A Method for Detecting Gas-Liquid Two-Phase Flow Rate in Bent Pipes Based on Acoustic Emission and ELMAN

Abstract

A flow rate detection method for gas-liquid two-phase flow in a bend, based on acoustic emission and ELMAN, includes: deriving single-phase flow rate measurement formulas applicable to circular and elliptical cross-sections using the forced swirl theory of bends; calculating the mass flow rate of the gas-liquid two-phase flow mixture generated on the inner and outer sides of the bend by the differential pressure between the gas and liquid phases; measuring the acoustic emission and differential pressure signals of the gas-liquid two-phase flow through the bend using acoustic emission and differential pressure sensors; identifying the flow pattern under various flow conditions; and measuring the liquid holdup at the cross-section. H L The invention calculates five parameters, including voltage amplitude; constructs an acoustic emission-ELMAN neural network; and uses the neural network to obtain the cross-sectional liquid holdup and flow pattern measurement formula to perform gas-liquid two-phase flow mixing mass flow rate measurement. This invention achieves accurate cross-sectional liquid holdup measurement through acoustic emission technology and the ELMAN neural network, and combines this cross-sectional liquid holdup with the bend-pipe gas-liquid two-phase flow mixing mass flow rate measurement formula to achieve safe online flow rate measurement.

Description

Technical Field

[0001] This invention relates to a method for detecting the mass flow rate of a gas-liquid two-phase flow mixture, specifically a method for detecting the flow rate of a gas-liquid two-phase flow in a bent pipe based on acoustic emission and ELMAN, belonging to the field of gas-liquid two-phase flow detection. Background Technology

[0002] In multiphase flow systems, gas-liquid two-phase flow is the most common. It is widely present in industrial production processes across various fields, including petroleum, natural gas, and energy. Examples include evaporation and concentration processes in pharmaceutical manufacturing, electrolysis and energy conversion processes in fuel cells, and power equipment such as boilers and heat pipe heat exchangers in energy processes. Furthermore, oil and gas extraction and transportation in the petroleum and natural gas industry also fall under the category of gas-liquid two-phase flow systems. Therefore, measuring gas-liquid two-phase flow plays a crucial role in industrial production and manufacturing, as well as in social development and progress.

[0003] Currently, methods for measuring parameters of gas-liquid two-phase flow are mainly divided into separation measurement methods and non-separation measurement methods. Separation measurement methods are relatively mature; CN104832155A discloses a method for separating oil and gas, which offers high accuracy in oil and gas measurement and a simple and fast measurement process. However, this method requires bulky equipment, which occupies a significant amount of installation space on offshore platforms, thus limiting its application. Non-separation measurement methods include electrical methods, which are divided into capacitance and conductivity methods. The measurement results are relatively accurate, but require numerous functional devices and are susceptible to interference when placed in dispersed locations. X-ray methods measure parameters based on the attenuation of rays in a medium. This method uses X-rays and gamma rays, offering rapid measurement, a wide measurement range, and high accuracy. CN101261236A discloses a dual-energy gamma-ray measurement method for gas and water content in crude oil; however, its structure is very complex, expensive, difficult to widely promote, and it involves radiation. Among numerous differential pressure generating devices, the bend method is effective for measuring single-phase fluids. This method changes the fluid flow direction, creating a pressure difference between the inside and outside of the bend, thus establishing a relationship between the pressure difference and the flow. Compared to other differential pressure sensors, it has advantages such as low pressure loss, overcoming clogging, and easy maintenance. However, it has not yet achieved accurate measurement for multiphase fluids. Furthermore, due to manufacturing processes, the circular cross-section of the bend is prone to deformation during industrial production, becoming an elliptical cross-section. Since common single-phase flow measurement formulas using bends require a circular cross-section, this imposes additional requirements on manufacturing and measurement.

[0004] Acoustic emission (AE) detection technology employs a non-invasive, non-destructive testing method. The equipment features a high sampling frequency, enabling real-time acquisition of a large amount of instantaneous acoustic signals and obtaining rich flow information. It offers advantages such as a wide detection frequency range (capable of detecting high-frequency signals above 20kHz), low energy consumption, no radiation, and convenient and flexible operation. Combining AE with artificial neural networks utilizes physical and flow parameters to achieve liquid holdup measurement and flow pattern recognition. For example, CN112114047A and CN111896616A, developed in our laboratory, utilize AE and BP neural networks to achieve liquid holdup measurement and flow pattern recognition in gas-liquid two-phase flow, respectively. However, research on the application of AE in gas-liquid two-phase flow flow measurement is still lacking. Therefore, it is necessary to study the application of AE in gas-liquid two-phase flow flow measurement. In the combined application of AE and artificial neural networks, the commonly used neural network is the BP neural network. Its advantages lie in its simple structure, strong nonlinear mapping ability, and the use of error backpropagation during training. However, the global stability of the BP neural network is weak, and it is prone to getting trapped in local optima. Summary of the Invention

[0005] To address the current problem of gas-liquid two-phase flow rate detection, this invention provides a method for detecting gas-liquid two-phase flow rate in a bend based on acoustic emission and ELMAN neural network. By measuring the differential pressure between the inner and outer sides and the liquid holdup at the 45° section of a right-angle bend with an elliptical or circular cross-section, and combining this with the flow rate measurement formula proposed in this invention, the mixed mass flow rate of the gas-liquid two-phase flow in the bend is measured, thus developing a novel gas-liquid two-phase flow rate measurement technology.

[0006] A method for detecting the flow rate of a gas-liquid two-phase flow in a bend pipe based on acoustic emission and ELMAN is characterized by the following steps:

[0007] Step 1: Using the forced vortex theory of bent pipes, the single-phase flow measurement formula for circular cross-sections is extended to elliptical cross-sections, resulting in single-phase flow measurement formulas applicable to both circular and elliptical cross-sections, as shown in Equation 1:

[0008]

[0009] Wherein, G - single-phase mass flow rate, kg / s; α - flow coefficient, determined by experimental measurement; a - half-axis length of the 45° section of the bend perpendicular to the direction of differential pressure measurement, m; ρ - single-phase fluid density, kg / m³ 3 ; Δp - differential pressure between the inside and outside of the 45° section of the bend, Pa; R - radius of curvature at the axis of the bend, m; b - half-axis length of the 45° section of the bend along the direction of differential pressure measurement, m; C - constant term, zero-point offset correction of the differential pressure instrument, which is 0 when no offset occurs; For this invention, a is the long half-axis, b is the short half-axis, and when the section of the bend is circular, a = b;

[0010] Step 2: Assume the gas phase and liquid phase are located on the inside and outside of the bend, respectively, and that the actual velocities of the gas and liquid phases are equal and equal to the mixing velocity v. m The differential pressure between the inner and outer sides of the gas and liquid phases at the 45° section of the bend is obtained, as shown in Equations 2 and 3:

[0011]

[0012]

[0013] Where, Δp g - The differential pressure between the inner and outer sides of the gas phase region at a 45° bend in the pipe, Pa; H L - Cross-sectional liquid holdup, value to be measured later; ρ g - Gas phase fluid density, kg / m³ 3 ;v m - Gas-liquid two-phase flow mixing velocity, m / s; Δp L - The differential pressure between the inner and outer sides of the liquid phase region at a 45° bend in the pipe, Pa; ρ L -Liquid phase fluid density, kg / m³ 3 ;

[0014] Step 3: Obtain the differential pressure of each phase according to the differential pressure formulas for the inner and outer sides of the bend for the gas and liquid phases shown in Equations 2 and 3. Add the results calculated by Equations 2 and 3 to obtain the total differential pressure Δp between the inner and outer sides of the bend and the mixing velocity v of the gas-liquid two-phase flow. m The relationship is shown in Equations 4 and 5. Further, by combining Equation 6, we obtain the formula for calculating the mass flow rate of the gas-liquid two-phase flow in the bend, as shown in Equation 7.

[0015]

[0016]

[0017] G m =ρ m v m πab (6)

[0018]

[0019] Wherein, Δp is the total differential pressure between the inner and outer sides measured at the 45° section of the bend, in Pa; G m - Mass flow rate of gas-liquid mixture, kg / s; ρ m -Density of the mixture fluid, kg / m³ 3 ;ρ L -Liquid phase fluid density, kg / m³ 3 ;ρ g - Gas phase fluid density, kg / m³ 3 H L- Liquid holdup of the section; a - Major semi-axis of the bend section; b - Minor semi-axis of the bend section, when the bend section is circular, a = b; R - Radius of curvature at the bend axis, m;

[0020] Equation 7 is then corrected using the flow coefficient α and the correction constant C. The optimized formula for measuring the mass flow rate of the gas-liquid two-phase flow mixture is shown in Equation 8.

[0021]

[0022] Where α is the flow coefficient, determined by experimental measurements; C is a constant term, used for zero-point offset correction of the differential pressure instrument, which is 0 when no offset occurs.

[0023] Step 4: Determine the structural parameters of the pipe bending measuring element, including the radius of curvature R at the pipe bending axis and the major and minor semi-axles of the cross section;

[0024] Step 5: Allow the gas-liquid two-phase flow to pass through the bend measurement structure under different flow conditions. At the 45° section of the bend with an elliptical or circular cross-section, measure the acoustic emission signal and pressure difference signal of the gas-liquid two-phase flow passing through the 45° section of the bend using a differential pressure sensor and an acoustic emission sensor. Combine this with a visualization experimental mode and the existing acoustic emission-BP neural network-based flow pattern recognition system to obtain the flow pattern for each flow condition, thereby obtaining the acoustic emission parameters corresponding to different gas-liquid two-phase flows under different flow conditions.

[0025] When the gas-liquid two-phase flow medium passes through the bend measuring structure, the differential pressure sensor acquires the differential pressure signal Δp and transmits it to the computer via a data acquisition card. Based on existing technology, the cross-sectional liquid holdup H is then measured. L The mixed mass flow rate is calculated according to Equation 8.

[0026] When the gas-liquid two-phase flow medium flows through the bend measuring structure, the acoustic emission sensor acquires the acoustic emission signal and transmits it to the computer to calculate five statistical parameters of the acoustic emission signal of the gas-liquid two-phase flow in the bend: voltage amplitude, effective ring count, average signal level, root mean square, and absolute energy value.

[0027] Step 6: Process the acoustic emission signal of the gas-liquid two-phase flow in the bend pipe using 4-scale wavelet decomposition.

[0028] p Ei =E i / E total (9)

[0029] Where, p Ei - The ratio of wavelet energy in the i-th frequency band to the total wavelet energy; E i - Wavelet energy of the i-th frequency band, J; E total - Total wavelet energy, J; i - Wavelet decomposition frequency band code;

[0030] Step 7: Use the five acoustic emission statistical parameters from Step 5, the wavelet energy ratios obtained in Step 6, and the four bend structure parameters: the plane where the bend is located (0 is used when the bend is in a horizontal plane, and 1 is used when the bend is vertically upward), the major axis length 2a of the 45° section of the bend, the minor axis length 2b of the 45° section of the bend, and the radius of curvature R at the axis of the 45° section of the bend as the input layer neurons of the ELMAN neural network; the output parameter of this ELMAN neural network is the cross-sectional liquid holding capacity, therefore the number of output layer neurons is 1.

[0031] Step 8: A neural network model for measuring liquid holdup is thus constructed, denoted as the acoustic emission-ELMAN neural network;

[0032] Step 9: When actually measuring the flow rate of the gas-liquid two-phase flow in the bend, obtain the neurons of each input layer of the acoustic emission-ELMAN neural network and their respective flow conditions in accordance with the methods in steps 4, 5, and 6. Input each input layer neuron into the acoustic emission-ELMAN neural network to obtain the cross-sectional liquid holdup. Substitute the obtained cross-sectional liquid holdup into the flow measurement formula 8 with the obtained coefficients to perform the mass flow rate measurement of the gas-liquid two-phase flow mixture, thereby realizing the flow rate measurement.

[0033] In step 5, during the calculation process of Equation 8, the flow coefficient α and the constant term C are variable parameters. During the calculation, the differential pressure Δp and the cross-sectional liquid holdup H are first used... L The mixed mass flow rate without flow coefficient α and constant term C is calculated according to Equation 7. Next, the mixed mass flow rate obtained from Equation 7 and the known actual mixed mass flow rate are linearly fitted to obtain the slope and intercept of the linear fit, which correspond to the flow coefficient α and constant term C, respectively. According to this step, the flow coefficient α and constant term C under each bend structure can be obtained according to different flow patterns.

[0034] In step 6, the wavelet decomposition frequency band code i ranges from 1 to 16, therefore the number of input layer neurons is 25.

[0035] In step 7, the number of neurons in the hidden layer of the ELMAN neural network of the present invention is determined to be 50, the activation function is the tansig function, and the training function is traingdm, thereby obtaining the acoustic emission-ELMAN neural network liquid holding rate measurement model and realizing the measurement of cross-sectional liquid holding rate.

[0036] Advantages of the invention

[0037] This invention measures the liquid holdup at bends in pipe sections using acoustic emission combined with an ELMAN neural network. It then utilizes a proposed formula for measuring the mass flow rate of gas-liquid two-phase flow mixtures in bends to achieve flow detection. This establishes a formula applicable to both circular and elliptical cross-sections of bends, expanding its applicability. Furthermore, for bends, common interventional measurement techniques such as conductivity probes require pipe insulation, which is difficult to install on high-pressure steel pipes. Additionally, these methods can damage the bend structure when measuring the liquid holdup at a 45° angle, posing potential safety hazards. Acoustic emission detection, on the other hand, can collect data non-invasively through contact with the outer wall of the pipe, offering convenient on-site application value. The ELMAN neural network adds a receiving layer to the BP neural network structure, receiving feedback signals from the hidden layer and memorizing the output value of the hidden layer units from the previous moment, then transmitting it to the hidden layer along with the current input. Compared to feedforward neural networks, the ELMAN neural network has stronger dynamic information processing capabilities, global stability, and computational power. The use of acoustic emission combined with an ELMAN neural network and bends for gas-liquid two-phase flow rate detection has significant inventive value. This invention utilizes acoustic emission technology and ELMAN neural network learning to achieve accurate measurement of cross-sectional liquid holdup. Furthermore, by combining this cross-sectional liquid holdup with the formula for measuring the mass flow rate of gas-liquid two-phase flow in a bend, safe online flow rate measurement can be achieved. Attached Figure Description

[0038] Figure 1 This is a flowchart of the acoustic emission-ELMAN neural network-bend pipe gas-liquid two-phase flow detection process used in this invention.

[0039] Figure 2 This is a schematic diagram of an acoustic emission-ELMAN neural network-bent tube gas-liquid two-phase flow measurement device.

[0040] Among them, 1-differential pressure sensor, 2-acoustic emission sensor, 3-preamplifier, 4-signal acquisition card, 5-computer.

[0041] Figure 3 For liquid as a single-phase medium, the single-phase mass flow rate measurement curves of the liquid under four structural parameter bends used in this laboratory are presented according to Equation 1. The horizontal axis represents the actual measured value of the water mass flow rate at the inlet flow meter, and the vertical axis represents the calculated value of the water mass flow rate under the bend.

[0042] Figure 4For four common gas-liquid two-phase flow patterns in industry: (a) bubbly flow, (b) stratified flow, (c) annular flow, and (d) intermittent flow, the calculated mixed mass flow rate according to Equation 7 is compared with the measured value, and the fitted curve is Equation 8. The horizontal axis represents the bend-pipe calculated value of the air-water mixed mass flow rate according to Equation 7, and the vertical axis represents the gas-liquid mixed mass flow rate obtained from the actual measured values ​​of each phase of air and water through the inlet single-phase flow meter. Detailed Implementation

[0043] The hardware used in this invention includes a differential pressure sensor 1, an acoustic emission sensor 2, a preamplifier 3, a signal acquisition box 4, and a computer 5, such as... Figure 2 As shown, the acoustic emission sensor 2 is installed at the 45° section of the bend. The high-pressure tap of the differential pressure sensor 1 is located on the outside of the 45° section of the bend, and the low-pressure tap is located on the inside of the 45° section of the bend. The pipe material is stainless steel. The bend is detachable. When the bend is placed horizontally, both the upstream and downstream pipes are in a horizontal plane. When the bend is installed vertically upward, the upstream is horizontal and the downstream is vertically upward, and the entire bend system is in a vertically upward plane. The acoustic emission signal and the differential pressure signal are acquired by the preamplifier 3 and then by the data acquisition card 4, and then transmitted to the computer 5. The computer performs raw data processing and realizes mixed mass flow measurement.

[0044] The specific method is as follows:

[0045] A method for detecting the flow rate of a gas-liquid two-phase flow in a bend pipe based on acoustic emission and ELMAN is characterized by the following steps:

[0046] Step 1: Using the forced vortex theory of bent pipes, the single-phase flow measurement formula for circular cross-sections is extended to elliptical cross-sections, resulting in single-phase flow measurement formulas applicable to both circular and elliptical cross-sections, as shown in Equation 1:

[0047]

[0048] Wherein, G - single-phase mass flow rate, kg / s; α - flow coefficient, determined by experimental measurement; a - semi-axis length of the 45° section of the bend perpendicular to the differential pressure direction between the inside and outside, m; ρ - single-phase fluid density, kg / m³ 3 ; Δp - differential pressure between the inside and outside of the 45° section of the bend, Pa; R - radius of curvature at the axis of the bend, m; b - half-axis length of the 45° section of the bend along the differential pressure direction between the inside and outside, m; C - constant term, zero-point offset correction of the differential pressure instrument, which is 0 when no offset occurs; For this invention, a is the major half-axis, b is the minor half-axis, and when the bend section is circular, a = b;

[0049] Figure 3Using liquid as a single-phase medium, the single-phase mass flow rate measurement curves of liquid under four structural parameters of bends used in this laboratory were obtained according to Equation 1. The figure shows that Equation 1 provides excellent linearity for the single-phase mass flow rate measurement curves under both circular and elliptical cross-section bends, reflecting the effectiveness of Equation 1.

[0050] Step 2: Assume the gas phase and liquid phase are located on the inside and outside of the bend, respectively, and that the actual velocities of the gas and liquid phases are equal to the mixing velocity v. m The differential pressure between the inner and outer sides of the gas and liquid phases at the 45° section of the bend is obtained, as shown in Equations 2 and 3:

[0051]

[0052]

[0053] Where, Δp g - The differential pressure between the inner and outer sides of the gas phase region at a 45° bend in the pipe, Pa; H L - Cross-sectional liquid holdup, value to be measured later; ρ g - Gas phase fluid density, kg / m³ 3 ;v m - Gas-liquid two-phase flow mixing velocity, m / s; Δp L - The differential pressure between the inner and outer sides of the liquid phase region at a 45° bend in the pipe, Pa; ρ L -Liquid phase fluid density, kg / m³ 3 ;

[0054] Step 3: Obtain the differential pressure of each phase according to the differential pressure formulas for the gas and liquid phases inside and outside the bend, as shown in Equations 2 and 3. Add the results calculated by Equations 2 and 3 to obtain the total differential pressure Δp between the inside and outside of the bend and the mixing velocity v of the gas-liquid two-phase flow. m The relationship is shown in Equations 4 and 5. Further, by combining Equation 6, we obtain the formula for calculating the mass flow rate of the gas-liquid two-phase flow mixture in the bend, as shown in Equation 7.

[0055]

[0056]

[0057] G m =ρ m v m πab (6)

[0058]

[0059] Wherein, Δp is the differential pressure between the inner and outer sides of the 45° section of the bend, in Pa; G m - Mass flow rate of gas-liquid mixture, kg / s; ρ m -Density of the mixture fluid, kg / m³3 ;ρ L -Liquid phase fluid density, kg / m³ 3 ;ρ g - Gas phase fluid density, kg / m³ 3 H L - Liquid holdup of the section; a - Major semi-axis of the bend section; b - Minor semi-axis of the bend section, when the bend section is circular, a = b; R - Radius of curvature at the bend axis, m;

[0060] However, in actual measurement, the theoretical measurement value may be deviated due to the influence of structure and instrumentation. Therefore, correction is made by using the flow coefficient α and the correction constant C. Thus, the optimized formula for measuring the mass flow rate of gas-liquid two-phase flow mixture adopted in this invention is shown in Equation 8:

[0061]

[0062] Where α is the flow coefficient, determined by experimental measurements; C is a constant term, used for zero-point offset correction of the differential pressure instrument, which is 0 when no offset occurs.

[0063] Step 4: Determine the structural parameters of the pipe bending measuring element, including the radius of curvature R at the pipe bending axis and the major and minor semi-axles of the cross section;

[0064] Step 5: Under different flow conditions, allow the gas-liquid two-phase flow to pass through the bend measurement structure. At the 45° section of the bend with an elliptical or circular cross-section, measure the acoustic emission signal and pressure difference signal of the gas-liquid two-phase flow passing through the 45° section of the bend using acoustic emission sensors and differential pressure sensors. Combine this with a visualization experimental mode and the flow pattern recognition system based on acoustic emission-BP neural network previously developed in our laboratory to obtain the flow pattern for each flow condition; where:

[0065] When the gas-liquid two-phase flow medium flows through the bent tube measuring structure, the differential pressure sensor 1 acquires the differential pressure signal Δp and transmits it to the computer 5 via the data acquisition card 4. Based on existing technology, the cross-sectional liquid holdup H is measured. L And calculate according to Equation 8;

[0066] In the calculation process of Equation 8, the flow coefficient α and the constant term C are variable parameters. During the calculation, the differential pressure Δp and the cross-sectional liquid holdup H under a certain flow pattern are first used. LThe mixed mass flow rate without flow coefficient α and constant term C is calculated according to Equation 7. Next, the mixed mass flow rate obtained from Equation 7 is linearly fitted with the known actual mixed mass flow rate (the sum of the inlet single-phase flowmeter measurements, obtained through a gas-liquid two-phase flow calibration experimental system) to obtain the slope and intercept, which correspond to the flow coefficient α and constant term C under this flow pattern. It can be seen that, based on the gas-liquid flow rate experimental calibration process of the bend measurement system, the flow coefficient α and constant term C corresponding to various structural parameters of the bend (Table 1) under various flow patterns can be obtained, as shown in Table 2.

[0067] When the gas-liquid two-phase flow medium flows through the bend measuring structure, while acquiring the pressure difference signal, the acoustic emission sensor 2 acquires the acoustic emission signal and transmits it to the computer 5 to calculate five statistical parameters of the acoustic emission signal of the gas-liquid two-phase flow in the bend: voltage amplitude, effective ring count, average signal level, root mean square, and absolute energy value.

[0068] Step 6: The acoustic emission signal of the gas-liquid two-phase flow in the bend pipe is processed using 4-scale wavelet decomposition. The sampling frequency of the acoustic emission in this laboratory is 2000kHz. According to Shannon's sampling theorem, the Nyquist frequency is 1000kHz. Therefore, after 4-scale wavelet decomposition, this example will obtain 16 wavelet signals with a frequency band interval of 62.5kHz and their respective wavelet energies E. i And according to Equation 7, the energy ratio of the 16 wavelet nodes is obtained;

[0069] p Ei =E i / E total (9)

[0070] Where, p Ei - The ratio of wavelet energy in the i-th frequency band to the total wavelet energy; E i - Wavelet energy of the i-th frequency band, J; E total - Total wavelet energy, J; i - Wavelet decomposition frequency band code, preferably ranging from 1 to 16;

[0071] Step 7: The 5 acoustic emission statistical parameters from Step 5, the 16 wavelet energy ratios from Step 6, and the 4 bend structure parameters—the plane where the bend is located (0 for a horizontal plane and 1 for a vertically upward bend), the major axis length 2a of the 45° section of the bend, the minor axis length 2b of the 45° section of the bend, and the radius of curvature R at the axis of the 45° section of the bend—are used as the input layer of the ELMAN neural network of this invention. Therefore, the number of neurons in the input layer is 25. Since the output parameter of this invention is the cross-sectional liquid holding capacity, the number of neurons in the output layer is 1.

[0072] Step 8: After extensive experimental verification, it was determined that the number of neurons in the hidden layer of the ELMAN neural network of the present invention is 50, the activation function is the tansig function, the training function is traingdm, and the model is learned by 942 sets of gas-liquid two-phase flow experimental data obtained in this laboratory, thereby constructing an acoustic emission-ELMAN neural network liquid holdup measurement model for realizing cross-sectional liquid holdup measurement.

[0073] Step 9: When actually measuring the flow rate of the gas-liquid two-phase flow in the bend, obtain the neurons of each input layer of the acoustic emission-ELMAN neural network and the corresponding flow conditions in accordance with the methods in steps 4, 5, and 6. Input each input layer neuron into the acoustic emission-ELMAN neural network to obtain the cross-sectional liquid holdup. Substitute the obtained cross-sectional liquid holdup into the flow measurement formula with known parameters (Equation 8 and Table 2) to perform the gas-liquid two-phase flow mixing mass flow rate measurement and obtain the measurement result.

[0074] Example

[0075] To describe in detail the measurement steps of this invention, the following is in conjunction with the appendix. Figure 1 Described using a flowchart:

[0076] First, to achieve flow rate measurement of gas-liquid two-phase flow, this invention requires measuring the structural parameters of the bend, including the length of the major axis of the 45° section of the bend, the length of the minor axis of the 45° section of the bend, and the radius of curvature at the axis of the 45° section of the bend.

[0077] Then, the flow pattern of each gas-liquid two-phase flow condition is obtained according to the acoustic emission-BP neural network flow pattern recognition system of this laboratory. In this example, the flow patterns of gas-liquid two-phase flow include bubbly flow, stratified flow, annular flow and intermittent flow.

[0078] After determining the structural parameters of the bend component, it is necessary to obtain the flow coefficient α and correction constant C under each flow pattern in Equation 8. The specific implementation steps are as follows: First, determine the flow conditions of the gas-liquid two-phase flow according to the flow pattern on the gas-liquid two-phase flow experimental device, determine the inlet apparent gas velocity and apparent liquid velocity, and obtain the actual mixed mass flow rate; Next, obtain the cross-sectional liquid holdup and the differential pressure inside and outside the bend under the flow conditions through dual parallel conductivity probes and differential pressure sensors; Then, substitute the obtained cross-sectional liquid holdup and differential pressure inside and outside the bend into Equation 7; Finally, perform linear fitting between the measurement results of Equation 7 and the actual mixed mass flow rate value to determine the flow coefficient α and correction constant C, and obtain Equation 8.

[0079] Figure 4 The figure shows the fitted curves for measurements based on Equation 8 under four common industrial flow patterns: bubbly flow, stratified flow, annular flow, and intermittent flow. To correct the measurement results using Equation 8, it is necessary to calculate the flow coefficient α and the correction constant C caused by the zero-point offset of the measuring instrument. The specific implementation process is as follows... Figure 4As shown: First, the predicted value without measurement parameters is calculated using Equation 7 based on the measurement data. Figure 4 (discrete data points), and then a linear fit is applied to the predicted and actual values ​​(...). Figure 4 The slope and intercept are obtained from the fitted curve, thus determining the measurement parameters of Equation 8. Figure 4 It can be seen that the measurement curves under each flow pattern have a good linear relationship.

[0080] After modification, Formula 8 becomes capable of measuring gas-liquid two-phase flow in industrial pipelines. While dual parallel conductivity probes can obtain the cross-sectional liquid holdup, they pose potential installation hazards, and this method is unsuitable for high-pressure steel pipes in the field. Therefore, to obtain the cross-sectional liquid holdup more safely and effectively, this invention employs an acoustic emission-ELMAN neural network measurement method. The specific steps involve first measuring the acoustic emission signal generated at a 45° cross-section of the gas-liquid two-phase flow passing through a bend, and simultaneously measuring the pressure difference signal between the inside and outside of the bend. After measuring the acoustic emission signal, the proposed input layer of this invention consists of 25 neurons: the installation plane (0 for a horizontal bend and 1 for a vertically upward bend), the major axis length 2a of the 45° section of the bend, the minor axis length 2b of the 45° section of the bend, the radius of curvature at the axis of the 45° section of the bend, the wavelet energy ratio of 16 frequency bands in the 4-scale wavelet decomposition of the acoustic emission signal, and five statistical parameters: acoustic emission signal voltage amplitude, effective ring count, average signal level, root mean square, and absolute energy value; the hidden layer consists of 50 neurons; and the output layer consists of one neuron (section liquid holdup). This is used for learning, resulting in an acoustic emission-ELMAN neural network capable of measuring section liquid holdup. Experimental verification in our laboratory of 94 sets of gas-liquid two-phase flow conditions in bends shows that the acoustic emission-ELMAN neural network proposed in this invention achieves a prediction accuracy of over 90% for measuring the section liquid holdup of pipes.

[0081] Finally, the liquid holdup obtained from the acoustic emission-ELMAN neural network and the pressure difference signal between the inside and outside of the bend measured by the differential pressure instrument were substituted into Equation 8, which determines the flow coefficient α and the correction constant C under the known flow pattern. The mass flow rate of the gas-liquid two-phase flow mixture was then calculated, thus achieving measurement. The verification results of the acoustic emission-ELMAN neural network-bend gas-liquid two-phase flow mass flow measurement in this study show that the average measurement errors for bubbly flow, stratified flow, annular flow, and intermittent flow are 4.5%, 11.5%, 13%, and 22.6%, respectively.

[0082] The following parameters were used to verify this invention:

[0083] To verify the general applicability of this measurement method, four types of bent pipe structures were used in the experiment, including circular and elliptical cross-sections. The cross-sectional dimensions and radii of curvature at the 45° section of the bent pipe under the four structures are given in Table 1.

[0084] Table 1 Structural Parameters of the Bend

[0085]

[0086] To further demonstrate the applicability of the present invention, the flow coefficient α and correction constant C of the four types of bends in the horizontal and vertical planes were measured. The measurement parameters are shown in Table 2.

[0087] Table 2 Measurement parameters for each bend structure

[0088]

[0089] Based on the acoustic emission-ELMAN neural network, the liquid holdup of the cross section under different working conditions in four types of bend structures under two types of mounting planes was measured. The results are shown in Table 3. According to the comparison in Table 3, it can be seen that the method can accurately measure the liquid holdup of the cross section under various bend structures.

[0090] Table 3 Examples of liquid holdup measurement results based on acoustic emission-ELMAN neural network

[0091]

[0092]

[0093] Note: There are a total of 95 test cases. Due to space limitations, only 4 cases of each structure are shown, for a total of 32 examples.

[0094] Table 4 shows an example of the flow measurement results of the acoustic emission-ELMAN neural network-bent pipe gas-liquid two-phase flow used in this invention. According to the results in the table, it can be seen that the gas-liquid two-phase flow mixing mass flow measurement technology used in this invention can accurately measure bubbly flow, stratified flow, annular flow and intermittent flow under various bend structures.

[0095] Table 4 shows the measurement results of the mass flow rate of gas-liquid two-phase flow based on acoustic emission-ELMAN neural network-bend pipe.

[0096] Serial Number flow pattern <![CDATA[u GS m / s]]> <![CDATA[u LS m / s]]> <![CDATA[True G m kg / s]]> <![CDATA[Measurement of G m kg / s]]> error% 1 bubble flow 0.72 1.89 1.001 0.975 2.66 2 bubble flow 0.39 1.68 0.891 0.901 1.18 3 bubble flow 0.73 2.09 1.098 1.152 4.94 4 bubble flow 0.54 1.65 0.875 0.855 2.32 5 Circular flow 25.12 0.57 0.275 0.272 0.69 6 Circular flow 17.27 0.26 0.183 0.171 6.93 7 Circular flow 20.88 0.38 0.178 0.173 2.78 8 Circular flow 30.70 0.52 0.242 0.222 8.10 9 Layered Flow 0.30 0.26 0.169 0.147 12.78 10 Layered Flow 0.27 0.68 0.446 0.495 11.08 11 Layered Flow 2.46 0.50 0.219 0.237 8.71 12 Layered Flow 6.10 0.26 0.115 0.129 12.28 13 intermittent flow 0.45 0.65 0.282 0.277 1.67 14 intermittent flow 3.72 1.49 0.649 0.721 10.96 15 intermittent flow 0.72 0.42 0.222 0.222 0.07 16 intermittent flow 7.27 1.55 0.816 0.959 17.60

[0097] Note: There are a total of 95 test cases. Due to space limitations, only 4 cases are shown for each flow type, for a total of 16 examples.

Claims

1. A method for detecting the flow rate of a gas-liquid two-phase flow in a bent pipe based on acoustic emission and ELMAN, characterized in that: The following steps are involved: Step 1: Using the forced vortex theory of bent pipes, the single-phase flow measurement formula for circular cross-sections is extended to elliptical cross-sections, resulting in single-phase flow measurement formulas applicable to both circular and elliptical cross-sections, as shown in Equation 1: Wherein, G - single-phase mass flow rate, kg / s; α - flow coefficient, determined by experimental measurement; a - half-axis length of the 45° section of the bend perpendicular to the direction of differential pressure measurement, m; ρ - single-phase fluid density, kg / m³ 3 ; Δp - differential pressure between the inside and outside of the 45° section of the bend, Pa; R - radius of curvature at the axis of the bend, m; b - half-axis length of the 45° section of the bend along the direction of differential pressure measurement, m; C - constant term, zero-point offset correction of the differential pressure instrument, which is 0 when no offset occurs; For this invention, a is the long half-axis, b is the short half-axis, and when the section of the bend is circular, a = b; Step 2: Assume the gas phase and liquid phase are located on the inside and outside of the bend, respectively, and that the actual velocities of the gas and liquid phases are equal and equal to the mixing velocity v. m The differential pressure between the inner and outer sides of the gas and liquid phases at the 45° section of the bend is obtained, as shown in Equations 2 and 3: Where, Δp g - The differential pressure between the inner and outer sides of the gas phase region at a 45° bend in the pipe, Pa; H L - Cross-sectional liquid holdup, value to be measured later; ρ g - Density of gas phase fluid, kg / m³ 3 ;v m - Gas-liquid two-phase flow mixing velocity, m / s; Δp L - The differential pressure between the inner and outer sides of the liquid phase region at a 45° angle to the bend in the pipe, Pa; ρ L -Liquid phase fluid density, kg / m³ 3 ; Step 3: Obtain the differential pressure of each phase according to the differential pressure formulas for the gas and liquid phases inside and outside the bend, as shown in Equations 2 and 3. Add the results calculated by Equations 2 and 3 to obtain the total differential pressure Δp between the inside and outside of the bend and the mixing velocity v of the gas-liquid two-phase flow. m The relationship is shown in Equations 4 and 5. Further, by combining Equation 6, we obtain the formula for calculating the mass flow rate of the gas-liquid two-phase flow in the bend, as shown in Equation 7. G m =ρ m v m pub (6) Among them, G m - Mass flow rate of gas-liquid mixture, kg / s; ρ m -Density of the mixture fluid, kg / m³ 3 ; Equation 7 is then corrected using the flow coefficient α and the correction constant C. The optimized formula for measuring the mass flow rate of the gas-liquid two-phase flow mixture is shown in Equation 8. Step 4: Determine the structural parameters of the pipe bending measuring element, including the radius of curvature R at the pipe bending axis and the major and minor semi-axles of the cross section; Step 5: Allow the gas-liquid two-phase flow to pass through the bend measurement structure under different flow conditions. At the 45° section of the bend with an elliptical or circular cross-section, measure the acoustic emission signal and pressure difference signal of the gas-liquid two-phase flow passing through the 45° section of the bend using a differential pressure sensor and an acoustic emission sensor. Combine this with a visualization experimental mode and the existing acoustic emission-BP neural network-based flow pattern recognition system to obtain the flow pattern for each flow condition, thereby obtaining the acoustic emission parameters corresponding to different gas-liquid two-phase flows under different flow conditions. When the gas-liquid two-phase flow medium passes through the bend measuring structure, the differential pressure sensor acquires the differential pressure signal Δp and transmits it to the computer via a data acquisition card. Based on existing technology, the cross-sectional liquid holdup H is then measured. L The mixed mass flow rate is calculated according to Equation 8. When the gas-liquid two-phase flow medium flows through the bend measuring structure, the acoustic emission sensor acquires the acoustic emission signal and transmits it to the computer to calculate five statistical parameters of the acoustic emission signal of the gas-liquid two-phase flow in the bend: voltage amplitude, effective ring count, average signal level, root mean square, and absolute energy value. Step 6: Process the acoustic emission signal of the gas-liquid two-phase flow in the bend pipe using 4-scale wavelet decomposition. p Ei =And i / AND total (9) Where, p Ei - The ratio of wavelet energy in the i-th frequency band to the total wavelet energy; E i - Wavelet energy of the i-th frequency band, J; E total - Total wavelet energy, J; i - Wavelet decomposition frequency band code; Step 7: Use the five acoustic emission statistical parameters from Step 5, the wavelet energy ratios obtained in Step 6, and the four structural parameters of the bend: the plane where the bend is located, the length of the major axis of the 45° section of the bend (2a), the length of the minor axis of the 45° section of the bend (2b), and the radius of curvature R at the axis of the 45° section of the bend as the input layer neurons of the ELMAN neural network. A value of 0 is used when the bend is in a horizontal plane, and a value of 1 is used when the bend is vertically upward. The output parameter of this ELMAN neural network is the liquid holding capacity of the section; therefore, the number of neurons in the output layer is one. Step 8: A neural network model for measuring liquid holdup is thus constructed, denoted as the acoustic emission-ELMAN neural network; Step 9: When actually measuring the flow rate of the gas-liquid two-phase flow in the bend, obtain the neurons of each input layer of the acoustic emission-ELMAN neural network and their respective flow conditions in accordance with the methods in steps 4, 5, and 6. Input each input layer neuron into the acoustic emission-ELMAN neural network to obtain the cross-sectional liquid holdup. Substitute the obtained cross-sectional liquid holdup into the flow measurement formula 8 with the obtained coefficients to perform the mass flow rate measurement of the gas-liquid two-phase flow mixture, thereby realizing the flow rate measurement.

2. The method for detecting the flow rate of a gas-liquid two-phase flow in a bent pipe based on acoustic emission and ELMAN as described in claim 1, characterized in that: In step 6, the wavelet decomposition frequency band code i ranges from 1 to 16, therefore the number of input layer neurons is 25.

3. The flow rate detection method for gas-liquid two-phase flow in a bent pipe based on acoustic emission and ELMAN as described in claim 2, characterized in that: In step 7, the number of neurons in the hidden layer is determined to be 50, the activation function is the tansig function, and the training function is traingdm, thereby obtaining the acoustic emission-ELMAN neural network liquid holding rate measurement model and realizing the measurement of cross-sectional liquid holding rate.

Images