Study On Identification Of Core Aperture Based On Hardware Accelerated Electrical Spectrum Distribution Type

The complex resistivity spectrum curve and capillary pressure curve of rock are important related data in logging field.The petrophysical experiment shows that the complex resistivity spectrum has relatively strong correlation with the pore structure of rock,and the capillary pressure curve can indicate the pore throat distribution of rock.Therefore,this paper focuses on the joint identification of core aperture distribution type by complex resistivity spectrum curve and capillary pressure curve.A method of constructing characteristic points based on the shape of capillary pressure curves was proposed.The core capillary pressure curves measured in the laboratory were divided into four categories,and the correctness of the classification results was verified.The stochastic forest algorithm was used to classify and predict the capillary pressure curves of rock complex resistivity spectrum,which was convenient for the subsequent setting of capillary pressure measurement points.The random forest algorithm runs efficiently on CPU.With the increase of the number of samples of complex resistivity spectrum curve and the number of decision trees,the running time of the algorithm increases exponentially,which cannot meet the requirements of real-time logging.However,hardware FPGA has a significant acceleration effect on the implementation of the algorithm,and the calculation process of decision tree can be designed in pipeline.The random forest algorithm is composed of several independent decision trees,which can be designed by taking advantage of the characteristics of FPGA parallel computation to achieve the algorithm acceleration effect.The experimental results show that the accuracy of the random forest algorithm for predicting the type of capillary pressure curve with complex resistivity spectrum is consistent with that of MATLAB and FPGA,and the accuracy can reach 82.98%,which indicates that the principle of predicting the type of capillary pressure curve with complex resistivity spectrum is feasible,but the time of the algorithm running on MATLAB is more than 50 times that of FPGA.With the increase of the number of complex resistivity spectrum samples and decision tree,the acceleration effect of FPGA on the algorithm becomes more and more significant.