导　　师： 黄志尧

学科专业： 081102

授予学位： 博士

作　　者： ;

机构地区： 浙江大学

摘　　要： 两相流参数检测是目前国内外迫切需要解决而又长期未能得以很好解决的测量难题。数据挖掘技术是目前信息科学领域研究热点之一。本学位论文基于主成分分析/(PCA/)//主成分回归/(PCR/)、偏最小二乘/(PLS/)//偏最小二乘回归/(PLSR/)和最小二乘支持向量机/(LS-SVM/)等数据挖掘技术,研究气液两相流流型辨识、空隙率测量和液液两相流参数测量等问题。 本学位论文的主要创新点和贡献有: 1.基于12电极ECT阵列式电容传感器,提出并比较研究了三种基于数据挖掘的油气两相流流型辨识的新方法:1/)LS-SVM流型辨识方法,2/)PCA与LS-SVM相结合的流型辨识方法,3/)PLS与LS-SVM相结合的流型辨识方法。研究结果表明,该三种流型辨识方法均是有效的,其中以PLS与LS-SVM相结合的流型辨识新方法准确度最高,与现有的其它流型辨识方法相比较,具有流型辨识准确度较高、实时性能佳等优点。 2.基于12电极ECT阵列式电容传感器,提出了基于数据挖掘的油气两相流空隙率测量新方法。针对水平管油气两相流每一种典型流型分别建立有相应的空隙率测量模型,实际测量时首先采用PLS与LS-SVM相结合的方法辨识流型,然后根据辨识结果选择相应的空隙率测量模型计算获得空隙率测量值。空隙率测量模型采用数据挖掘技术建立,对比研究了三类建模方法:1/)线性回归方法/(PCR和PLSR/),2/)非线性回归方法/(LS-SVM/),3/)特征量提取/(PCA或PLS/)与LS-SVM相结合的方法。研究结果表明,所提出的空隙率测量新方法是有效的,该方法直接利用阵列式电容传感器所获得的电容测量值实现空隙率测量,回避了所有的图像重建过程并考虑了流型对空隙率的影响。研究结果同时表明,所建立的各种空隙率测量模型中,利用PLS与LS-SVM相结合的方法所建立的模型具有最好的测量表现。与现有的基于电容层析成像技术或其他数据挖掘方法的空隙率测量方法相比较,若采用PLS与LS-SVM相结合的方法建立的空隙率测量模型,则所提出的新方法的最大测量误差小于3.6/%,具有精度高和实时性能佳等优点。 3.将16电极ERT阵列式电导传感器、组合式流量测量原理和数据挖掘技术相结合,提出了一种液液两相流参数测量的新方法。基于阵列式电导传感器所获得的电导测量值,利用PLS和LS-SVM等数据挖掘方法设计了主导相判别器,以克服相含率对液液两相流参数测量的影响。组合式仪表由文丘里管和椭圆齿轮流量计组成。实际测量过程中,文丘里管和椭圆齿轮流量计获得油水两相流差压和总体积流量,主导相判别器识别两相流中起主导作用的相,然后根据判别结果选择文丘里管仪表系数,最终获得油水两相流的总体积流量、混合密度和总质量流量。研究结果表明,所提出的油水两相流测量新方法是有效的,总体积流量、混合密度和总质量流量的最大测量误差分别小于1.0/%、4.6/%和4.5/%,并具有成本低和易于工程实现等优点。 Parameter measurement of two-phase flow has been an urgent and difficult task in the world for a long time.However,it has not been solved satisfactorily up to now.Data mining technique is currently a hot spot in the research field of information science.Based on data mining techniques such as Principle Component Analysis/(PCA/)// Principle Component Regression/(PCR/),Partial Least Squares/(PLS/)//Partial Least Squares Regression/(PLSR/) and Least Squares Support Vector Machine/(LS-SVM/),this dissertation studies the flow pattern identification and the voidage measurement methods of gas-liquid two-phase flow and the parameter measurement method of liquid-liquid two-phase flow. The main innovative points and contributions of this dissertation are listed as follows: 1.Based on the ECT capacitance sensor array and data mining techniques,three new methods are proposed and compared for the flow pattern identification of oil-gas two-phase flow:1/) flow pattern identification method based on LS-SVM,2/) flow pattern identification method based on PCA and LS-SVM,3/) flow pattern identification method based on PLS and LS-SVM.Research results show that these three proposed methods are all effective and the method based on PLS and LS-SVM has the highest flow pattern identification accuracy. Compared with other flow pattern identification methods,the method based on PLS and LS-SVM has the advantages of relatively high accuracy and good real-time performance. 2.Based on the ECT capacitance sensor array and data mining techniques,a new method is proposed for the voidage measurement of oil-gas two-phase flow.Voidage measurement models are established in advance for each typical flow pattern of oil-gas two-phase flow in horizontal pipe.In the practical measurement process,the flow pattern of oil-gas two-phase flow is firstly identified according to the method based on PLS and LS-SVM.Then,the suitable voidage measurement model is selected by using the flow pattern identification result to calculate the voidage value.The voidage measurement models are established based on data mining techniques.Three kinds of modeling methods are studied and compared:1/) Modeling methods based on linear regression/(PCR and PLSR/),2/) Modeling methods based on non-linear regression/(LS-SVM/),3/) Modeling methods based on feature extraction/(PCA or PLS/) and LS-SVM.Research results show that the proposed method is effective.Without any image reconstruction process,it directly uses the measured capacitance values to implement the voidage measurement and can overcome the influence of flow pattern on the voidage measurement.Research results also indicate that among all the established voidage measurement models,the PLS and LS-SVM based models shows the best measurement performance.Compared with the other voidage measurement methods based on ECT or data mining,if the PLS and LS-SVM based model is adopted,the proposed voidage measurement method has the advantages of high accuracy and good real-time performance and its maximum voidage measurement error is less than 3.6/%. 3.With the 16-electrode ERT conductivity sensor array,the measurement principle of hybrid flowmeter and data mining techniques,a new method is proposed for the parameter measurement of liquid-liquid two-phase flow.To overcome the influence of oil fraction on the measurement results,a dominant phase identifier is developed based on the conductivity values obtained by the conductivity sensor array and data mining techniques/(PLS and LS-SVM/).The hybrid flowmeter consists of a Venturi meter and an oval gear flowmeter.In the practical measurement process,the differential pressure and the total volume flowrate of two-phase flow are firstly obtained by the Venturi meter and the oval gear flowmeter. Meanwhile,the phase that dominates the global characteristics of the oil-water two-phase flow is identified by the dominant phase identifier.Then,the meter coefficient of the Venturi meter is selected according to the domiant phase identification result.Finally,the total volume flowrate,the density and the the total mass flowrate of the oil-water two-phase flow are calculated.Research results show that the proposed oil-water two-phase flow measurement method is effective.The maximum measurement error of the total volume flowrate,the density and the the total mass flowrate of the oil-water two-phase flow are 1.0/%,4.6/%and 4.5/%,respectively.Additionally,the proposed method is low-cost and easy to be implemented.

关 键 词： 两相流 测量 数据挖掘 流型 空隙率 主成分分析 主成分回归 偏最小二乘 偏最小二乘回归 最小二乘支持向量机

分 类 号： [O359.1]

领　　域： [理学] [理学]