Geochemical Data Analytical Methods Based On Factor Analysis And Fractal Theory

As a branch of geoscience,geochemistry plays an irreplaceable role in geological exploration and resource prediction.When the geochemical data are processed and analyzed by different analytical methods,researchers usually get relatively different geological analysis results,which will have a certain influence on the actual geological exploration and resource prediction.Therefore,how to select and put forward the best geochemical data analysis method according to the actual geological tectonic events,regional geological situation and hydrogeological conditions in the study area is particularly important.According to the complexity of geological bodies,the heterogeneity of element assemblage and enrichment and the anisotropy of element migration,through the optimum selection analysis of geochemical data,we can find out the enrichment regularity of some chemical elements in time and space,establishing regional quantitative geological model and exploring the ore-forming regularity and mineralization,which provides a certain basis for regional geological survey and detailed investigation,resource exploration and prediction.Therefore,this paper uses two methods to process and analyze the stream sediment geochemical data of Guyang County in the northern periphery of Ordos Basin,namely factor analysis based on compositional data processing and fractal theory,and obtains good results.The details are as follows:1.Geochemical data can be regarded as a typical compositional data,being combined with factor analysis,the relationship among different data can be explored effectively.On the one hand,component data has equivalence,closed property and simplex space property.Studying and discussing the difference between the ratio of each component to the total vector between data sets can find out the hidden internal relations between data.On the other hand,the traditional factor analysis is essentially a linear dimension-reduction method,which is not suitable for nonlinear geochemical data processing.However,through different logarithmic ratio transformations,the data can not only satisfy the normal distribution law,but also eliminate the closure effect and the negative correlation of covariance,which is more suitable for subsequent statistical analysis.In this experiment,for trace elements and oxides,additive logarithmic ratio transformation and centralized logarithmic ratio transformation were respectively used for factor analysis,and the results of the factor analysis were compared with the original data.The results show that the factor combination under the additive logarithmic ratio transformation is better than the factor combination under the centralized logarithmic ratio transformation and the original data,the correlation is more relevant,and the factor score graph is in good accordance with the regional geological background:2.Being based on the traditional compositional data analysis,the Contain-Area(C-A)fractal method is adopted to identify the weak anomalies of iron in the study area.The results show that the lower limit of anomaly determined by the fractal method is 2.73%,while the lower the traditional method is 7.92%.The fractal method not only preserves two strong anomaly fields of iron elements,but also identifies five weak anomaly fields of iron elements well,and the anomaly distribution is consistent with the location of iron ore.Compared with the traditional method,this method has better effect in identifying anomalies.At the same time,some iron anomalies are also identified in the non-iron-ore area,which can be used as a verification for regional iron-ore prospecting.The origin of these iron ore deposits is complex,mainly metamorphic iron ore,and the essence of the surrounding rock is different.3.On the basis of multifractal theory and combined with fast Fourier transform,the geochemical data are processed from the perspective of frequency domains so that different background fields could be created.The geochemical data could be treated as a one-dimensional discrete signal,and the power spectrum image obtained by fast Fourier transform has the multifractal characterization.Based on the fractal theory,the cut-off frequencies f1 and f2 of regional and local background anomalies are determined.After low-pass filtering,the inverse Fourier transform is carried out,thus the element content distribution under the two fields is obtained.Among them,the regional background anomaly can be regarded as geochemical structure under the value of "dynamic"background,(X+2S)can be used to calculate the threshold of dynamic background.The local background anomaly can be regarded as the geochemical anomaly field filtered by the background field.The anomaly lower limit of each point is different,so the anomaly can be delineated directly.The results show that the delineation of regional background anomalies includes the element enrichment caused by different geological conditions,and the extraction effect of weak anomalies is poor.However,the delineation of the local background anomaly not only preserves the information of the original geochemical field,but also identifies the weak anomaly field,which is more effective.The two methods can be combined with each other to verify the validity of anomaly delineation.