Pipeline cavitation prediction model training method, prediction method, device and medium

Abstract

The invention provides a pipeline cavitation prediction model training method, a prediction method, a device and a medium, and relates to the technical field of computers, in particular to the technical fields of fluid mechanics, artificial intelligence, deep learning and the like. The pipeline cavitation prediction model training method comprises the following steps: acquiring a sample data set; wherein the sample data set comprises flow field body data in a sample pipeline; the flow field volume data in the sample pipeline comprises at least one of the following parameters: pipeline geometric parameters, pipeline inlet flow parameters and pipeline fluid physical property parameters; utilizing a to-be-trained pipeline cavitation prediction model to obtain predicted flow field physical parameters based on the flow field volume data in the sample pipeline; inputting the predicted physical parameters of the flow field into a target loss function to obtain a loss value; and based on the loss value, adjusting model parameters of the pipeline cavitation prediction model to obtain a final pipeline cavitation prediction model.

Description

TECHNICAL FIELD

[0001] The present disclosure relates to the technical field of computer, in particular to the technical field of fluid mechanics, artificial intelligence, deep learning and the like, and specifically relates to a pipeline cavitation prediction model training method, a pipeline cavitation phenomenon prediction method, a device and a medium. BACKGROUND

[0002] In a pipeline transportation system, cavitation refers to a phase change process in which cavitation bubbles are generated, developed and collapsed due to the fact that the local pressure of a liquid is lower than the saturated vapor at the same state and temperature, and a series of physical or chemical changes are triggered thereby. Cavitation phenomenon can reduce the reliability and service life of the pipeline transportation system. SUMMARY

[0003] The present disclosure provides a pipeline cavitation prediction model training method, a prediction method, a device and a medium.

[0004] According to an aspect of the present disclosure, a pipeline cavitation prediction model training method is provided, comprising: obtaining a sample data set; wherein the sample data set comprises: sample pipeline internal flow field volume data; the sample pipeline internal flow field volume data comprises at least one of the following: pipeline geometric parameters, pipeline inlet flow parameters and pipeline internal fluid physical property parameters; using a pipeline cavitation prediction model to be trained, based on the sample pipeline internal flow field volume data, to obtain predicted flow field physical parameters; the flow field physical parameters comprise a vapor volume fraction; inputting the predicted flow field physical parameters into a target loss function to obtain a loss value; wherein the target loss function is constructed based on the predicted flow field physical parameters and flow field physical parameters corresponding to the sample pipeline internal flow field volume data; based on the loss value, adjusting model parameters of the pipeline cavitation prediction model to obtain a final pipeline cavitation prediction model.

[0005] According to another aspect of the present disclosure, a pipeline cavitation phenomenon prediction method is provided, comprising: obtaining pipeline internal flow field volume data; the pipeline internal flow field volume data comprises at least one of the following: pipeline geometric parameters, pipeline inlet flow parameters and pipeline internal fluid physical property parameters; inputting the pipeline internal flow field volume data into a pipeline cavitation prediction model to obtain a prediction result output by the pipeline cavitation prediction model; wherein the prediction result comprises a vapor volume fraction distribution field, and the vapor volume fraction distribution field is used to represent a pipeline cavitation position and a pipeline cavitation intensity; wherein the pipeline cavitation prediction model is trained by using the method according to any one of the preceding aspects.

[0006] According to another aspect of the present disclosure, a training device of a pipeline cavitation prediction model is provided, comprising: a first obtaining unit configured to obtain a sample data set; wherein the sample data set comprises sample pipeline internal flow field volume data; the sample pipeline internal flow field volume data comprises at least one of pipeline geometric parameters, pipeline inlet flow parameters, and pipeline internal fluid physical property parameters; a first processing unit configured to obtain predicted flow field physical parameters based on the sample pipeline internal flow field volume data by using a pipeline cavitation prediction model to be trained; the flow field physical parameters comprise vapor volume fractions; a second processing unit configured to input the predicted flow field physical parameters into a target loss function to obtain a loss value; wherein the target loss function is constructed based on the predicted flow field physical parameters and flow field physical parameters corresponding to the sample pipeline internal flow field volume data; an adjusting unit configured to adjust model parameters of the pipeline cavitation prediction model based on the loss value to obtain a final pipeline cavitation prediction model.

[0007] According to another aspect of the present disclosure, a device for predicting pipeline cavitation phenomena is provided, comprising: a second obtaining unit configured to obtain pipeline internal flow field volume data; the pipeline internal flow field volume data comprises at least one of pipeline geometric parameters, pipeline inlet flow parameters, and pipeline internal fluid physical property parameters; a third processing unit configured to input the pipeline internal flow field volume data into a pipeline cavitation prediction model to obtain a prediction result output by the pipeline cavitation prediction model; wherein the prediction result comprises a vapor volume fraction distribution field, and the vapor volume fraction distribution field is used to represent pipeline cavitation positions and pipeline cavitation intensities; wherein the pipeline cavitation prediction model is trained by using the device as described above.

[0008] According to another aspect of the present disclosure, an electronic device is provided, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method of any of the embodiments of the present disclosure.

[0009] According to another aspect of the present disclosure, a non-transitory computer readable storage medium storing computer instructions is provided, wherein the computer instructions are used to cause the computer to perform the method according to any of the embodiments of the present disclosure.

[0010] According to another aspect of the present disclosure, there is provided a computer program product comprising a computer program which, when executed by a processor, implements the method according to any of the embodiments of the present disclosure.

[0011] The pipeline cavitation prediction model training method, the pipeline cavitation prediction method, the device and the medium provided by the present disclosure can realize real-time, high-precision and high-efficiency prediction when the pipeline cavitation model is used to predict the cavitation phenomenon in the pipeline transportation system, and can be applied to various pipeline inlet working conditions and various pipeline geometrical structures.

[0012] It should be understood that the content described in this part is not intended to identify key or important features of the embodiments of the present disclosure, nor to limit the scope of the present disclosure. Other features of the present disclosure will become apparent from the following description. BRIEF DESCRIPTION OF DRAWINGS

[0013] The accompanying drawings are used to better understand the present scheme and do not constitute a limitation on the present disclosure. Among them: Figure 1 is a schematic diagram of a simplified pipeline transportation system; Figure 2 is a flowchart of a pipeline cavitation prediction model training method according to an embodiment of the present disclosure; Figure 3 is a schematic diagram of a pipeline cavitation prediction model according to an embodiment of the present disclosure; Figure 4 is a flowchart of a pipeline cavitation prediction model training method according to another embodiment of the present disclosure; Figure 5 is a flowchart of a pipeline cavitation phenomenon prediction method according to an embodiment of the present disclosure; Figure 6 is a structural block diagram of a pipeline cavitation prediction model training device according to an embodiment of the present disclosure; Figure 7 is a structural block diagram of a device for predicting pipeline cavitation phenomenon according to an embodiment of the present disclosure; Figure 8 is a block diagram of an electronic device according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0014] Exemplary embodiments of the present disclosure are described below with reference to the accompanying drawings, which include various details of the embodiments of the present disclosure to assist in understanding, and should be considered as merely exemplary. Therefore, those of ordinary skill in the art should recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope of the present disclosure. Also, in order to be clear and concise, descriptions of well-known functions and structures are omitted in the following description.

[0015] In a pipeline transportation system, there are usually throttling elements and geometric mutation structures, such as pressure reducing valves, orifice plates, elbow sections, and reducing pipe sections. As shown in a simplified pipeline transportation system, Figure 1 including a straight pipe section with a length of L1, an elbow section with a curvature of K, a straight pipe section with a length of L2, a reducing pipe section with a reducing rate of dR / dx, and a straight pipe section with a length of L1 and a radius of R. When flowing through the region where the throttling element or the geometric mutation structure is located, the flow rate of the liquid increases sharply, and the local static pressure decreases significantly. When the pressure drops below the saturated vapor pressure of the liquid at the current temperature, phase change occurs inside the liquid and generates vapor bubbles. As shown in Figure 1 the inner region of the elliptical circle is a dangerous position where cavitation is likely to occur. When the vapor bubbles generated by the phase change inside the liquid (i.e., cavitation bubbles) enter the downstream high-pressure region with the main flow, they will quickly collapse (implode), generating a high-intensity shock wave, a local high temperature (up to several thousand Kelvin), and a high-speed microjet. This physical process can easily cause material erosion (cavitation erosion), broadband noise, and structural vibration, thereby reducing the reliability and service life of the pipeline transportation system. Therefore, effectively predicting the location of cavitation and controlling the cavitation phenomenon has always been a difficult problem in pipeline design.

[0016] In some related technologies, the pipeline geometry is usually optimized by means of experience and simplified theoretical models, numerical simulations, and small-scale experiments, such as the position of the throttling element, the curvature of the elbow, the change rate of the reducing pipe, and the like. The use of experience and theoretical models requires the estimation of dimensionless parameters such as cavitation number and semi-empirical formulas, which are difficult to accurately describe the influence of three-dimensional complex flow and real boundary conditions, especially in actual pipeline networks with multiple component coupling and asymmetric geometry. The prediction error is significant.

[0017] In another related technology, a scaled-down experimental system is used to predict the scheme to improve the accuracy of the prediction results. However, this scheme cannot directly observe the cavitation region and needs to use monitoring equipment to detect the region where cavitation may occur locally. In addition, this scheme has the problems of long experimental period, high cost, and difficulty in supporting rapid iterative optimization design of multiple parameters and multiple working conditions.

[0018] In yet another related technology, a multiphase flow model (such as the Schnerr-Sauer, Zwart-Gerber-Belamri cavitation model) is coupled with a turbulence solver (such as RANS / LES) to realize three-dimensional prediction of the cavitation flow field. However, this prediction scheme consumes a large amount of computing resources, such as several hours to several days for a single simulation. In the process of pipeline optimization design, only a small adjustment of the pipeline geometry parameters (such as the curvature radius of the elbow, the reducing cone angle, and the valve core opening) requires the generation of a new grid and the start of full flow field solving. There are a large number of repetitive and small-difference "approximate working condition" calculations, which leads to low design optimization efficiency.

[0019] It is evident that the relevant technologies struggle to balance prediction accuracy, computational efficiency, and real-time performance.

[0020] Based on this, this application provides a pipeline cavitation prediction model training method, prediction method, device and medium, which can realize high-precision and high-efficiency prediction of cavitation phenomena under different inlet working conditions (such as flow velocity, pressure and fluid quality) and complex pipeline geometric parameters (such as bend curvature radius, diameter ratio, valve core opening, etc.).

[0021] The following example illustrates the implementation process of the pipeline cavitation prediction model training method, prediction method, device, and medium provided in this embodiment.

[0022] Figure 2 This is a schematic flowchart of a pipeline cavitation prediction model training method according to an embodiment of the present disclosure.

[0023] Please refer to Figure 2 The pipeline cavitation prediction model training method provided in this embodiment includes the following steps S210 to S240: Step S210: Obtain the sample dataset; wherein, the sample dataset includes: flow field data inside the sample pipe; the flow field data inside the sample pipe includes at least one of the following: pipe geometric parameters, pipe inlet flow parameters, and fluid property parameters inside the pipe; Step S220: Using the pipeline cavitation prediction model to be trained, based on the flow field data inside the sample pipeline, the predicted flow field physical parameters are obtained; the flow field physical parameters include the vapor phase volume fraction. Step S230: Input the predicted flow field physical parameters into the target loss function to obtain the loss value; wherein, the target loss function is constructed based on the predicted flow field physical parameters and the flow field physical parameters corresponding to the flow field volume data in the sample pipe; Step S240: Based on the loss value, adjust the model parameters of the pipeline cavitation prediction model to obtain the final pipeline cavitation prediction model.

[0024] In step S210, a sample dataset can be obtained based on at least one of the following: flow field data inside the pipe obtained from simulation, flow field data inside the pipe obtained from experiment, and historical flow field data inside the pipe.

[0025] The pipeline cavitation prediction model can learn the nonlinear mapping relationship from the pipeline geometry and pipeline inlet conditions to the internal flow field (such as pressure field and velocity field) based on the acquired sample dataset, and realize end-to-end proxy simulation of the entire flow field.

[0026] Optionally, the pipe geometry parameters include at least one of the following: the length of the straight pipe section (length L), the curvature of the bend (curvature κ), the diameter change ratio of the reducing pipe section (diameter change ratio dR / dx), and the opening degree of the throttling element (opening degree = ). The curvature of the bend is continuously adjustable. Pipeline geometric parameters characterize the pipeline's geometric structure.

[0027] Optionally, the pipe inlet flow parameter includes at least one of the following: the average inflow velocity at the pipe inlet. Inflow pressure at the pipe inlet The flow parameters at the pipe inlet can characterize the operating conditions at the pipe inlet.

[0028] Optionally, the fluid properties within the pipeline include at least one of the following: saturated vapor pressure Pv, liquid viscosity μ, thermodynamic parameter Cp, and specific heat ratio γ. These fluid properties encompass the key thermodynamic and transport characteristics of vaporizable liquids and characterize the physicochemical properties of the fluid medium. By altering these fluid properties, rapid cavitation phenomena can be predicted for pipelines transporting different fluids, including common steam-water cavitation, oil-gas cavitation, and liquefied natural gas cavitation.

[0029] It is understood that the specific parameters included in the flow field data inside the pipeline are not limited to these. This embodiment is only an example, and the specific parameters can be selected according to actual needs.

[0030] In this embodiment, by introducing the above parameters, it is possible to adapt to any continuously adjustable pipe shape, improve geometric generalization ability, adapt to various working conditions, be compatible with various fluid media, meet the complex needs of pipeline transportation systems, and make the pipeline cavitation prediction model have a wide range of applicability.

[0031] In step S220, the cavitation prediction model of the pipeline to be trained is used to construct a three-dimensional geometric representation of the flow field inside the pipeline based on the parametric modeling method, and the key variables of the entire flow field, namely the predicted physical parameters of the flow field, are output through forward propagation calculation.

[0032] In some examples, the predicted flow field physical parameters include pressure p, velocity u, and mixing mass. Vapor phase volume fraction .in, It can characterize the cavitation intensity at a certain point. , Indicates pure liquid phase, This indicates a pure gas phase. That is, It can characterize whether cavitation has occurred at a local location and the intensity of cavitation. The distribution field can be used to characterize the cavitation location and corresponding cavitation intensity of a pipeline transportation system.

[0033] It is understandable that during the model training phase, the predicted flow field physical parameters include: mixed mass. Pressure p, velocity u, vapor volume fraction In the model application phase, the predicted flow field physical parameters include vapor volume fraction. .

[0034] Before step S230, a target loss function is constructed based on the predicted flow field physical parameters and the flow field physical parameters corresponding to the flow field volume data in the sample pipe. In step S230, the predicted flow field physical parameters obtained in step S220 are input into the target loss function to obtain the loss value.

[0035] Please refer to Figure 3 In step S240, for example, if the loss value is greater than a preset convergence threshold, backpropagation is performed to adjust the model parameters in the pipeline cavitation prediction model. For example, an adaptive gradient optimizer can be used to update the model parameters in the pipeline cavitation prediction model. When the loss value is less than or equal to the preset convergence threshold, the final pipeline cavitation prediction model is obtained. The set convergence threshold can be 0.010 or 0.015, etc., and can be set according to actual needs. For different pipeline geometries or pipeline inlet conditions, the set convergence threshold can be the same or different.

[0036] The training method provided in this embodiment uses a sample dataset including pipeline geometric parameters, pipeline inlet flow parameters, and fluid property parameters within the pipeline when training the pipeline cavitation prediction model. This dataset enables the pipeline cavitation prediction model to generate predicted flow field physical parameters. Based on the predicted flow field physical parameters and the target loss function, a loss value is obtained. Then, based on the loss value, the model parameters of the pipeline cavitation prediction model are adjusted until the final pipeline cavitation prediction model is obtained. In this way, when using the pipeline cavitation model to predict cavitation phenomena in pipeline transportation systems, real-time, high-precision, and high-efficiency predictions can be achieved, and it is applicable to various pipeline inlet conditions and various pipeline geometries.

[0037] In some embodiments, the pipe cavitation prediction model may include a parameter input module and an encoding / decoding frequency domain neural operator network module. The parameter input module constructs a three-dimensional geometric representation of the flow field within the pipe based on a parametric modeling method and generates unstructured mesh nodes through mesh discretization. The three-dimensional geometric representation is a digital geometric model used to describe the physical shape and topology of the flow field within the pipe.

[0038] The encoding / decoding frequency domain neural operator network module is used to construct a high-fidelity mapping from multi-source inputs (pipe geometry parameters, pipe inlet flow parameters, and fluid properties within the pipe) to the physical quantities of the entire flow field. Please refer to [link / reference]. Figure 3 and Figure 4For example, the encoding / decoding frequency domain neural operator network module includes: an encoder, a frequency domain neural operator, and a decoder. The encoder is used to map structured field data to latent space, the frequency domain neural operator is used to perform fast frequency domain learning within the latent space, and the decoder is used to map the latent space data back to the original geometry. The implementation process of the encoding / decoding frequency domain neural operator network module is illustrated below.

[0039] In some embodiments, step S220, using a pipe cavitation prediction model, obtains the predicted flow field physical parameters based on the flow field data within the sample pipe, including the following steps a10 to a30: Step a10: Use the encoder of the pipeline cavitation prediction model to encode the unstructured grid node data corresponding to the flow field data in the sample pipeline, and generate structured field data based on the encoded data. Step a20: Use the frequency domain neural operator of the pipeline cavitation prediction model to perform Fourier transform on the structured field data, and process the frequency domain data after Fourier transform to obtain a low-dimensional latent map. Step a30: Use the decoder of the pipeline cavitation prediction model to decode the low-dimensional latent map and obtain the predicted flow field physical parameters.

[0040] In step a10, the encoder encodes the unstructured grid node data using a graph neural network to generate unstructured node latent features; wherein, the graph neural network may include at least one of the following: graph convolutional network, graph attention network. Encoding unstructured grid node data into unstructured node latent features using a graph neural network (GNN) can preserve the physical details of local topological relationships and abrupt flow field changes, thereby improving the geometric generalization ability of the pipeline cavitation prediction model.

[0041] Subsequently, differentiable interpolation methods, such as inverse distance weighting, radial basis function interpolation, moving least squares, or natural neighbor interpolation, are used to map the latent features of unstructured nodes onto a regular Cartesian grid, generating structured field data. This satisfies the input regularity requirements of subsequent frequency-domain neural operators, improves frequency-domain compatibility, and helps ensure high fidelity of flow field features. The grid resolution of the structured field data can be set to be equal to the maximum Fourier mode number of the frequency-domain neural operator.

[0042] This embodiment constructs a fully differentiable preprocessing chain by combining graph neural network encoding and differentiable interpolation, allowing gradients to propagate backward from the frequency domain neural operator to the original unstructured grid nodes. This facilitates the modeling of the direct impact of pipeline geometry parameters and pipeline inlet flow parameters on cavitation prediction results, thereby enabling the optimization design of pipeline geometry parameters and pipeline inlet flow parameters.

[0043] In step a20, the frequency domain neural operator performs a Fourier transform on the structured field data, converting the structured field data from the learning space to the frequency domain.

[0044] Next, the frequency-domain neural operator models the long-range dependence and multi-scale coupling effects of the flow field in the frequency domain. For example, step a20, processing the Fourier-transformed frequency-domain data to obtain a low-dimensional latent map, includes steps a21 to a23: Step a21: Perform a learnable linear transformation and spectral truncation on the frequency domain data after Fourier transform to obtain the target frequency domain data within the target spectrum range. Step a22: Apply differential weighting to the target frequency domain data using a channel attention mechanism; Step a23: Perform an inverse Fourier transform on the differentiated weighted target frequency domain data to obtain a low-dimensional latent map.

[0045] In step a21, spectral truncation can be performed first, followed by a learnable linear transformation, to reduce the amount of data processing. In other examples, a learnable linear transformation can also be performed first, followed by spectral truncation, to reduce errors.

[0046] The target frequency range can be set according to actual needs. The target frequency domain data includes low-frequency and high-frequency components. The low-frequency components correspond to the components in the Fourier modes with wavenumbers below a preset wavenumber threshold, representing slowly changing, long-wavelength (large-scale) characteristics in the flow field, and are used to model the evolution characteristics of mainstream pressure gradients (such as pipe pressure drop) and large-scale backflow (such as backflow zones at bends). The high-frequency components correspond to the components in the Fourier modes with wavenumbers above a preset wavenumber threshold, representing rapidly changing, short-wavelength (small-scale) characteristics in the flow field, and are used to capture small-scale vortex structures (such as boundary layer vortices) and cavitation initiation disturbances (such as microbubble nucleation) in the flow field. The preset wavenumber threshold can be determined based on the attenuation characteristics of the energy distribution in the spectrum or the physical characteristics of the problem domain.

[0047] In step a22, a channel attention mechanism is used to differentially weight the low-frequency and high-frequency components. This mechanism utilizes a learnable parameter mapping network and dynamically generates weights based on pipe geometry parameters and inlet flow parameters. That is, the weights for low-frequency and high-frequency components can be adaptively adjusted; their weights can differ depending on the pipe inlet conditions and pipe geometry. For example, in scenarios with multiple bends, if a greater focus is on modeling large-scale recirculation zones, the weight of low-frequency components is increased; if a greater focus is on local flow separation details, the weight of high-frequency components is increased. Furthermore, the weight of low-frequency components can be correlated with pressure gradients to incorporate physical characteristics and ensure the accuracy of the pipe cavitation prediction model.

[0048] In this embodiment, by applying a learnable linear transformation to the frequency domain data and by using spectral truncation and channel attention mechanisms to differentiate the weighting of high-frequency and low-frequency components, the ability of the pipeline cavitation prediction model to capture the non-uniformity of the flow field in geometrically abrupt regions (such as elbows and orifice plates) and its generalization performance can be significantly improved, thereby improving the accuracy of the prediction results.

[0049] In step a23, the weighted low-frequency components and high-frequency components are merged, and the merged data is subjected to inverse Fourier transform to convert the data to the spatial domain, thereby obtaining a low-dimensional latent map.

[0050] In step a30, the decoder can project the low-dimensional latent map back to the original unstructured grid nodes through a graph neural network to obtain the predicted flow field physical parameters.

[0051] In this embodiment, the above settings enable the transformation of the internal flow field grid into the internal flow field physical parameters of any pipeline transportation system. This facilitates the rapid prediction of the location and intensity of cavitation that may occur in different pipeline transportation systems, allowing for timely optimization of the pipeline transportation system.

[0052] In some embodiments, a target loss function is constructed based on the predicted flow field physical parameters and the flow field physical parameters corresponding to the flow field volume data in the sample pipe, including the following steps b10 to b30: Step b10: Based on the predicted flow field physical parameters and the flow field physical parameters corresponding to the flow field volume data in the sample pipe, the first loss term is obtained; wherein, the flow field physical parameters corresponding to the flow field volume data in the sample pipe are obtained based on the CFD numerical simulation results; Step b20 derives the second loss term based on the predicted flow field physical parameters and the mass transport source term; Step b30 performs a weighted summation of the first loss term and the second loss term to obtain the target loss function.

[0053] In step b10, the flow field physical parameters corresponding to the flow field volume data in the sample pipe can be obtained based on the CFD numerical simulation results. These simulation results can be generated by the open-source solver OpenFOAM. The predicted flow field physical parameters can be compared point by point with the flow field physical parameters in the corresponding CFD numerical simulation results to determine the error; the average of the squared errors is then taken to obtain the first loss term.

[0054] It can be understood that when the physical parameters of the flow field include multiple parameters, such as the mixed mass, then... Pressure p, velocity u, vapor volume fraction In this case, the first loss term can be a mixture of mass. Pressure p, velocity u, vapor volume fraction The mean square error weighted sum; the weight coefficients of each parameter can be dynamically adjusted according to their dimensional characteristics or the gradient magnitude during the training process.

[0055] In this embodiment, by introducing a data-driven first loss term into the target loss function, the output of the pipeline cavitation prediction model can be made to approximate the real flow field distribution, thus ensuring the accuracy of the basic prediction.

[0056] In step b20, the divergence terms of multiple sampling points can be determined based on the predicted flow field physical parameters; the residuals of the divergence terms of multiple sampling points and the corresponding mass transport source terms can be determined; and the average of the squares of the multiple residuals can be used to obtain the second loss term.

[0057] To accelerate the convergence of the pipeline cavitation prediction model and ensure its predicted solutions satisfy physical consistency, avoiding getting trapped in non-physical solution spaces, a physical constraint term based on the Schnerr–Sauer cavitation model, also known as the second loss term (or regularization term), can be embedded in the objective loss function. This pipeline cavitation prediction model assumes that the liquid and its vaporized gas satisfy the following mass conservation equation involving mass transfer: (Formula 1) in, , Indicates the mixed density, indicating gas phase density, Indicates the density of the liquid phase; The vapor phase volume fraction is represented by u; velocity is represented by u. This indicates a mass transport source item.

[0058] For quasi-steady-state problems, the effect of time-varying derivative terms can be neglected. The Schnerr-Sauer cavitation model assumes the existence of cavitation in the flow field. A radius is The evolution of individual bubbles can be approximated by the Rayleigh-Plesset equation, ultimately leading to the derivation of the mass transport source term. The expression is: (Formula 2) in, Indicates a mass transport source term; Indicates the evaporation process; Indicates the condensation process; Indicates the density of the gas phase; Indicates the volume fraction of the vapor phase; This represents the fluid pressure within the pipe. Indicates saturated vapor pressure; Indicates the density of the liquid phase; The radius of the air bubble present in the flow field inside the pipe can be obtained through empirical calculation or is a set constant.

[0059] In this embodiment, cavitation dynamics and bubble radius are embedded in the second loss term. It can accurately reflect the initiation and collapse processes of cavitation, and can automatically switch between evaporation and condensation modes according to local pressure conditions to adapt to transient changes in the flow field. Furthermore... By binding the prediction results with the pipeline cavitation prediction model, a closed-loop feedback optimization can be formed.

[0060] The predicted flow field physical parameters (ρ) obtained in step S220 m , p,u, α v Substituting into the second formula for quasi-steady state, the second loss term is calculated: (Formula 3) in, Indicates the second loss item; Indicates the total number of sampling points; This indicates that the differential operator defined by Equation 2 applies to the parameters of the neural network. The residual function under the given conditions can be simply referred to as The specific form is as follows: ; This represents the coordinates of the i-th sampling point.

[0061] The first loss term of the data-driven part The second loss term of the mechanism-driven part By performing a weighted summation, we can obtain the target loss function: (Formula 4) in, Represent the objective function; This represents the weighting coefficient of the first loss term; This represents the weighting coefficient of the second loss term.

[0062] Optionally, this can be achieved by using an adaptive weight training tool. and Adaptive adjustments can be made to further accelerate model convergence. For example, adjustments can be made based on... and Dynamic adjustment of gradient magnitude and For example, it can be done in the initial stage... Set to 0, and adjust according to training rounds. To increase.

[0063] In this embodiment, by embedding physical information constraints based on the Schnerr–Sauer cavitation model into the target loss function, and adding the residual of the gas phase mass transport equation as the second loss term to the target loss function, it not only ensures that the prediction results strictly satisfy the physical mechanism of cavitation phase transition and effectively avoids the occurrence of non-physical solutions, but also significantly improves the generalization ability of the model under extrapolation conditions and accelerates the convergence process.

[0064] Figure 5 This is a schematic flowchart of a pipeline cavitation prediction method according to an embodiment of the present disclosure.

[0065] Please refer to Figure 5 This embodiment also provides a method for predicting pipeline cavitation, including steps S510 to S520: Step S510: Obtain flow field data inside the pipe; the flow field data inside the pipe includes at least one of the following: pipe geometric parameters, pipe inlet flow parameters, and fluid properties inside the pipe; Step S520: Input the flow field data inside the pipeline into the pipeline cavitation prediction model to obtain the prediction results output by the pipeline cavitation prediction model; wherein, the prediction results include the vapor phase volume fraction distribution field, which is used to characterize the cavitation location and cavitation intensity of the pipeline; the pipeline cavitation prediction model is trained using the method in any of the aforementioned embodiments.

[0066] The prediction method provided in this embodiment can predict pipeline cavitation phenomena in real time, with high accuracy and efficiency, and is applicable to various pipeline inlet conditions and various pipeline geometries.

[0067] The similarities between this embodiment and the previous embodiments will not be repeated here.

[0068] In some embodiments, the pipeline cavitation prediction method further includes: optimizing the pipeline geometric parameters and / or the pipeline inlet flow parameters until the pipeline cavitation risk area index reaches its minimum value or the number of iterations reaches its maximum value.

[0069] During the optimization process, global optimization algorithms such as Genetic Algorithm (GA), Particle Swarm Optimization (PSO), or Bayesian Optimization (BO) are selected. For example, using Bayesian optimization, a pipeline cavitation prediction model can be embedded in the Bayesian optimization system. Pipeline geometric parameters and inlet flow parameters are used as optimization variables to construct a high-dimensional optimization space, and Bayesian optimization drives the iterative search. The pipeline cavitation prediction model acts as an efficient objective function evaluator, replacing traditional time-consuming CFD calculations. Iteration stops when the cavitation risk area index determined by the prediction results output by the pipeline cavitation prediction model reaches its minimum value, or when the number of iterations reaches its maximum value.

[0070] Optionally, determining pipeline cavitation risk zone indicators includes: determining the volume of the pipeline cavitation risk zone based on the pipeline cavitation location where the pipeline cavitation intensity reaches a set intensity threshold, and determining the pipeline cavitation risk zone indicator based on the volume of the pipeline cavitation risk zone. The set intensity threshold can be set according to actual needs. In this way, by quantifying the cavitation risk zone, a clear target can be provided for parameter optimization.

[0071] In this embodiment, the optimized pipeline geometry parameters and pipeline inlet flow parameters can significantly suppress or even completely eliminate cavitation clouds with high collapse risk, thereby improving the reliability and service life of the pipeline transportation system.

[0072] Compared to simplified prediction methods based on experience in related technologies, the prediction method in this embodiment adopts a pipeline cavitation prediction model driven by a hybrid approach of data and mechanism. The training data for the pipeline cavitation prediction model comes from high-fidelity CFD data, and an additional physical constraint loss term for the cavitation model is introduced. This method is more accurate than simplified prediction methods that rely on simplified derivations and engineering empirical formulas.

[0073] Compared to physical verification methods based on scaled-down experiments in related technologies, the prediction method in this embodiment employs a pipeline cavitation prediction model driven by a hybrid approach of data and mechanism. This model ensures computational accuracy while reflecting the overall flow properties of the pipeline transportation system without relying on measurement equipment. Furthermore, it enables rapid optimization of the pipeline transportation system without the need to repeatedly build experimental systems.

[0074] Compared to the numerical simulation methods based on traditional CFD in related technologies, the prediction method in this embodiment involves repeated calculations and wastes a lot of computing resources when using traditional CFD methods for iterative design. In contrast, the pipeline cavitation prediction model driven by a combination of data and mechanism in this embodiment can quickly predict the cavitation situation of the flow field during the design stage, thereby supporting iterative optimization design.

[0075] It should be noted that the prediction method provided in this embodiment can be applied to at least one of the following scenarios: oil and gas pipeline systems, urban water supply and drainage networks, liquefied natural gas (LNG) cryogenic pipelines, chemical process industry pipelines, hydraulic transmission and control systems, ship propulsion system pipelines, aerospace fuel pipelines, high-speed fluid valves and throttling devices, hydropower water intake and tailrace pipelines, geothermal energy transmission pipelines, and internal flow channel design of various pumps, compressors and turbines.

[0076] Figure 6 This is a structural block diagram of a training apparatus for a pipeline cavitation prediction model according to an embodiment of the present disclosure.

[0077] Please refer to Figure 6 According to another aspect of this disclosure, a training apparatus for a pipeline cavitation prediction model is provided, comprising: The first acquisition unit 610 is used to acquire a sample dataset; wherein, the sample dataset includes: flow field data inside the sample pipe; the flow field data inside the sample pipe includes at least one of the following: pipe geometric parameters, pipe inlet flow parameters, and fluid property parameters inside the pipe; The first processing unit 620 is used to obtain predicted flow field physical parameters based on the flow field volume data inside the sample pipe using the pipeline cavitation prediction model to be trained; the flow field physical parameters include vapor phase volume fraction. The second processing unit 630 is used to input the predicted flow field physical parameters into the target loss function to obtain the loss value; wherein the target loss function is constructed based on the predicted flow field physical parameters and the flow field physical parameters corresponding to the flow field volume data in the sample pipe; The adjustment unit 640 is used to adjust the model parameters of the pipeline cavitation prediction model based on the loss value to obtain the final pipeline cavitation prediction model.

[0078] In some embodiments, the first processing unit 620 is specifically used for: The encoder of the pipeline cavitation prediction model is used to encode the unstructured grid node data corresponding to the flow field data in the sample pipeline, and structured field data is generated based on the encoded data. The structured field data is subjected to Fourier transform using the frequency domain neural operator of the pipeline cavitation prediction model, and the frequency domain data after Fourier transform is processed to obtain a low-dimensional latent map. The low-dimensional latent map is decoded using the decoder of the pipeline cavitation prediction model to obtain the predicted flow field physical parameters.

[0079] In some embodiments, the first processing unit 620 is specifically used for: A learnable linear transformation and spectral truncation are performed on the frequency domain data after Fourier transform to obtain the target frequency domain data within the target frequency range. The low-frequency and high-frequency components in the target frequency domain data are differentially weighted using a channel attention mechanism; The low-dimensional latent map is obtained by performing an inverse Fourier transform on the differentially weighted target frequency domain data.

[0080] In some embodiments, the first processing unit 620 is specifically used for: The unstructured grid node data corresponding to the predicted flow field data in the pipeline is encoded using a graph neural network to generate unstructured node latent features. The unstructured node latent features are mapped to a regular Cartesian grid using a differentiable interpolation method to generate structured field data.

[0081] In some embodiments, the training apparatus for the pipeline cavitation prediction model further includes a construction unit for: Based on the predicted flow field physical parameters and the flow field physical parameters corresponding to the flow field volume data in the sample pipe, a first loss term is obtained; wherein, the flow field physical parameters corresponding to the flow field volume data in the sample pipe are obtained based on CFD numerical simulation results; Based on the predicted flow field physical parameters and mass transport source term, the second loss term is obtained; The first loss term and the second loss term are weighted and summed to obtain the target loss function.

[0082] In some embodiments, the building unit is specifically used for: The predicted flow field physical parameters are compared point by point with the flow field physical parameters corresponding to the flow field volume data in the sample pipe, and the error is determined. The average of the squares of the obtained errors is used to obtain the first loss term.

[0083] In some embodiments, the building unit is specifically used for: Based on the predicted flow field physical parameters, the divergence terms of multiple sampling points are determined; Determine the residual between the divergence term and the corresponding mass transport source term at the sampling point; The second loss term is obtained by averaging the squares of the multiple residuals.

[0084] In some embodiments, the training apparatus for the pipeline cavitation prediction model further includes: The mass transport source term under quasi-steady state is determined using the following formula: ; in, Indicates a mass transport source term; Indicates the evaporation process; Indicates the condensation process; Indicates the density of the gas phase; Indicates the volume fraction of the vapor phase; This represents the fluid pressure within the pipe. Indicates saturated vapor pressure; Indicates the density of the liquid phase; This represents the radius of the air bubble present in the flow field inside the pipe.

[0085] In some embodiments, the pipe geometry parameters include at least one of the following: the length of the straight pipe section, the curvature of the bend, the diameter change rate of the variable diameter pipe section, and the opening degree of the throttling element; The flow parameters at the pipe inlet include at least one of the following: the average inflow velocity at the pipe inlet and the inflow pressure at the pipe inlet; The fluid properties in the pipeline include at least one of the following: saturated vapor pressure, liquid viscosity, thermodynamic parameters, and specific heat ratio.

[0086] The specific functions and examples of each module and submodule of the apparatus in this disclosure can be found in the relevant descriptions of the corresponding steps in the above method embodiments, and will not be repeated here.

[0087] Figure 7 This is a structural block diagram of an apparatus for predicting pipeline cavitation phenomena according to an embodiment of the present disclosure.

[0088] Please refer to Figure 7 According to another aspect of this disclosure, an apparatus for predicting cavitation phenomena in pipelines is provided, comprising: The second acquisition unit 710 is used to acquire flow field data inside the pipeline; the flow field data inside the pipeline includes at least one of the following: pipeline geometric parameters, pipeline inlet flow parameters, and fluid property parameters inside the pipeline. The third processing unit 720 is used to input the flow field data inside the pipeline into the pipeline cavitation prediction model to obtain the prediction results output by the pipeline cavitation prediction model; wherein, the prediction results include a vapor phase volume fraction distribution field, which is used to characterize the cavitation location and cavitation intensity of the pipeline; wherein, the pipeline cavitation prediction model is trained using the device described above.

[0089] In some embodiments, the apparatus for predicting pipeline cavitation further includes an optimization unit for: The pipeline geometric parameters and / or pipeline inlet flow parameters are optimized until the pipeline cavitation risk area index reaches its minimum value or the number of iterations reaches its maximum value.

[0090] In some embodiments, the apparatus for predicting pipeline cavitation further includes a determining unit for: The volume of the pipeline cavitation risk zone is determined based on the pipeline cavitation location where the pipeline cavitation intensity reaches a set intensity threshold, and the pipeline cavitation risk zone index is determined based on the volume of the pipeline cavitation risk zone.

[0091] The specific functions and examples of each module and submodule of the apparatus in this disclosure can be found in the relevant descriptions of the corresponding steps in the above method embodiments, and will not be repeated here.

[0092] According to embodiments of this disclosure, this disclosure also provides an electronic device, a readable storage medium, and a computer program product.

[0093] Figure 8 A schematic block diagram of an example electronic device 800 that can be used to implement embodiments of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0094] like Figure 8 As shown, device 800 includes a computing unit 801, which can perform various appropriate actions and processes based on a computer program stored in read-only memory (ROM) 802 or a computer program loaded from storage unit 808 into random access memory (RAM) 803. RAM 803 may also store various programs and data required for the operation of device 800. The computing unit 801, ROM 802, and RAM 803 are interconnected via bus 804. Input / output (I / O) interface 805 is also connected to bus 804.

[0095] Multiple components in device 800 are connected to I / O interface 805, including: input unit 806, such as keyboard, mouse, etc.; output unit 807, such as various types of monitors, speakers, etc.; storage unit 808, such as disk, optical disk, etc.; and communication unit 809, such as network card, modem, wireless transceiver, etc. Communication unit 809 allows device 800 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0096] The computing unit 801 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 801 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 801 performs the various methods and processes described above. For example, in some embodiments, the model training method or prediction method can be implemented as a computer software program tangibly contained in a machine-readable medium, such as storage unit 808. In some embodiments, part or all of the computer program can be loaded and / or installed on device 800 via ROM 802 and / or communication unit 809. When the computer program is loaded into RAM 803 and executed by the computing unit 801, one or more steps of the model training method or prediction method described above can be performed. Alternatively, in other embodiments, the computing unit 801 can be configured to perform the model training method or prediction method by any other suitable means (e.g., by means of firmware).

[0097] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0098] The program code used to implement the methods of this disclosure may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing apparatus, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0099] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0100] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0101] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as a data server), or computing systems that include middleware components (e.g., an application server), or computing systems that include frontend components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., a communication network). Examples of communication networks include local area networks (LANs), wide area networks (WANs), and the Internet.

[0102] Computer systems can include clients and servers. Clients and servers are generally located far apart and typically interact via communication networks. Client-server relationships are created by computer programs running on the respective computers and having a client-server relationship with each other. Servers can be cloud servers, servers in distributed systems, or servers incorporating blockchain technology.

[0103] It should be noted that the collection, storage, use, processing, transmission, provision, and disclosure of data involved in the technical solution disclosed herein all comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0104] It should be noted that all data involved in this disclosure (including but not limited to data used for analysis, data stored, data displayed, etc.) are authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with the relevant laws, regulations and standards of the relevant countries and regions.

[0105] It is worth noting that in the embodiments disclosed herein, certain software, components, models, and other existing solutions in the industry may be mentioned. These should be considered as exemplary, and their purpose is only to illustrate the feasibility of implementing the technical solutions disclosed herein. However, this does not mean that the applicant has used or necessarily used such solutions.

[0106] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this disclosure can be achieved, and this is not limited herein.

[0107] The specific embodiments described above do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the principles of this disclosure should be included within the scope of protection of this disclosure.

Claims

1. A method for training a pipeline cavitation prediction model, comprising: Obtain a sample dataset; wherein the sample dataset includes: flow field data inside the sample pipe; the flow field data inside the sample pipe includes at least one of the following: pipe geometric parameters, pipe inlet flow parameters, and fluid properties inside the pipe; Using the pipeline cavitation prediction model to be trained, the predicted flow field physical parameters are obtained based on the flow field data inside the sample pipeline; the flow field physical parameters include the vapor phase volume fraction. The predicted flow field physical parameters are input into the target loss function to obtain the loss value; wherein, the target loss function is constructed based on the predicted flow field physical parameters and the flow field physical parameters corresponding to the flow field volume data in the sample pipe; Based on the loss value, the model parameters of the pipeline cavitation prediction model are adjusted to obtain the final pipeline cavitation prediction model.

2. The method according to claim 1, wherein, The pipeline cavitation prediction model, based on the flow field data within the sample pipeline, yields the predicted flow field physical parameters, including: The encoder of the pipeline cavitation prediction model is used to encode the unstructured grid node data corresponding to the flow field data in the sample pipeline, and structured field data is generated based on the encoded data. The structured field data is subjected to Fourier transform using the frequency domain neural operator of the pipeline cavitation prediction model, and the frequency domain data after Fourier transform is processed to obtain a low-dimensional latent map. The low-dimensional latent map is decoded using the decoder of the pipeline cavitation prediction model to obtain the predicted flow field physical parameters.

3. The method according to claim 2, wherein, The process of processing the frequency domain data after Fourier transform to obtain a low-dimensional latent map includes: A learnable linear transformation and spectral truncation are performed on the frequency domain data after Fourier transform to obtain the target frequency domain data within the target frequency range. The low-frequency and high-frequency components in the target frequency domain data are differentially weighted using a channel attention mechanism; The low-dimensional latent map is obtained by performing an inverse Fourier transform on the differentially weighted target frequency domain data.

4. The method according to claim 2, wherein, The process of encoding the predicted flow field data within the pipeline and generating structured field data based on the encoded data includes: The unstructured grid node data corresponding to the predicted flow field data in the pipeline is encoded using a graph neural network to generate unstructured node latent features. The unstructured node latent features are mapped to a regular Cartesian grid using a differentiable interpolation method to generate structured field data.

5. The method according to claim 1, wherein, Based on the predicted flow field physical parameters and the flow field physical parameters corresponding to the flow field volume data in the sample pipe, a target loss function is constructed, including: Based on the predicted flow field physical parameters and the flow field physical parameters corresponding to the flow field volume data in the sample pipe, a first loss term is obtained; wherein, the flow field physical parameters corresponding to the flow field volume data in the sample pipe are obtained based on CFD numerical simulation results; Based on the predicted flow field physical parameters and mass transport source term, the second loss term is obtained; The first loss term and the second loss term are weighted and summed to obtain the target loss function.

6. The method according to claim 5, wherein, The second loss term, derived based on the predicted flow field physical parameters and mass transport source term, includes: Based on the predicted flow field physical parameters, the divergence terms of multiple sampling points are determined; Determine the residual between the divergence term and the corresponding mass transport source term at the sampling point; The second loss term is obtained by averaging the squares of the multiple residuals.

7. The method according to claim 6, wherein, Also includes: The mass transport source term under quasi-steady state is determined using the following formula: ； in, Indicates a mass transport source term; Indicates the evaporation process; Indicates the condensation process; Indicates the density of the gas phase; Indicates the volume fraction of the vapor phase; This represents the fluid pressure within the pipe. Indicates saturated vapor pressure; Indicates the density of the liquid phase; This represents the radius of the air bubble present in the flow field inside the pipe.

8. The method according to claim 5, wherein, The first loss term is obtained based on the predicted flow field physical parameters and the flow field volume data corresponding to the sample pipe. This loss term includes: The predicted flow field physical parameters are compared point by point with the flow field physical parameters corresponding to the flow field volume data in the sample pipe, and the error is determined. The average of the squares of the obtained errors is used to obtain the first loss term.

9. The method according to claim 1, wherein, The pipeline geometry parameters include at least one of the following: the length of the straight pipe section, the curvature of the bend section, the diameter change rate of the variable diameter pipe section, and the opening degree of the throttling element. The flow parameters at the pipe inlet include at least one of the following: the average inflow velocity at the pipe inlet and the inflow pressure at the pipe inlet; The fluid properties in the pipeline include at least one of the following: saturated vapor pressure, liquid viscosity, thermodynamic parameters, and specific heat ratio.

10. A method for predicting pipeline cavitation, comprising: Acquire flow field data inside the pipeline; the flow field data inside the pipeline includes at least one of the following: pipeline geometric parameters, pipeline inlet flow parameters, and fluid properties inside the pipeline. The flow field data inside the pipeline is input into the pipeline cavitation prediction model to obtain the prediction results output by the pipeline cavitation prediction model; wherein, the prediction results include a vapor phase volume fraction distribution field, which is used to characterize the cavitation location and cavitation intensity of the pipeline. The pipeline cavitation prediction model is trained using the method described in any one of claims 1 to 9.

11. The method according to claim 10, wherein, Also includes: The pipeline geometric parameters and / or pipeline inlet flow parameters are optimized until the pipeline cavitation risk area index reaches its minimum value or the number of iterations reaches its maximum value.

12. The method according to claim 11, wherein, Also includes: The volume of the pipeline cavitation risk zone is determined based on the pipeline cavitation location where the pipeline cavitation intensity reaches a set intensity threshold, and the pipeline cavitation risk zone index is determined based on the volume of the pipeline cavitation risk zone.

13. A training device for a pipeline cavitation prediction model, comprising: The first acquisition unit is used to acquire a sample dataset; wherein, the sample dataset includes: flow field data inside the sample pipe; the flow field data inside the sample pipe includes at least one of the following: pipe geometric parameters, pipe inlet flow parameters, and fluid physical property parameters inside the pipe; The first processing unit is used to obtain predicted flow field physical parameters based on the flow field volume data inside the sample pipe using the pipeline cavitation prediction model to be trained; the flow field physical parameters include vapor phase volume fraction. The second processing unit is used to input the predicted flow field physical parameters into the target loss function to obtain the loss value; wherein the target loss function is constructed based on the predicted flow field physical parameters and the flow field physical parameters corresponding to the flow field volume data in the sample pipe; An adjustment unit is used to adjust the model parameters of the pipeline cavitation prediction model based on the loss value to obtain the final pipeline cavitation prediction model.

14. An apparatus for predicting cavitation phenomena in pipelines, comprising: The second acquisition unit is used to acquire flow field data inside the pipeline; the flow field data inside the pipeline includes at least one of the following: pipeline geometric parameters, pipeline inlet flow parameters, and fluid physical property parameters inside the pipeline. The third processing unit is used to input the flow field data inside the pipeline into the pipeline cavitation prediction model to obtain the prediction result output by the pipeline cavitation prediction model; wherein, the prediction result includes a vapor phase volume fraction distribution field, which is used to characterize the cavitation location and cavitation intensity of the pipeline; wherein, the pipeline cavitation prediction model is trained using the device as described in claim 13.

15. An electronic device comprising: At least one processor; as well as A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1 to 12.

16. A non-transitory computer-readable storage medium storing computer instructions, wherein, The computer instructions are used to cause the computer to perform the method according to any one of claims 1 to 12.

17. A computer program product comprising a computer program that, when executed by a processor, implements the method according to any one of claims 1 to 12.

Images