The System Of On-line Identification For Gas-Liquid Two-Phase Flow Pattern

Gas-liquid two-phase flow pattern affects the characteristic of flow and heat transfer of two-phase system, the performance characteristic of two-phase system, and the exact measurement of flow parameter. After the on-line identification of gas-liquid two-phase flow pattern is realized, pressure drop can be calculated right and bad flow pattern can be controlled, then the economy and safe performance of the whole flow system can be ensured. With the use of computer and the development of the modern measuring technology and signal processing techniques, the study and development of on-line identification for flow pattern becomes feasible.In this paper, the characteristic of parameter which can reflect the fluctuating characteristic of gas-liquid two-phase horizontal flow is discussed, the pressure drop is selected. Aimed at the characteristic of nonlinear and non-stationary of pressure drop, statistic theory and wavelet theory are considered, and average, standard deviation, skewness, kurtosis, and wavelet packet energy are calculated, one feature vector is obtained. In the end, the feature vector values is inputted to BP neutral network, and the BP neutral network is trained. Meanwhile, the mapping from characteristic space to flow pattern space is erected, the on -line identification of flow pattern can be finished by BP neutral network which has been trained.The software of on-line identification for gas-liquid two-phase flow pattern is designed with Visual Basic 6.0 based on the WindowsXP operating system. According to measurement of signal, extraction of signal characteristic and state identification, which are three progresses of realizing external and automatic identification, an on-line identification system is established. The interface of system is friendly and conveniently. In the end, 180 groups of simulating samples are gathered on air-water two-phase flow test equipment.The results show that the accurate identification rate is 91.7%, the system not only has better identificationrate,but also can automatically send out on-line flow pattern information and reach the target of on-line identification.