Method for changing phase uniformity of flow field based on deep learning to reduce pipeline vibration

Abstract

The invention discloses a method for changing phase uniformity of a flow field based on deep learning to reduce vibration of a pipeline, and belongs to the technical field of pipeline transportation. The pipeline comprises a static assembly, a dynamic assembly and a sealing base, and the static assembly comprises a rectification grating element which rectifies fluid and enables the phases to be mixed, dispersed and uniform; the dynamic assembly comprises a blade rotating shaft and blades; the method further comprises an angle measuring module, a vibration process data collecting module, a deep learning system and a servo driving system, vibration, process data and angle data serve as learning parameters of the deep learning system, an agent model of the deep learning system and vibration is established, when the working condition changes, the optimal angle is automatically fed back, and blade rotation is completed through the servo driving system. The method is high in adaptability, and vibration can be adaptively adjusted and reduced; the device is particularly suitable for the vibration problem caused by multiphase flow conveying, large-diameter pipelines and gravity control flow pipelines, the uniformity of a flow field is actively adjusted and controlled, accumulated bias flow between phases is destroyed, and therefore pipeline vibration is reduced.

Description

Technical Field

[0001] The present invention relates to the field of pipeline transportation, and specifically to a method for reducing pipeline vibration by changing the phase uniformity of the flow field based on deep learning. Background Art

[0002] The problem of pipeline vibration occupies a crucial position in fields such as fluid transportation and mechanical engineering. Especially in the process of industrial production and energy transmission, in the transportation of long-distance and large-diameter pipelines, the coalescence effect of different phases cannot be ignored. The flow pattern of fluid flow changes into complex flow, and the resulting pressure fluctuations and the unstable gas-liquid two-phase flow in the pipeline will couple with the pipeline wall, thereby inducing severe pipeline vibration, affecting the safe operation of the entire pipeline system, and causing a series of safety hazards and economic losses.

[0003] In Patent CN107387635A, the resonance of the sub-absorber is used to absorb vibration energy, thereby reducing pipeline vibration. The pre-compression amount adjustment of the metal rubber block requires precise control, otherwise it may affect the vibration reduction effect, and the metal rubber block in a special environment may fail, affecting the vibration reduction performance. In Patent CN113007447A, the vibration energy is transmitted through the pipe clamp, and the metal rubber block relies on non-linear deformation to absorb and dissipate vibration energy, thereby achieving the vibration reduction effect. However, the metal bellows may experience fatigue failure during long-term use, affecting the vibration reduction effect, and the setting of the anti-impact limit block may limit the movement range of the metal mass block, affecting the dynamic performance of the absorber.

[0004] The existing pipeline vibration reduction and stress optimization technologies have the following deficiencies: 1. Limited adjustment ability: Traditional vibration reduction devices usually adopt fixed structures or simple mechanical adjustment methods, and are set outside the pipeline, making it difficult to adapt to complex working condition changes. Especially, the strong low-frequency vibration of the pipeline caused by flow patterns such as stratified flow and slug flow caused by phase accumulation. 2. Slow response speed: After the existing technology detects vibration or stress changes, it requires manual intervention or a complex control system for adjustment, with a slow response speed, unable to respond to sudden working condition changes in a timely manner, and easily causing excessive wear and damage to equipment. 3. Poor adaptability: The characteristics of different pipeline systems, fluid media, and pipeline installation environments vary greatly, and existing devices often struggle to meet diverse application requirements, increasing costs and complexity. Summary of the Invention

[0005] In order to solve the above problems, strengthen the pipeline vibration reduction adjustment ability, response speed, and environmental adaptability, and provide real-time feedback on the vibration reduction and stress control effects of the adjustment device, it is necessary to design a method for reducing pipeline vibration by changing the phase uniformity of the flow field based on deep learning to solve the above technical problems.

[0006] The technical solution adopted by the present invention is as follows: A method for changing the phase uniformity of the flow field based on deep learning to reduce pipeline vibration, characterized in that the pipeline is respectively connected to the incoming flow pipeline and the outgoing flow pipeline through flanges; the pipeline includes a static component, a dynamic component, and a sealing base; the dynamic component includes a blade rotating shaft and blades, the blade rotating shaft is arranged on the sealing base, and the blades are connected to the blade rotating shaft; An angle measurement module and a servo drive system are arranged on the blade rotating shaft, and a vibration process data acquisition module is arranged on the incoming flow pipeline and the outgoing flow pipeline; the deep learning system is electrically connected to the angle measurement module, the vibration process data acquisition module, and the servo drive system; The angle measurement module includes positioning sensors respectively arranged on different blade rotating shafts; The vibration process data acquisition module includes a pressure sensor, a temperature sensor, and a flow sensor arranged on the incoming flow pipeline; a displacement sensor, a velocity sensor, an acceleration sensor, a pressure sensor, a temperature sensor, and a flow sensor arranged on the outgoing flow pipeline; The servo drive system adopts a motor set connected to the blade rotating shaft; The deep learning system uses the data output by the angle measurement module and the vibration process data acquisition module as a data set, and based on the principle of deep learning, automatically learns and establishes a surrogate model between the incoming flow condition, vibration displacement, velocity acceleration, and angle; when the incoming flow condition changes, the surrogate model is used to feedback the optimal configuration angle, and the servo drive system responds according to the feedback signal and automatically adjusts the rotation of the blade rotating shaft to the optimal configuration angle.

[0007] Further, the static component includes a rectifying grille element, and the rectifying grille element includes a baffle grille and a grid grille.

[0008] Further, the baffle grille includes baffle plates arranged in a staggered manner, and two adjacent baffle grilles are respectively installed at 180° along the pipeline axis; The grid grille is evenly distributed with square holes.

[0009] Further, the grille is connected to the sealing base by welding; The blade rotating shaft is fixed on the sealing base through bearings, and the motor set drives the blade rotating shaft to achieve the rotation function, and the blades are welded and fixed on the blade rotating shaft.

[0010] Further, the static component: includes a rectifying grille element, and the grille includes various types, including but not limited to baffle grilles, "grid" grilles, orifice plates, etc., which destroy the accumulation of phases in the fluid through cutting, baffle, etc., and stabilize the incoming flow and reduce vibration; the number of the grille, blades, and blade rotating shafts can be increased or decreased; Angle measurement module: A positioning sensor is configured on the blade rotation axis; Vibration process data acquisition module: A first pressure sensor, a first temperature sensor, and a first flow sensor are arranged on the incoming flow pipeline; a first displacement sensor, a first velocity sensor, a first acceleration sensor, a second pressure sensor, a second temperature sensor, and a second flow sensor are arranged on the outgoing flow pipeline; Servo drive system: Includes a first motor, a second motor, a third motor to a fourteenth motor connected to the blade rotation axis; Deep learning system: After removing abnormal signals from the output data sets of the angle measurement module and the vibration process data acquisition module, the available data sets are used. The data sets are divided into a training set, a validation set, and a test set. Based on the principle of deep learning, the parameters of the model are initially determined through the training set and the validation set, and the generalization ability of the model is tested using the test set. An agent model between the incoming flow conditions, vibration displacement, velocity acceleration, and angle is automatically learned and established; when the incoming flow conditions change, the optimal angle command signal is fed back using the agent model, and the servo drive system quickly responds according to the feedback signal to automatically adjust the rotation of the blade rotation axis. The angle measurement module feeds back the rotated angle to the deep learning system in real time to confirm the deformation accuracy.

[0011] Output display system: Used to display relevant data and information; The sealing base is connected to the incoming flow pipeline and the outgoing flow pipeline through flanges, and a bearing is provided inside. The blade rotation axis is fixed to the sealing base through the bearing, and the blade is fixed to the blade rotation axis by welding; The first motor, the second motor, the third motor to the fourteenth motor respectively control the angle adjustment of the blade rotation axis.

[0012] The deep learning system uses the output data of the angle measurement module and the vibration process data acquisition module as available data sets, divides the data sets into a training set, a validation set, and a test set, and automatically learns and establishes an agent model between the incoming flow conditions, vibration displacement, velocity acceleration, and angle based on the principle of deep learning; when the incoming flow conditions change or phenomena such as external load changes or water hammer phenomena occur upstream, the optimal shape and position angle is fed back using the agent model, and the servo drive system quickly responds according to the feedback signal to automatically adjust the rotation of the blade rotation axis. The angle measurement module feeds back the rotated angle to the deep learning system in real time to confirm the deformation accuracy.

[0013] The beneficial effects of the present invention are as follows: 1. The existing pipeline vibration reduction methods usually adopt fixed structures or simple mechanical adjustment methods, and are arranged outside the pipeline, making it difficult to meet the requirements of complex working conditions. The static components and dynamic components proposed in the present invention achieve changes in the flow field uniformity under different structures through combined spoiler and mixing designs, and solve the stratified flow or slug flow caused by phase accumulation without replacing components.

[0014] 2. The present invention has an adaptive adjustment function: integrating an angle measurement module and a vibration process data acquisition module, it can monitor the vibration state of the pipeline in real time. The deep learning system uses the vibration, process data, and angle data as the learning parameters of the deep learning system to establish a proxy model with vibration. When the working conditions change, it automatically feedbacks the optimal angle to reduce vibration without manual intervention, improving the timeliness and accuracy of adjustment.

[0015] 3. The present invention adopts high-precision measurement and control components, which can accurately obtain vibration data and achieve high-precision rotational adjustment through a servo motor. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The present invention will be further described below with reference to the drawings and embodiments.

[0017] Figure 1 It is a structural diagram of a method for changing the phase uniformity of the flow field based on deep learning to reduce pipeline vibration.

[0018] Figure 2 It is a structural diagram of a folded grid.

[0019] Figure 3 It is a schematic arrangement diagram of the folded grid.

[0020] Figure 4 It is a structural diagram of a grid in the shape of a well.

[0021] Figure 5 It is a structural diagram of the pipeline part of this method.

[0022] Figure 6 It is a top view of the arrangement of the blade rotation axes in the pipeline.

[0023] In the figure: 1. Folded grid; 2. Grid in the shape of a well; 3. Blade rotation axis; 3a. First horizontal blade rotation axis; 3b. Second horizontal blade rotation axis; 3c. Third horizontal blade rotation axis; 3d. Fourth horizontal blade rotation axis; 3e. Fifth horizontal blade rotation axis; 3f. Sixth horizontal blade rotation axis; 3g. Seventh horizontal blade rotation axis; 3h. First vertical blade rotation axis; 3i. Second vertical blade rotation axis; 3j. Third vertical blade rotation axis; 3k. Fourth vertical blade rotation axis; 3l. Fifth vertical blade rotation axis; 3m. Sixth vertical blade rotation axis; 3n. Seventh vertical blade rotation axis; 4. Blade; 5. Sealing base. Detailed implementation manners Embodiment

[0024] Figure 1 The structure diagram of a method for changing the phase uniformity of a flow field based on deep learning to reduce pipeline vibration is shown. In the figure, this method includes a static component, a dynamic component, and a sealing base 5.

[0025] The static component includes a rectifying grid element, where the folded grid 1 includes folded plates arranged staggeredly (see Figure 2 ), and two folded grids 1 are respectively installed at 180° along the pipeline axis to disperse the fluid and improve the uniformity (see Figure 3 ), and the grid 2 in a grid pattern has a plurality of square holes with a side length of 2 mm (see Figure 4 ).

[0026] Figure 5 The structure diagram of the pipeline part of this method is shown. In the figure, the sealing base 5 is provided with a sealing ring to prevent fluid leakage, and the sealing base 5 is connected to the incoming pipeline (INLET) and the outgoing pipeline (OUTLET) through flanges.

[0027] The dynamic component includes a blade rotating shaft 3, blades 4, an angle measurement module, a vibration process data acquisition module, a deep learning system, and a servo drive system.

[0028] Figure 6 The top view of the arrangement of the blade rotating shaft in the pipeline is shown. In the figure, the blade rotating shaft 3 is connected to the bearing provided on the sealing base 5, and the blade 4 is welded to the blade rotating shaft 3 and rotates with the rotation of the blade rotating shaft 3. Two layers of blade rotating shafts 3 are provided, with 7 in each layer, ensuring that the blade rotating shafts 3 in each layer are perpendicular to each other (the first transverse blade rotating shaft 3a, the second transverse blade rotating shaft 3b, the third transverse blade rotating shaft 3c, the fourth transverse blade rotating shaft 3d, the fifth transverse blade rotating shaft 3e, the sixth transverse blade rotating shaft 3f, the seventh transverse blade rotating shaft 3g, the first longitudinal blade rotating shaft 3h, the second longitudinal blade rotating shaft 3i, the third longitudinal blade rotating shaft 3j, the fourth longitudinal blade rotating shaft 3k, the fifth longitudinal blade rotating shaft 3l, the sixth longitudinal blade rotating shaft 3m, the seventh longitudinal blade rotating shaft 3n); the blade 4 has a taper, which is beneficial for guiding the flow.

[0029] The angle measurement module includes positioning sensors respectively arranged on different blade rotating shafts 3, which can measure the angle information of the blade rotation and transmit it to the deep learning system.

[0030] The vibration process data acquisition module includes a first pressure sensor, a first temperature sensor, and a first flow sensor provided on the incoming flow pipeline (INLET); a first displacement sensor, a first velocity sensor, a first acceleration sensor, a second pressure sensor, a second temperature sensor, and a second flow sensor provided on the outgoing flow pipeline (OUTLET), which can transmit pipeline vibration information and fluid process information to the deep learning system.

[0031] The servo drive system includes a first motor M1, a second motor (M2), a third motor (M3), a fourth motor M4, a fifth motor M5, a sixth motor M6, a seventh motor M7, an eighth motor M8, a ninth motor M9, a tenth motor M10, an eleventh motor M11, a twelfth motor M12, a thirteenth motor M13, and a fourteenth motor M14 connected to different blade rotation shafts 3. That is, each blade rotation shaft 3 is connected to a motor, and these motors respectively control the adjustment of the rotation angle of the corresponding blade rotation shaft 3, and the rotation angle range is controlled between 0° and 180°.

[0032] Under different working conditions, the deep learning system uses the output data of the angle measurement module and the vibration process data acquisition module as input signals, continuously collects vibration displacement, vibration velocity, and vibration acceleration data of the pipeline during operation to enhance the diversity of sample data. The deep learning system uses a BP neural network to establish an accurate non-linear mapping relationship between the rotation angle of the blade rotation shaft 3 and the vibration displacement, vibration velocity, and vibration acceleration data, obtains the function relationship curve between the rotation angle of the key structure and the vibration intensity under different working conditions, uses the control variable method and the trainlm learning method for neural network training, continuously optimizes the structure of the BP neural network model, topologically optimizes the key structure dimensions, finds the laws of the rotation angle of the key structure and the vibration intensity under different working conditions, enables the device to quickly feedback and find the optimal rotation angle under different working conditions and use it as the output signal, and then adjusts the rotation angle of the blade rotation shaft 3 through the feedback signal received by the servo motor, so that the device can adaptively adjust the key structure to the optimal angle under different working conditions and improve the working performance of the device.

[0033] The output data of the angle measurement module and the vibration process data acquisition module under different working conditions collected by the deep learning system are sample data of the available data set. First, correlation analysis is performed to eliminate abnormal data, and the remaining data are divided into training set, verification set and test set. Then, deep learning training is performed under the BP neural network using the control variable method and trainlm learning method. The parameters of the model are preliminarily determined through the training set and verification set, and the BP neural network model structure is determined. The generalization ability of the model is tested using the test set. The system can use the proxy model to feedback the optimal angle, guide the servo drive system to respond quickly according to the feedback signal, and automatically adjust the rotation of the blade rotating shaft 3 to achieve the best vibration reduction optimization effect, thereby improving the safety and reliability of the entire pipeline system.

[0034] The output display system is used to display the vibration and stress state of the pipeline in real time, as well as the parameter changes during the adjustment process. Through intuitive data display, operators can timely understand the operation status and adjustment effect of the device, which is convenient for monitoring and further optimization and adjustment, ensuring that the pipeline system operates in an efficient and stable state.

[0035] The above embodiments are only used to illustrate the present invention, and any equivalent transformations and improvements based on the technical solution of the present invention should not be excluded from the protection scope of the present invention.

Claims

1. A method for changing the phase uniformity of a flow field based on deep learning to reduce pipeline vibration, characterized in that, The pipeline is connected to the incoming flow pipeline and the outgoing flow pipeline through flanges respectively; the pipeline includes a static component, a dynamic component, and a sealing base (5); the dynamic component includes multiple groups of blade rotating shafts (3) and blades (4), each blade rotating shaft (3) is arranged on the sealing base (5), and the blade (4) is fixedly connected to the blade rotating shaft (3); An angle measurement module and a servo drive system are arranged on the blade rotating shaft (3), and a vibration data acquisition module is arranged on the incoming flow pipeline and the outgoing flow pipeline; the deep learning system is electrically connected to the angle measurement module, the vibration process data acquisition module, and the servo drive system; The angle measurement module includes positioning sensors respectively arranged on each blade rotating shaft (3); The vibration process data acquisition module includes a pressure sensor, a temperature sensor, and a flow sensor arranged on the incoming flow pipeline; a displacement sensor, a speed sensor, an acceleration sensor, a pressure sensor, a temperature sensor, and a flow sensor arranged on the outgoing flow pipeline; The servo drive system includes a motor group, and one motor is connected to one blade rotating shaft (3); The deep learning system uses the data output by the angle measurement module and the vibration process data acquisition module as a data set, and based on the deep learning principle, automatically learns and establishes a proxy model between the incoming flow condition, vibration displacement, speed acceleration, and angle; when the incoming flow condition changes, the optimal configuration angle is fed back by using the proxy model, and the servo drive system responds according to the feedback signal and automatically adjusts the rotation of different blade rotating shafts (3) to the optimal configuration angle.

2. A method for changing the phase uniformity of a flow field based on deep learning to reduce pipeline vibration according to claim 1, characterized in that: The static component includes a rectifying grille element, and the rectifying grille element includes a baffle grille (1) and a grid grille (2).

3. A method for changing the phase uniformity of a flow field based on deep learning to reduce pipeline vibration according to claim 2, characterized in that: The baffle grille (1) includes baffle plates arranged in a staggered manner, and two adjacent baffle grilles (1) are respectively installed at 180° along the pipeline axis; The grid grille (2) is evenly provided with square holes.

4. A method for changing the phase uniformity of a flow field based on deep learning to reduce pipeline vibration according to claim 1, characterized in that: The grille is connected to the sealing base (5) by welding; The blade rotating shaft (3) is fixed on the sealing base (5) through a bearing, and the motor drives the blade rotating shaft (3) to achieve the rotation function, and the blade (4) is welded and fixed on the blade rotating shaft (3).

5. A method for changing the phase uniformity of a flow field based on deep learning to reduce pipeline vibration according to claim 1, characterized in that: Two layers of blade rotating shafts (3) are arranged on the sealing base (5), with 7 roots in each layer, and the two layers of rotating shafts (3) are perpendicular to each other.

Images