【作者】 彭珍瑞；

【导师】 黄志尧； 王保良；

【作者基本信息】 浙江大学 ， 控制科学与工程， 2007， 博士

【摘要】 两相流广泛存在于动力、石油、冶金、核能和化工等领域。两相流参数的测量对生产过程的计量、控制以及环保等具有非常重要的意义，但是两相流流动特性复杂，两相流参数检测的难度相当大。空隙率是众多两相流参数中的一个重要检测参数，现有的多种空隙率测量方法还未能满足实际工业应用要求，空隙率测量技术仍有待进一步研究。本文重点针对气液两相流空隙率的测量进行了研究，主要工作和创新点如下：1．提出了一种最小二乘支持向量机(LS-SVM)的改进方法。针对现有LS-SVM的不足，运用训练数据点筛选策略，根据LS-SVM模型训练误差的大小，剔除训练数据中误差过大和过小的数据点，从而获得模型的“稀疏性”，提高了模型的泛化能力和鲁棒性。仿真验证和实际空隙率测量的实验验证，表明LS-SVM的改进是成功的。2．将实数编码的遗传算法(RC-GA)引入到LS-SVM参数优化中，解决了LS-SVM使用时存在的参数选取困难的问题。将LS-SVM的参数选取问题看作优化问题，建立优化问题的目标函数，凭借RC-GA强大的全局搜索能力，搜索最优LS-SVM参数。仿真结果和实际空隙率测量实验表明RC-GA方法是有效的。3．提出了基于12电极电容层析成像技术(ECT)和LS-SVM的油气两相流空隙率在线测量的新方法。运用该方法测量空隙率时，首先辨识流型，然后选择与流型辨识结果相对应的空隙率模型计算获得空隙率。该方法省去了常用ECT方法测量空隙率时复杂而耗时的图像重建过程，提高了空隙率测量的实时性。实验结果证明了本方法的有效性。4．提出了基于16电极电阻层析成像技术(ERT)和LS-SVM的气水两相流空隙率测量的新方法。该方法由ERT传感器获得104个测量值，然后把ERT传感器得到的104个(去除16个与激励电极对相邻的测量值后为88个)测量值作为已经建立好的空隙率测量模型的输入，计算获得空隙率。实验结果表明本方法是有效的。更多还原

【Abstract】 Two-phase flow exists widely in many industrial fields such as power,petroleum, metallurgy, nuclear energy, chemical engineering and so on.Measurement of two-phase flow parameters is very important for the metrology,control and environment protection in modem production processes. However, the inherent complexity of two-phase flow system leads to many difficulties in measuring the parameters of two-phase flow. Voidage is one of the most important parameters. Although many voidage measurement methods have been proposed, it is still difficult to measure the voidage due to the complexity of the characteristics of two-phase flow. It is necessary to explore new methods for voidage measurement.The author mainly focuses on the research of voidage measurement of gas-liquid two-phase flow. The main works of the dissertation are listed as follows:1. An improved Least Squares Support Vector Machine (LS-SVM) was proposed to overcome the drawback existing in the present LS-SVM. According to the training errors of LS-SVM model, data points with too large or too small errors were discarded, and thus the generalization ability and robustness of LS-SVM model was improved. The improved LS-SVM was verified by simulation data.Experimental results of voidage measurements also demonstrate that the improvement of LS-SVM is effective.2. Real-Coded Genetic Algorithm (RC-GA) was introduced to solve the difficult problem of parameters selection in LS-SVM. The issue of parameters selection in LS-SVM was regarded as an optimization problem. A RC-GA with global searching capability was employed to search the optimal parameters in LS-SVM. Experimental results show that the RC-GA based parameters optimization method is effective.3. Based on Electrical Capacitance Tomography (ECT) and LS-SVM, a new method was proposed for on-line voidage measurement of oil-gas two-phase flow. LS-SVM was used to establish the voidage measurement models under different flow patterns. In the measurement process, the flow pattern of oil-gas two-phase flow was first identified, and then, the voidage was computed using the voidage model corresponding to the identified flow pattern. This new method implemented voidage measurement without complicated and time-consuming image reconstruction. And thus the real-time performance of voidage measurement was improved. Experimental results show that the new method is effective.4. Based on Electrical Resistance Tomography (ERT) technique and LS-SVM, a new voidage measurement method of gas-liquid two-phase flow was proposed. In this method, the ERT sensor was employed to obtain the 104 independent resistance values. These 104 measurements values (88 measurements after reducing dimension by eliminating 16 adjacent values) were the input of the voidage measurement model which was established by LS-SVM. The output of the model was the voidage. Experimental results prove the effectiveness of the new method.更多还原