Application Of Data Mining Methods In Engineering Electrical Geophysical Data Interpretation

Yunnan is one of the serious water shortage provinces. Yunnan water diversion project aims to ease water shortage state, which has an important significance for the national economy. Digging tunnel for diversion in the complex geological mountain environment is an important part of the water diversion project. The correct evaluation of the quality and characteristics of rock, accurate and timely determination the location of the abnormal geological bodies, which not only can reduce the cost of construction, but also has great significance for guiding tunnel construction safely. Because conventional drilling and geochemical methods are difficult to be applied in the large depth tunnel project located in the alpine region, it can only get some information by means of geophysical exploration. For apparent resistivity data obtained by means of geophysical exploration, the results of conventional geophysical interpretation is often rather rough. Therefore, a variety of data mining methods are used in this paper, which is to obtained richer and more refined information from the apparent resistivity data, will help the geological exploration on the diversion project.Metalliferous mineral, carbonaceous and clay minerals with good conductivity have significant influence to the resistivity data. Under normal circumstances, the main factors that affect the resistivity of rock are porosity, aquosity and water salinity of the rock. For the abnormal geological body just like overburden, fault fracture zone and erosion areas, etc., compared with the surrounding normal rock, they will be reflected in the physical characteristics which apparent resistivity will occur low resistivity changes, especially in its interactive influence with water (surface water or groundwater). Based on the apparent resistivity responsing the physical properties of rock, this study aims at finding a suitable method for the detection of apparent resistivity anomaly cutoff points, while using image processing methods to get the effective visual expression of the apparent resistivity data, so that it can show more intuitive level variation, and then combine the two to divide the cover layer, analyze and find the abnormal geological body, and grade the tunnel wall rock according the physical properties.A variety of data mining methods are used in the process of the study, for example, combined with drilling data, Association Rule Mining (ARM) method is used to find out the relationship between apparent resistivity and lithology so as to perform lithology determination; The improved Minimax Variance Stratification (MVS) method is used to detect abnormal points in the apparent resistivity data; Activity Stratification Method (ASM) is used to detect abnormal points in the apparent resistivity data; Wavelet Transform Singularity Detection (WTSD) theory is used to detect abnormal points in the apparent resistivity data, and analyze the abnormal distribution on geological cross-sectional, etc. The reservoir is between the Youche village and Yangxian tea plantation in Yixing city Fuhu town of Jiangsu Province, whose apparent resistivity data is obtained by high density electric method; there is clear explanation on the results of exploration in some areas in the report, which is convenient to make the results contrast and effectiveness evaluation of research methods. Therefore, we choose this reservoir data as the experimental data for the methods’comparison. After the comparison and selection, a set of data mining methods based on wavelet transform combined with image processing to divide the cover layer, analyze and find the abnormal geological body, and get the classification of the tunnel wall rock according the physical properties.Firstly, Histogram Equalization (HE) method is used to deal with the apparent resistivity data and more intuitively exhibit the characteristic changes in the level of the data; On this basis, the drilling data is used to mark the position of overburden on the apparent resistivity image; its performance characteristics is analyzed, and the rules of dividing the cover layer is summarized to divide covering layer.Then, we adopt one-dimensional dyadic wavelet transform to detect abnormal points of the apparent resistivity data, and combining with the resistivity image to divide the abnormal area on the section, make reasonable explanations on the causes of the abnormal area reference the geological plane graph.Finally, wavelet transform is used to detect abnormality breakpoints of the apparent resistivity data in horizontal orientation; and then we take the upper and lower 25-meter cross-section of the runnel centerline axis as the study area, to zone the study area according to the demarcation points; finally, The ArcGIS partition statistical tools is used to count the apparent resistivity mean of each partition, grade and visualize them.Shek Kwu Wangchengpo Xianglushan tunnel is the control project of Yunnan water diversion. Because of the long distance and large depth tunnel, it is more likely to encounter karst leakage and other issues on the surrounding rock stability suddenly during the construction process, which brings high risk to the construction security. The data mining methods which was chosen above are applied in this project, in contrast to the interpretation results of geophysicist, the results of this study is consistent to them in the overall trend, but partial is more refined. The use of horizontal wavelet abnormality breakpoints on partitioning tunnel regional ensures that the internal identity in the same segment and differences between the different segments, which is scientific as well as reasonable. It can ensure the rationality of the results of the statistics, evaluation and grading on such a partitioning. The data mining methods combined Wavelet transform with image processing are implemented based on data processing, which reduces reliance on personnel experience of interpreters. The methods can assist bore-hole arrangements; it also can carry out the properties classification evaluation along the construction section effectively, and provide a reference for tunnel construction safety.