| In modern industry,multi-phase flow technology is more used in petroleum,chemical,lubricants,light industry and other industries,and these industries are the leading industries of the national economy.The essence of studying multiphase flow technology is to identify the flow pattern of multiphase flow.In general,multi-phase flow is mainly divided into three states: solid,liquid,and gas.This paper studies the two-phase state of oil and water.For example,in the petroleum field,multi-phase flow detection technology plays an important role in the underground crude oil extraction.In this paper,the measurement principle of capacitance tomography is briefly explained,and the three major components of the system are introduced.Based on Maxwell's equations and combining measurement principles,the positive and inverse problems of the Electrical Capacitance Tomography system are introduced.That is to conduct a theoretical study on the forward simulation and reverse evolution of the system,and obtain a mathematical model for solving the capacitance value and introduce the method of reconstructing the image during the inversion process.Aiming at the traditional flow pattern recognition technology,the traditional recognition methods are introduced respectively: nearest neighbor recognition method,feature extraction method,neural network method and compressed sensing method.The capacitance sensor is established through COMSOL simulation software,the size of the reference sensor is set,and the parameters such as the number of plates of the sensor,the thickness of the tube wall,and the coverage of the plates are simulated and studied according to a single control.The structure parameters of the sensor were further optimized,and then the parameters of the 16-pole sensor used in this paper were determined.And through the COMSOL with MATLB simulation of the sensor,the data of each flow type in the experiment is obtained.Based on the flow pattern data obtained by COMSOL simulation software,an ECT flow pattern recognition method based on integrated learning theory is proposed for the problem of low recognition rate of traditional electrical capacitance tomography(ECT)flow pattern recognition method.A large amount of streaming data obtained by the ECT system is normalized,and 60% of the sample data is randomly assigned to the training data,and 40% is used as the test data.Improve on the basis of the existing random forest classifier,obtain the identification labels of various types of flow types by identifying a single type of flow type,and reorganize the label results obtained by multiple small types through the principle of combination strategy,And finally obtain a variety of flow classification results.The simulation results of 8 typical flow patterns show that the method inherits the classification characteristics of traditional random forest.In the case of 5-40 d B signal-to-noise ratio,the recognition rate has been significantly improved,and the recognition rate can reach up to 99.93%,indicating that this method is strong against noise interference and is a method suitable for industrial detection.By learning the ECT flow pattern recognition method based on integrated learning theory,and comparing with the learning of commonly used recognition methods.First of all,the neural network method mainly analyzes the identification of each flow pattern through the BP neural network.Secondly,it uses pre-processing in the feature extraction method,and then uses the neural network method to analyze the five neural network methods from the results,and finds that the RNN neural network has the best recognition effect.By combining Kmeans clustering,ANN,Bayesian,integrated learning and other methods,under the noise condition of 20-40 d B,the overall recognition of the flow pattern was tested and analyzed,and the recognition of the integrated learning method was verified in other methods.Above. |
PDF Full Text Request