The Study On Soft Measurement Technology Of Two Phase-Flow Parameters

Two-phase flow phenomenon exists widely in nature and many industrial processes. At present, in many industries, such as power, chemical, nuclear, refrigeration, oil and metallurgy industries, in these production equipments involve multiphase flow condition. Two-phase flow medium distribution, is named flow pattern, which greatly affects the two-phase flow pressure loss and the heat and mass transfer characteristics, at the same time affects the flow parameters of the accurate measurement and flow system movement characteristics. So the gas-liquid two-phase flow pattern recognition research has been two phase flow parameter measurement is an important research direction. The intelligent identification is used in two-phase flow, which provides a new thought and method in this filed. The using of modern signal processing techniques in two-phase flow parameters of the soft measurement method provides the main parameters online estimation about the easily measured auxiliary process variables and off-line analysis information.This article mainly is to use a Hidden Markov model (HMM) for two phase flow parameter measurement. A Hidden Markov model (HMM) HMM (Hidden Markov model) has a strong ability of time-series modeling. Due to the HMM can handle dynamic information, therefore, it usually widely used in speech recognition, character recognition, identification and fault diagnosis, etc. In view of HMM has perfect recognition effect, and is currently in the flow pattern identification field seldom use HMM to identify flow pattern, so this paper uses the collected data and puts forward a method for identifying gas-liquid two-phase flow pattern based, on Hidden Markov model, with the method of the linear prediction coding(LPC) and support vector machine (SVM) method. First, processing the experimental parameters, and obtains the characteristic vector; second, inputting to the trained HMM model, so as to realize the typical flow pattern identification. In training to identify process, first of all to the HMM is initialized, and then using the HMM algorithm based on the three basic computing, which used the Forward-Backward algorithm for the logarithmic likelihood probability measurement signals; Using the Viterbi algorithm for a given observation sequence in each kind of model is optimized; Using Baum-Welch revaluation of parameters is realized, so as to realize the flow pattern identification of two phase flow process. Through the calculation and simulation, the results were very good. Using Hidden Markov model parameter analysis is a kind of new two-phase flow pattern identification of the soft measurement method.