed. There are High-power DC electric traction locomotives in the state-owned large and medium-sized underground mines. They can produce a wide range of free electric current in the lower reaches, which made electric channel data had many large-energy square noise and there were different degrees of near-source effects in resistivity curves. When interference is serious, they can make data completely loose value, so they become the main source of strong interference. Mountains in the region are undulating. The base stations of Mobile, Unicom and PHS are close. Although signal frequency up to more than MHz, the signal strength is large, and exist the phenomenon of failure frequency. Communication base stations use high-voltage underground transmission and the power supply distribution substation facilities, at the same time, they make a great impact on data. In addition, the vibration interference caused by civil mining and low-frequency interference often appear. Free electric current, roads, railways, carrier phone, high-tension lines, substations and other infrastructure in the cities of the region also have a greater interference with the measured data.Having researched on the basic principles of remote reference method and wavelet de-nosing method, combined with analysis on the effect of examples in the study area, we found that the two methods in the study area have limitations on the existence of noise. Remote reference method can improve the quality of high-frequency data well and can better recover the MT weak signals. If time for observation is long enough, the effect of low-frequency part improvement is also obvious, while the humanities interference can be less inhibited. But it is powerless to near-source effect in mines and communication base stations. Reference near the source still exists in the improved data. So they still can not be used for inverse calculation; wavelet de-nosing method can improve a single type of noise and strong energy data, but when noise energy is weak and number is large, the capacity of noise identification and de-noising is often very poor.For the characteristics of noise in the area we pose man-machine de-nosing method. Because the original data and the MT curve have a direct relationship, the purpose of data synchronization de-noising can be achieved by modifying the curve. This method introduced in de-nosing theory of manual translation and linear interpolation fitting. Analyzing the effect of de-nosing measured data shows that this method not only can meet the shortfall of remote reference method and wavelet analysis, but also can deal with noise of different types and intensity.We prepared man-machine de-noising software after combination of VB language and API function. The software interface is divided into three modules, including data acquisition and storage modules, value and the related graphics data display module, and de-nosing operation module. When we designed the interface, we added at a glance menu bar, buttons and logo etc... The interface is succinct and clear overall, and improves operation efficiency.In the preparation of software we solved a number of major key technologies, including the V5-2000 original document storage format, coordinate conversion technology, curve display and zoom function, select the exact sampling point and noise identification and correction curve technology.The functions of man-machine de-noising software as follows: (1). Quick read and store the V5-2000 original data files.(2).Drawing curves and accurately display graphical.(3). Choose original curves of any observation and any one time period to analysis freely. (4). Automatic calculation and display the relevant information. (5) Could be selected with function of automatic identification of the mouse and fine-tune of the keyboard which can accurate realization of the noise waveform. (6). the realization of the manual shifting and linear interpolation function fitting the de-noising. (7). Add function of real-time saving which could avoid the loss of data.The software has many characteristics, such as user-friendly interface design, curves show the visual data, auxiliary to distinguish noise, better maneuverability, excellent de-noising effect. Users can choose any one observation and any one time period curve to analysis, and they have the freedom to control the display cycle of the curves, enlarge and reduce the curves to understand and evaluate the data quality. When de-noising is operated, the control buttons concise, noise waveform recognition technology, de-noising scale and selection of speed greatly enhanced operational efficiency. In addition, we also added display data to the software to support the users to check SNR before and after de-noising the data.Objectively speaking, any kind of de-noising method is not omnipotent. When the noise interference is complex and any kind of automatic de-noising method can not solve the problems, the man-machine de-noising method becomes the last way. It not only can greatly improve data quality, but also can make staff to understand data interpretation. At the same time, staffs can understand the data quality objectively. Although it needs spend a lot of energy and time, it is also worthwhile.
|
PDF Full Text Request