Throttle valve gas-solid erosion prediction method
The gas flow state is divided by Reynolds coefficient Re and clustering algorithm, combined with mechanism model and data-driven model, and the accuracy and adaptability of throttle gas-solid erosion prediction are solved, and the flow safety of shale gas field is improved.
Patent Information
- Application Number
- CN202311598192.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2023-11-27
- Publication Date
- 2025-07-11
AI Technical Summary
The prior art is difficult to achieve high-precision and strong adaptation prediction of throttle gas-solid erosion, which has led to the threat of the flow safety of shale gas fields. The erosion mechanism model is limited to specific operating conditions, and the data-driven model lacks physical explanatory nature.
The gas flow state is divided by Reynolds coefficient Re and clustering algorithm, and nonlinear expressions are generated by combining the mechanism model. Local prediction models are constructed through data-driven models, and model fusion is carried out to achieve high-precision prediction.
The accuracy and adaptability of throttle valve gas-solid erosion prediction is improved, the flow safety of shale gas field is enhanced, and high-precision erosion prediction is achieved.
Smart Images
Figure CN120299535A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the technical equipment field of oil and gas field development. Specifically, it relates to a method for predicting gas-solid erosion of a throttle valve. Background Art
[0002] To accelerate the exploration and development of shale gas in southern Sichuan, western Hubei, and Yunnan-Guizhou regions, multi-stage throttling needs to be set at the wellhead during shale gas field development. The throttle valve is a commonly used key throttling component and a crucial link connecting the surface gathering and transportation system with the underground wellbore. It can control the backpressure and flow rate by adjusting its opening degree to prevent accidents such as overflow, well kick, and even blowout. However, although there are downhole sand fixation and sand blocking measures, the produced shale gas still inevitably contains mud, sand, and other solid impurities. The solid particles continuously impact the throttle valve under the carrying of high-speed gas flow, causing the removal of internal valve materials and losses in mass and thickness, that is, erosion wear. Compared with straight pipe sections, elbows, and tees, the throttle valve has a higher flow rate, more particles, and a structure that is more likely to cause particle aggregation and reciprocating impact on the wall surface. Accidents such as perforation, fracture, and leakage caused by throttle valve erosion occur frequently, seriously threatening the flow safety of shale gas fields. Therefore, it is very important to achieve high-precision and strong-adaptability prediction of gas-solid flow erosion of throttle valves, which can enhance the flow safety during shale gas field development and contribute to the construction of intelligent oil and gas fields.
[0003] Since gas-solid two-phase pipe flow mostly causes erosion at key parts such as elbows and tees, scholars mostly focus on easily eroded components such as elbows and tees when studying gas-solid erosion, and rarely pay attention to throttle valves. Therefore, there is very little research on gas-solid erosion of throttle valves. Currently, the understanding of the gas-solid flow erosion mechanism of needle-type throttle valves is still far from sufficient. For example: What are the gas-solid flow characteristics and erosion characteristics under different opening degrees and variable flow channels of needle-type throttle valves? How does the flow characteristic affect the erosion characteristic? What is the internal relationship between the two? In addition, as a commonly used key throttling component, once problems such as perforation, fracture, and leakage are caused by throttle valve erosion, it will seriously threaten the flow safety of shale gas fields. Therefore, gas-solid erosion prediction of throttle valves is of great significance for promoting oil and gas flow assurance. Currently, the technologies for predicting erosion rate include mechanism models and data-driven models.
[0004] In terms of the erosion mechanism model, due to the different experimental working conditions and conditions, the constructed erosion models are limited to a certain working condition, or even a certain type of particle, a certain impact angle, a certain medium, and a certain type of erosion material. There is still no common conclusion and systematic conclusion. The erosion mechanism model has parameters with clear physical meanings, but due to the complexity of the erosion phenomenon, it is difficult to establish a model. During the modeling process, the conditions are often simplified, resulting in the erosion mechanism model being difficult to avoid calculation deviation problems. Due to the limitations of the erosion model itself, the adaptability of the established multiphase flow erosion prediction method also needs to be improved. In addition, the data-driven model directly depends on data for training, cannot extract relevant erosion physical information from parameters, is disconnected from the physical erosion mechanism, and is difficult to be practically applied. Summary of the Invention
[0005] In order to solve the problems that the current technology for predicting erosion rate still has calculation deviation and is disconnected from the physical erosion mechanism and is difficult to be practically applied, this application provides a method for predicting gas-solid erosion of a throttle valve.
[0006] The embodiments of this application are implemented as follows:
[0007] In a first aspect, this application provides a method for predicting gas-solid erosion of a throttle valve, including:
[0008] Clustering according to the erosion gas flow pattern;
[0009] Construct a mechanism-data series model for various gas flow patterns;
[0010] Based on the clustering results, cluster each gas flow pattern, construct a local prediction model, and generate a high-precision prediction model.
[0011] In a possible implementation manner, the clustering according to the erosion gas flow pattern further includes:
[0012] Simulate the gas-solid erosion of the throttle valve through computational fluid dynamics (CFD) simulation;
[0013] Obtain the parameter values required for calculating the Reynolds number Re;
[0014] Calculate the Re value;
[0015] Perform gas flow pattern clustering according to the Re value.
[0016] In a possible implementation manner, the construction of the mechanism-data series model for various gas flow patterns further includes:
[0017] Generate a non-linear expression according to the mechanism model;
[0018] Calculate the non-linear expression;
[0019] Input the value obtained by calculating the non - linear expression into the data - driven model.
[0020] In a possible implementation, based on the clustering results, clustering for each gas flow regime, constructing local prediction models, and generating a high - precision prediction model further includes:
[0021] Construct local prediction models for various gas flow regimes;
[0022] Divide the sample data and normalize it;
[0023] Evaluate the prediction performance of each model;
[0024] Select high - performance models and fuse them to generate a high - precision prediction model for throttle valve gas - solid erosion.
[0025] In a possible implementation, the Reynolds number Re is extremely important data in the field of fluid mechanics to characterize the flow state of a fluid. The calculation formula of the Reynolds number Re is as follows:
[0026]
[0027] In the formula: ρ is the fluid density, v is the fluid velocity, d is the characteristic diameter of the flow field, and μ is the dynamic viscosity coefficient.
[0028] In a possible implementation, the fluid density ρ, fluid velocity v, characteristic diameter d of the flow field, and dynamic viscosity coefficient μ are all obtained through CFD for their simulation values;
[0029] Substitute the simulation values of these parameters into the calculation formula of the Reynolds number Re, and the Reynolds coefficient Re under different working conditions can be obtained;
[0030] Then, according to the Reynolds coefficient Re, the CFD simulation results can be divided through a clustering algorithm.
[0031] In a possible implementation, data - driven models, that is, local prediction models, are constructed for various gas flow regimes through multiple methods such as the Random Forest (RF) algorithm, LightGBM algorithm, XGBoost algorithm, and BP neural network.
[0032] In a possible implementation, according to the root - mean - square error, mean absolute error percentage, and coefficient of determination as the regression prediction performance evaluation indicators of the data - driven model, evaluate the prediction performance of each model.
[0033] In a possible implementation, according to the evaluation results of the prediction performance of each model, select the relatively better - performing data - driven models corresponding to various gas flow regimes as the base learners, use the linear regression model as the meta - learner, and perform model fusion through the Stacking integration algorithm to generate a higher - performance data - driven model.
[0034] In a second aspect, the present application provides a throttling valve gas-solid erosion prediction device, including:
[0035] A gas clustering module for clustering according to the erosion gas flow pattern;
[0036] A model construction module for constructing a mechanism-data series model for various gas flow patterns;
[0037] An optimization and fusion module for constructing a local prediction model based on the clustering results of each gas flow pattern and generating a high-precision prediction model.
[0038] The technical solution provided by the present application can at least achieve the following beneficial effects:
[0039] The throttling valve gas-solid erosion prediction method and device provided by the present application use the Reynolds number Re and a clustering algorithm to divide the CFD simulation results with similar gas flow patterns into one category, thereby dividing into multiple clusters with large differences in gas flow patterns. Secondly, according to the erosion mechanism model, a corresponding non-linear expression is generated. For each gas flow pattern, the value of this expression is calculated and passed into the data-driven model to predict the CFD simulation erosion rate, forming a mechanism-data series model, thereby adding physical interpretability of the erosion model to the data-driven model. Finally, for each gas flow pattern, a local prediction model is constructed, and the model is optimized through hyperparameter tuning and model fusion to achieve high-precision prediction of throttling valve gas-solid erosion. BRIEF DESCRIPTION OF THE DRAWINGS
[0040] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.
[0041] Figure 1 It is a schematic flowchart of a throttling valve gas-solid erosion prediction method shown in an exemplary embodiment of the present application;
[0042] Figure 2 It is a specific flowchart of a throttling valve gas-solid erosion prediction method shown in an exemplary embodiment of the present application
[0043] Figure 3 It is a schematic flowchart of gas flow pattern clustering shown in an exemplary embodiment of the present application
[0044] Figure 4 It is a schematic flowchart of generating a non-linear expression shown in an exemplary embodiment of the present application
[0045] Figure 5 is a schematic flow chart of erosion prediction shown in an exemplary embodiment of the present application;
[0046] Figure 6 is a schematic structural diagram of a throttle valve gas-solid erosion prediction device shown in an exemplary embodiment of the present application.
[0047] Reference numerals:
[0048] 1. Gas clustering model; 2. Model construction module; 3. Optimization and fusion module. Detailed implementation manners
[0049] In order to make the purpose, implementation manners and advantages of the present application clearer, the following will clearly and completely describe the exemplary implementation manners of the present application with reference to the accompanying drawings in the exemplary embodiments of the present application. Obviously, the described exemplary embodiments are only a part of the embodiments of the present application, rather than all the embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application, and are not used to limit the present application.
[0050] It should be noted that the brief description of the terms in the present application is only for the convenience of understanding the subsequent described implementation manners, rather than intending to limit the implementation manners of the present application. Unless otherwise specified, these terms should be understood in their ordinary and general meanings.
[0051] The terms "first", "second", "third", etc. in the specification, claims and above-mentioned drawings of the present application are used to distinguish similar or homogeneous objects or entities, and do not necessarily mean to limit a specific order or sequence, unless otherwise noted. It should be understood that such terms can be interchanged under appropriate circumstances.
[0052] The terms "comprising" and "having" and any variations thereof are intended to cover but not exclude inclusion. For example, a product or device comprising a series of components does not necessarily have to be limited to all the clearly listed components, but may include other components not clearly listed or inherent to these products or devices.
[0053] For the convenience of describing the technical solutions of the application, some concepts related to the present application will be described first below.
[0054] Erosion: Erosion refers to the phenomenon that solid particles contained in a fluid flowing at high speed impact the wall surface, causing damage to the wall surface.
[0055] CFD: Computational Fluid Dynamics (CFD) is a classical research method for studying pipe flow and wall interaction, which is widely adopted by scholars at home and abroad. CFD (Computational Fluid Dynamics) software is used to model the fluid flow. The CFD modeling mainly uses fluid flow and turbulence models as well as existing empirical erosion models to track particles. By using CFD modeling, numerical schemes, turbulence models, near-wall treatment, meshing treatment, and discrete particle model parameters can be obtained in detail, thus providing guidance for the industry to prevent erosion.
[0056] K-means algorithm: Clustering is a commonly used data analysis method, which aims to divide a large amount of data into multiple categories so that as many data as possible are the same within each category, while having the greatest differences among different categories. The K-means algorithm is a clustering analysis method based on partitioning, aiming to divide data that are relatively similar in some aspects into the same region.
[0057] Mechanism Model (MM): The Mechanism Model (MM), also known as the First Principle Model (FPM), mainly uses principles such as energy, mass, momentum conservation, and reaction kinetics to describe the physical laws of the studied process, and at the same time obtains corresponding mathematical expressions based on rigorous logical deductions. Since the model parameters have clear physical meanings and the function transfer relationships are very clear, the mechanism model is also called the "White Box Model" (WBM).
[0058] Data-Driven Model (DDM): The Data-Driven Model (DDM) uses machine learning algorithms to explore the internal correlations of measurement data. Commonly used machine learning algorithms are divided into supervised learning and unsupervised learning. Supervised learning mainly includes support vector machine algorithms, neural networks, ensemble learning, classification algorithms, etc., and unsupervised learning mainly includes dimensionality reduction algorithms, clustering algorithms, etc. The data-driven model describes the system by correlating input and output data information. Since the model decision function does not reflect the process characteristics and relevant physical information cannot be extracted from the parameters, it is also called the "Black Box Model" (BBM).
[0059] Hybrid Model (HM): The advantage of the mechanism model lies in its ability to explore the physical essence of the research object, with clear physical meaning of the model and high reliability. The data-driven model can learn the correlation features contained in on-site samples and deduce the operation rules of the process under study. The organic integration of the mechanism model and the data-driven model forms a hybrid model (Hybrid Model, HM), which can promote the improvement of the overall accuracy and reliability of the model. The hybrid model is a research hotspot in the current modeling field. The hybrid model combines the advantages of both, so it is also called the "Grey Box Model" (Grey Box Model, GBM).
[0060] Random Forest (RF): The Random Forest (Random Forest, RF) algorithm is based on the Bagging ensemble idea and consists of multiple decision trees forming a "forest". The "randomness" is reflected in the training process of these decision trees. In this process, a subset containing multiple attributes is randomly selected from all candidate attributes. Then the optimal attribute is selected from this subset for partitioning. It has the advantages of high classification accuracy, fast learning speed, small generalization error, accurate output results, fast learning speed, small generalization error, and can output the importance ranking of model features, etc.
[0061] Light Gradient Boosting Machine (LightGBM): The Light Gradient Boosting Machine (Light Gradient Boosting Machine, LightGBM) algorithm is an ensemble learning method of the Boosting class released by Microsoft in 2017, with the advantages of high computational efficiency and high accuracy. Compared with other Boosting ensemble methods, LightGBM adds the Gradient-based One-Side Sampling (GOSS) and Exclusive Feature Bundling (EFB) algorithms.
[0062] Extreme Gradient Boosting (XGBoost): The extreme gradient boosting (extreme gradient boosting, XGBoost) algorithm belongs to the Boosting class of ensemble learning methods. This model can handle missing values without imputation processing and has the advantages of high accuracy and fast running speed.
[0063] BP Neural Network: The BP (back propagation neural network) neural network is a multi-layer feedforward neural network trained according to the error backpropagation algorithm and is one of the most widely used neural network models. The neural network consists of three parts: the input layer, the hidden layer, and the output layer. The standard BP algorithm includes two stages: forward propagation and backward propagation. If the expected output value is not finally obtained, the thresholds and weights of each layer can be further adjusted through the error until the ideal result is achieved and the algorithm ends. Therefore, this algorithm has a high degree of adaptability.
[0064] Linear Regression (LR): Linear Regression (LR) analysis mainly studies the relationships between variables. It uses a line to fit all data points and then studies how to minimize the distance differences from the line to all data points. Linear regression is the most common regression analysis algorithm. Linear regression processes the observed data to obtain a mathematical model expression that relatively conforms to the law, that is, to find the law between the independent variable data and the dependent variable data, so as to simulate the results of unknown data.
[0065] Stacking Ensemble Algorithm: The Stacking method is an ensemble learning method that includes two layers of prediction models. The first-layer prediction model is called the base learner, and the second-layer prediction model is called the meta-learner. The Stacking method combines the advantages of different learners by integrating multiple base learners, so that the prediction model has strong generalization ability. In addition, to improve the overall prediction accuracy, the meta-learner is used to further optimize the output of the base learners.
[0066] Before explaining the throttle valve gas-solid erosion prediction method provided by the embodiments of the present application, the application scenarios and implementation environments of the embodiments of the present application are introduced first.
[0067] It is necessary to accelerate the exploration and development of shale gas in southern Sichuan, western Hubei and Yunnan-Guizhou regions. In the development of shale gas fields, multi-stage throttling needs to be set at the wellhead. The throttle valve is a commonly used key throttling component and a key link connecting the surface gathering and transportation system and the underground wellbore. It can control the back pressure and flow rate by adjusting its opening degree to prevent accidents such as overflow, well kick, and even blowout. However, although there are downhole sand control and sand blocking measures, the produced shale gas still inevitably contains mud sand and other solid impurities. The solid particles continuously impact the throttle valve under the carrying of high-speed airflow, causing the removal of the internal materials of the valve and the loss of mass and thickness, that is, erosion wear. Compared with straight pipe sections, elbows and tees, the throttle valve has a higher flow rate, more particles and a structure that is more likely to cause particle aggregation and reciprocating impact on the wall surface. Accidents such as perforation, fracture and leakage caused by throttle valve erosion occur frequently, seriously threatening the flow safety of shale gas fields. Therefore, it is very important to achieve high-precision and strong-adaptability prediction of throttle valve gas-solid flow erosion, which can enhance the flow safety during the development of shale gas fields and contribute to the construction of intelligent oil and gas fields.
[0068] Since erosion in gas-solid two-phase pipe flow mostly occurs at key components such as elbows and tees, scholars mostly focus on easily eroded components like elbows and tees when studying gas-solid erosion, while rarely paying attention to throttle valves. Therefore, there is very little research on gas-solid erosion of throttle valves. Currently, the understanding of the gas-solid flow erosion mechanism of needle throttle valves is far from sufficient. For example, what are the gas-solid flow characteristics and erosion characteristics under different opening degrees and variable flow channels of needle throttle valves? How does the flow characteristic affect the erosion characteristic? What is the internal relationship between the two? In addition, as a commonly used key throttling component, once erosion of the throttle valve causes problems such as perforation, fracture, and leakage, it will seriously threaten the flow safety of shale gas fields. Therefore, gas-solid erosion prediction of throttle valves is of great significance for promoting the guarantee of oil and gas flow.
[0069] Currently, the techniques for erosion rate prediction include mechanism models and data-driven models. The schemes and disadvantages of the two are as follows:
[0070] (1) Mechanism models. When particles collide with the wall surface, the erosion rate can be calculated through erosion models. Scholars have proposed more than 200 empirical or semi-empirical erosion models through theoretical derivation combined with experimental research. Various erosion models consider a large number of physical parameters. However, due to the different experimental conditions and situations, the constructed erosion models are limited to a certain working condition, or even a certain type of particle, a certain impact angle, a certain medium, and a certain type of erosion material, etc., and it is still impossible to form a general conclusion and a systematic conclusion. The erosion mechanism model has parameters with clear physical meanings. Due to the complexity of the erosion phenomenon, it is difficult to establish a model. During the modeling process, the conditions are often simplified, resulting in the erosion mechanism model being difficult to avoid calculation deviation problems. Due to the limitations of the erosion model itself, the adaptability of the established multiphase flow erosion prediction method also needs to be improved.
[0071] (2) Data-driven models. With the maturity of machine learning, in recent years, many scholars have used machine learning algorithms such as random forests and BP neural networks to construct data-driven erosion prediction models to predict erosion wear. However, the resulting data-driven models directly rely on data for training, cannot extract relevant erosion physical information from the parameters, are disconnected from the physical erosion mechanism, and are difficult to apply in practice.
[0072] Based on this, the present application provides a throttle valve gas-solid erosion prediction method that uses the Reynolds number Re and a clustering algorithm to divide the CFD simulation results with similar gas flow states into one category, thereby dividing into multiple clusters with significantly different gas flow states. Secondly, according to the erosion mechanism model, a corresponding non-linear expression is generated. For each type of gas flow state, the value of this expression is calculated and passed into the data-driven model to predict the CFD simulation erosion rate, forming a mechanism-data series model, thereby adding physical interpretability of the erosion model to the data-driven model. Finally, for each type of gas flow state, a local prediction model is constructed, and the model is optimized through hyperparameter tuning and model fusion to achieve high-precision prediction of throttle valve gas-solid erosion.
[0073] Next, the technical solution of the present application, as well as how the technical solution of the present application solves the above technical problems, will be specifically described through embodiments in combination with the accompanying drawings. The embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. Obviously, the described embodiments are part of the embodiments of the present application, rather than all of the embodiments.
[0074] Figure 1 It is a flowchart showing a throttle valve gas-solid erosion prediction method shown in an exemplary embodiment of the present application.
[0075] In an exemplary embodiment, as Figure 1 shown, a throttle valve gas-solid erosion prediction method is provided. In this embodiment, the method may include the following steps:
[0076] Step 100: Cluster according to the erosion gas flow state.
[0077] Step 200: Construct a mechanism-data series model for each type of gas flow state.
[0078] Step 300: Based on the clustering results, for the clustering of each gas flow state, construct a local prediction model and generate a high-precision prediction model.
[0079] Figure 2 It is a specific flowchart showing a throttle valve gas-solid erosion prediction method shown in an exemplary embodiment of the present application, Figure 3 It is a flowchart showing the gas flow state clustering shown in an exemplary embodiment of the present application, Figure 4 It is a flowchart showing the generation of a non-linear expression shown in an exemplary embodiment of the present application, Figure 5 It is a flowchart showing the erosion prediction shown in an exemplary embodiment of the present application.
[0080] In a possible implementation, the throttle valve gas-solid erosion prediction method based on clustering and series model can be divided into 3 steps, as Figure 2As shown below, it specifically includes the following implementation steps:
[0081] (1) Cluster according to gas flow regime
[0082] Since it is difficult to obtain all the data required to affect erosion on site and it is also difficult to build a real erosion working condition in the laboratory, and CFD simulation has the advantages of saving time, low cost, and being able to simulate more complex processes, the method of CFD simulation is used in the present invention to obtain the required data.
[0083] Solid particles continuously impact the throttle valve under the carry of high-speed airflow, causing erosion wear of the internal materials of the valve. Due to the complex structure of the throttle valve, the flow characteristics inside the flow channel are variable. Therefore, when studying the gas-solid two-phase flow erosion of the throttle valve, it is necessary to identify the gas flow regimes of each working condition. And the Reynolds number Re is an extremely important data in the field of fluid mechanics, which can accurately characterize the flow state of the fluid. Specifically, paying attention to Re can better judge whether different samples have similar physical conditions and fluid flow patterns. The calculation formula of the Reynolds number Re is as follows:
[0084]
[0085] In the formula: ρ is the fluid density, v is the fluid velocity, d is the characteristic diameter of the flow field, and μ is the dynamic viscosity coefficient.
[0086] The simulated values of the above parameters can be obtained through CFD. Substituting the simulated values of these parameters into formula (1), the Reynolds coefficient Re under different working conditions can be obtained.
[0087] According to the Reynolds coefficient Re, the CFD simulation results can be divided through a clustering algorithm. For example, through the K-means clustering algorithm, samples with similar gas flow regimes can be divided into the same category, thereby dividing into multiple clusters with significantly different gas flow regime characteristics. Specifically as Figure 3 shown.
[0088] (2) Generate a non-linear expression
[0089] In order to make up for the shortcoming that the data-driven model cannot extract relevant physical information from parameters, a mechanism model can be used to provide physical laws to form a series model structure with the output of the mechanism model as the input of the data-driven model as Figure 4 shown. In this series model, the output of the mechanism model is incorporated into the data-driven model as prior knowledge, thereby guiding the data-driven model to optimize parameters, improving the physical interpretability of the data-driven model, and enhancing the fitting ability of the data-driven model.
[0090] The present invention establishes a non - linear expression G() of the erosion mechanism model, and uses the calculation result of the non - linear expression as the input of the data - driven model, so that the data - driven model absorbs the physical regularity in the semi - empirical and semi - mechanism equation. Taking the Oka erosion model as an example, considering the impact particles as quartz sand, some coefficients of the Oka model are set as reference values. The formula of the Oka model can be simplified to formula (2):
[0091]
[0092] At this time, the sample (x, y) can be converted to (G(x), y), and G() is shown in formula (3):
[0093]
[0094] Among them, for formula (2) and formula (3),
[0095] (3) Erosion prediction
[0096] Under the clustering of similar gas flow patterns, a local prediction model is constructed to effectively improve the prediction accuracy. As shown in, the specific steps are as follows: Figure 5 shown, specifically including the following steps:
[0097] Construct a data - driven model, such as the random forest (RF) algorithm, light gradient boosting machine (LightGBM) algorithm, extreme gradient boosting (XGBoost) algorithm, and BP neural network.
[0098] For the data samples of various gas flow patterns, first perform normalization processing, and then divide the normalized data into a training set, a validation set, and a test set according to the ratio of 6:2:2.
[0099] Optimize the hyperparameters of the data - driven model for various gas flow patterns to determine the optimal hyperparameter combination.
[0100] Take the root mean square error (RMSE), mean absolute error percentage (MAE), and coefficient of determination R 2 as the regression prediction performance evaluation indicators of the data - driven model, and screen out the high - performance data - driven models corresponding to various gas flow patterns.
[0101] For various gas flow patterns, select the data - driven model with better performance as the base learner, use the linear regression model as the meta - learner, and perform model fusion through the Stacking integration algorithm to generate a higher - performance data - driven model.
[0102] It should be understood that although the steps in the flowcharts involved in the above embodiments are shown in sequence as indicated, these steps are not necessarily executed in the order indicated. Unless there is a clear indication in this document, there is no strict order limit for the execution of these steps, and these steps can be executed in other orders. Moreover, at least some of the steps in the flowcharts involved in the above embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be executed alternately or in turn with at least some of the steps or stages in other steps or other steps.
[0103] Corresponding to the embodiments of the throttle valve gas-solid erosion prediction method described above and adopting the same technical concept, the present application also provides embodiments of a throttle valve gas-solid erosion prediction device.
[0104] Figure 6 It is a schematic structural diagram of a throttle valve gas-solid erosion prediction device shown in an exemplary embodiment of the present application.
[0105] In an exemplary embodiment, as Figure 6 shown, the throttle valve gas-solid erosion prediction device includes:
[0106] A gas clustering module 1, configured to perform clustering according to the erosion gas flow pattern;
[0107] A model construction module 2, configured to construct a mechanism-data series model for various types of gas flow patterns;
[0108] An optimization and fusion module 3, configured to construct a local prediction model based on the clustering results of each gas flow pattern and generate a high-precision prediction model.
[0109] For the specific limitations of the throttle valve gas-solid erosion prediction device, reference can be made to the limitations of the throttle valve gas-solid erosion prediction method in the above text, which will not be elaborated here. Each module in the above throttle valve gas-solid erosion prediction device can be implemented in whole or in part through software, hardware, and their combination. The above modules can be embedded in the processor of the computer device in hardware form or independent of it, or stored in the memory of the computer device in software form, so that the processor can call and execute the operations corresponding to the above modules.
[0110] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope recorded in this specification.
[0111] The embodiments described above merely represent several implementation manners of the present application. The description is relatively specific and detailed, but it should not be construed as a limitation on the scope of the invention patent. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present application, several modifications and improvements can still be made, and these all fall within the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the appended claims.
Claims
1. A method for predicting gas-solid erosion of a throttle valve, characterized in that, Including: Conduct clustering according to the erosion gas flow pattern; Construct a mechanism-data series model for various gas flow patterns; Based on the clustering results, construct local prediction models for the clustering of each gas flow pattern and generate a high-precision prediction model.
2. The throttle valve gas-solid erosion prediction method according to claim 1, characterized in that The clustering according to the erosion gas flow pattern further includes: Simulate the gas-solid erosion of the throttle valve through computational fluid dynamics (CFD) simulation; Obtain the parameter values required to calculate the Reynolds number Re; Calculate the Re value; Conduct gas flow pattern clustering according to the Re value.
3. The throttle valve gas-solid erosion prediction method according to claim 1, wherein The construction of the mechanism-data series model for various gas flow patterns further includes: Generate a non-linear expression according to the mechanism model; Calculate the non-linear expression; Input the value obtained by calculating the non-linear expression into the data-driven model.
4. The throttle valve gas-solid erosion prediction method according to claim 3, characterized in that The construction of local prediction models for the clustering of each gas flow pattern based on the clustering results and generating a high-precision prediction model further includes: Construct local prediction models for various gas flow patterns; Divide the sample data and normalize it; Evaluate the prediction performance of each model; Screen out high-performance models and fuse them to generate a high-precision prediction model for the gas-solid erosion of the throttle valve.
5. The throttle valve gas-solid erosion prediction method according to claim 2, characterized in that The Reynolds number Re is extremely important data in the field of fluid mechanics to characterize the flow state of the fluid. The calculation formula of the Reynolds number Re is as follows: In the formula: ρ is the fluid density, v is the fluid velocity, d is the characteristic diameter of the flow field, and μ is the dynamic viscosity coefficient.
6. The throttle valve gas-solid erosion prediction method according to claim 5, wherein, The simulated values of the fluid density ρ, fluid velocity v, characteristic diameter d of the flow field, and dynamic viscosity coefficient μ are all obtained through CFD; Substitute the simulated values of these parameters into the calculation formula of the Reynolds number Re to obtain the Reynolds coefficient Re under different working conditions; Then, according to the Reynolds coefficient Re, the CFD simulation results can be divided through a clustering algorithm.
7. The throttle valve gas-solid erosion prediction method according to claim 4, wherein Construct data-driven models, namely local prediction models, for various gas flow patterns through multiple methods such as the random forest (RF) algorithm, light gradient boosting algorithm, extreme gradient boosting algorithm, and BP neural network.
8. The throttle valve gas-solid erosion prediction method according to claim 4, characterized in that, Evaluate the prediction performance of each model according to the root mean square error, mean absolute error percentage, and coefficient of determination as the regression prediction performance evaluation indicators of the data-driven model.
9. The throttle valve gas-solid erosion prediction method according to claim 4, characterized in that, According to the evaluation results of the prediction performance of each model, screen out the data-driven models with better performance corresponding to various gas flow patterns as the base learners, use the linear regression model as the meta-learner, and perform model fusion through the Stacking integration algorithm to generate a higher-performance data-driven model.
10. A throttling valve gas-solid erosion prediction device, characterized in that, Including: A gas clustering module for clustering according to the erosion gas flow pattern; A model construction module for constructing a mechanism-data series model for various gas flow patterns; An optimization and fusion module for constructing local prediction models for the clustering of each gas flow pattern based on the clustering results and generating a high-precision prediction model.