| The riser is the main pipeline structure in deep-sea oil and gas exploitation.In order to save transportation cost,the pipeline is carried out by multiphase transportation.The ability to accurately measure and detect two-phase flow parameters in real time is therefore of great importance in engineering.Acoustic emission technology can realize real-time monitoring,with the advantages of high sensitivity and good environmental adaptability,which makes it has great potential in the field of multiphase flow.For long-distance multiphase transportation pipelines and subsea oil and gas exploration,it is important to study the mechanism of gasliquid two-phase flow in risers using acoustic emission technology and to establish a set of acoustic emission-based real-time detection model for two-phase flow parameters.In this paper,the effects of different pipe materials on the transmission acoustic emission signals of gas or liquid single-phase flow and two-phase flow in the vertical riser pipe is studied firstly,and the acoustic emission sensors with two sensitive frequency ranges are used to synchronously collect the flow state of the pipeline at the same section.The acoustic emission signals of severe sluging,slug flow,bubbly flow,churn flow and annular flow in plexiglas and stainless steel tubes were analyzed by parameter statistics analysis,time-frequency analysis and wavelet packet analysis.The results show that the gas-liquid two-phase flow is mainly caused by the movement of bubbles in the tube and the pressure on the wall of the tube.The signals mainly vary in the range of 0-130 k Hz,and the variation range of the signals in the other frequency bands is small.Compared with the stainless steel tube,the plexiglass tube has a strong attenuation effect on the acoustic emission signal,mainly concentrated in the range of 0-120 k Hz.The average voltage level(ASL),root mean square(RMS),absolute energy value(ABS),count and the wavelet packet energy spectrum of eight nodes with gas-liquid velocity show that the acoustic emission signal can be used for the detection of gas-liquid flow parameters..The acoustic emission signal in the pipe is weak,in order to increase the intensity of acoustic emission signal in the pipe,this paper refers to the structure of the vortex generator in the vortex flowmeter and selects the triangular column with the width of 5.5mm,6.5mm,7mm and 7.5mm on the stoss side as the bluff body.Firstly,it is demonstrated that the data with and without bluff body can be compared together by the probe voltage versus time graphe plot under the same operating conditions.Then the characteristic parameter variation plot with time is compared,and the results show that the characteristic parameter variation law with time varies greatly in different flow types,which proves that the characteristic parameter of acoustic emission signal can be used to detect the two-phase flow parameters,and the bluff body mainly enhances the acoustic emission signal existing in the pipe itself.Then the acoustic emission signal at the disturbance was analyzed in detail with a trigonal column with 7.5 mm on the flow surface as an example.The results show that due to the enhancement effect of the bluff body,the data with the bluff body are stronger than those without the bluff body,so the fluctuation with the change of gas-liquid velocity decreases and the slope of increase becomes larger.By comparing the spectrogram and wavelet packet energy spectrum,it is found that there is a signal with bluff body data in the range of 400-1000 k Hz whose size does not vary with the gas-liquid velocity,and the signal is considered to be related to the fluid hitting the trigonal column after analyzing the experimental data of 45° up-tilted pipe.Finally,the kurtosis factor is introduced,and it is found that the kurtosis factor of acoustic emission signal is different in different flow patterns,which can be used as a characteristic parameter to recognize the flow in the pipe.The existing genetic neural network method in the laboratory has disadvantages such as long optimization time,and this paper in order to solve the drawback introduces the support vector machine regression(SRV)method.In order to solve these disadvantages,this paper introduces SRV method.By comparing the detection effect of the two models with the same data,it is found that the SRV is superior to the genetic neural network in learning time and prediction error.Then,the optimal triangular cylinder is chosen by using the fitting effect of the wavelet energy spectrum of eight nodes in different flow patterns and the corresponding liquid holdup.Finally,the average voltage level(ASL),absolute energy value(ABS),root mean square(RMS),kurtosis factor(KV)and the relative wavelet energy of 8 nodes are selected as the input parameters of SRV.The performance of flow pattern identification and liquid holdup prediction of the detection model was evaluated by repeated sampling signals in random sampling conditions.By comparing acoustic emission sensors with different sensitive frequencies and different support vector machine parameter optimization methods,the results show that the expeimental data measured by SR150 N sensor and artificial bee colony method have the best detection performance.Therefore,acoustic emission technology can be used to monitor the gas-liquid flow in high-pressure subsea riser with thick wall,which is of great significance for deep-sea oil and gas exploitation. |
PDF Full Text Request