Intelligent Flow Prediction Method for Solenoid Valves Based on Structural Parameter and Operating Condition Fusion
By constructing an intelligent flow prediction method for solenoid valves that integrates structural parameters and operating parameters, and utilizing support vector regression and ridge regression models, the problems of low computational efficiency and insufficient data reuse in CFD simulation methods for solenoid valve flow analysis are solved, achieving efficient and accurate flow prediction and optimized design.
Patent Information
- Application Number
- CN202611244641.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-17
- Publication Date
- 2026-11-10
AI Technical Summary
Existing CFD simulation methods suffer from low computational efficiency, high cost, and insufficient data reuse capabilities in solenoid valve flow analysis, making it difficult to meet the needs of rapid design and optimization.
By establishing a dual-model fusion prediction method combining support vector regression and ridge regression, an intelligent prediction method for solenoid valve flow rate is constructed that integrates structural parameters and operating parameters. Existing simulation data is used to establish data mapping relationships, reducing redundant CFD calculations.
It improves the accuracy and efficiency of flow forecasting, reduces computational costs, enhances the ability to describe the coupling effect between structural parameters and operating conditions, and improves stability and data utilization efficiency under finite sample conditions.
Smart Images

Figure CN123069960A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of electromagnetic valve testing technology, specifically relating to an intelligent flow prediction method for electromagnetic valves based on the fusion of structural parameters and operating conditions. Background Technology
[0002] Solenoid valves are important fluid control components that use electromagnetic actuation to achieve fluid flow control, regulation, and regulation. They are widely used in hydraulic systems, pneumatic systems, automotive control systems, and industrial automation equipment. The flow characteristics of a solenoid valve directly affect the system's response speed, control accuracy, and operational stability. Therefore, accurate analysis and rapid prediction of the flow performance of solenoid valves are crucial technical aspects of solenoid valve structural design and performance optimization.
[0003] As fluid control systems evolve towards higher precision, higher response, and miniaturization, the internal flow channel structure of solenoid valves is becoming increasingly complex. Modern solenoid valves typically include stepped structures, cylindrical structures, plateau structures, flow orifices, and localized contraction and expansion regions. Their outlet flow rate is influenced by both structural parameters and operating conditions. Geometric parameters such as step height, cylindrical height, plateau height, orifice dimensions, and height differences between different regions alter the effective fluid flow area and localized pressure loss; operating parameters such as inlet pressure, outlet pressure, and pressure differential affect fluid velocity distribution and pressure variations. Therefore, solenoid valve flow rate variations exhibit a clear structure-operating condition coupling characteristic.
[0004] Currently, flow analysis of solenoid valves mainly employs methods such as theoretical calculations, experimental testing, and numerical simulations. Traditional theoretical calculation methods typically establish computational relationships based on parameters such as flow coefficient, pressure difference, and flow area, offering high computational efficiency. However, due to the use of idealized assumptions, they struggle to accurately describe the nonlinear flow behavior within complex internal channels. Experimental testing methods can obtain actual flow results, but require fabricating different structural samples and building test platforms, resulting in high costs and limited parameter coverage.
[0005] With the development of computer numerical computing technology, computational fluid dynamics (CFD) has gradually become an important method for flow analysis of solenoid valves. This method establishes a fluid computation model to numerically solve the internal pressure field, velocity field, and flow distribution, enabling accurate analysis of flow characteristics under different structural parameters and operating conditions, providing data for solenoid valve design optimization. A patent with publication number CN111767663A discloses a convenient method for calculating valve flow coefficients based on CFD simulation. The main steps include: establishing flow field models of the upstream pipeline, valve, and downstream pipeline based on a three-dimensional model of the valve assembly; dividing and optimizing the flow field model mesh; applying the re-optimized mesh to set boundary conditions for simulation calculation; collecting average pressure values from multiple cross-sections of the upstream pipeline; determining the linear fitting order based on the axial pressure distribution diagram and performing linear fitting; combining the total pressure difference of the pipeline with the valve obtained from the simulation calculation results; and calculating the valve flow coefficient using the valve pressure difference. Although the valve flow analysis method based on CFD simulation can accurately obtain the internal flow state and flow performance of the valve, the method still has the following shortcomings: (1) The computational efficiency is low and it is difficult to meet the needs of rapid multi-parameter design. The CFD method usually requires to rebuild the geometric model, divide the computational grid, set the boundary conditions and perform numerical solution for each valve structure. When it is necessary to analyze the combination of multiple structural parameters or carry out structural optimization design, a large number of simulation calculations need to be repeated, resulting in a long calculation cycle. (2) It is difficult to establish a unified mapping relationship between structural parameters and operating conditions. The existing CFD method mainly performs flow field analysis for a single structural model and specific operating conditions. Its calculation results are usually used as the basis for single design evaluation. When the structural parameters and operating conditions such as inlet pressure change at the same time, simulation calculations need to be carried out again, which is difficult to directly reflect the flow change law under the combined action of structural parameters and operating condition parameters. (3) The efficiency of simulation data utilization is insufficient. CFD calculation can generate a large amount of flow field data, but the existing method usually only uses the current calculation results for performance evaluation. Historical simulation data is difficult to be directly used for rapid prediction of subsequent structural schemes. When facing a large number of design scheme screening and parameter optimization tasks, a lot of computational resources are required. Therefore, although existing CFD simulation-based flow analysis methods have high accuracy, they suffer from high computational costs, low prediction efficiency, and insufficient data reuse capabilities, making it difficult to meet the development needs of rapid design and intelligent optimization of solenoid valves. Summary of the Invention
[0006] This invention addresses the problems existing in the prior art by designing an intelligent flow prediction method for solenoid valves based on the fusion of structural parameters and operating conditions. This invention establishes a data mapping relationship between structural parameters, operating conditions, and outlet flow rate using existing simulation data, enabling the trained model to directly predict flow results under the target structure and operating conditions, reducing redundant CFD calculations. By constructing a fusion input feature of structural and operating parameters, the geometric and operational factors affecting flow performance are uniformly expressed, improving the model's ability to represent the coupled influence of structure and corresponding operating conditions. Furthermore, by establishing a dual-model fusion prediction method combining support vector regression and ridge regression models, the nonlinear learning capability and trend constraint capability are utilized respectively to improve the model's prediction accuracy and generalization ability.
[0007] The objective of this invention is achieved through the following technical solution: a method for intelligent prediction of solenoid valve flow rate based on the fusion of structural parameters and operating conditions, comprising the following steps: Step 1: Determine the structural parameters and operating conditions of the solenoid valve; establish a parameterized three-dimensional model of the solenoid valve, and determine the key structural parameters affecting the outlet flow rate based on the internal flow channel structure and design requirements of the solenoid valve; Step 2: Generate parameter combinations corresponding to the inlet pressure of the structure, and generate multiple sets of solenoid valve structural schemes based on the allowable value range of each structural parameter; the structural parameter combinations are generated through a parameter sampling method; Step 3: Conduct CFD simulation and extract the steady-state outlet flow rate; conduct computational fluid dynamics simulation for each set of parameter combinations corresponding to the inlet pressure of the structure to obtain the steady-state outlet flow rate under the corresponding conditions. Step 4: Construct fusion features and divide the dataset. Based on the flow mechanism of the solenoid valve, construct interactive features of height difference, size ratio, effective flow area, and pressure corresponding to the structure. Step 5: Establish and screen flow prediction models. Using the standardized fusion features of the inlet pressure corresponding to the structure as input and the CFD steady-state outlet flow rate as output, establish multiple regression prediction models respectively. Step 6: Construct an SVR-Ridge weighted ensemble model. Input the same fusion features into the SVR model and the Ridge model respectively. Step 7: Output the flow prediction results. For the new solenoid valve structure to be predicted, obtain its key structural parameters and target inlet pressure, and construct fusion features and perform standardization processing according to the same method as in the training stage.
[0008] Preferably, the structural parameters in step one include valve port size parameters, step height parameters, column height parameters, platform height parameters, and height difference parameters between different regions; the structural parameters also include the number and diameter of flow channel holes, flow channel width, length, and other dimensional parameters affecting the effective flow area; the structural parameter vector of the i-th solenoid valve sample is defined as:
[0009]
[0010] in: This represents the step height of the i-th solenoid valve sample; ~ Indicates the column height parameters for different regions; Indicates the platform height parameter; Indicates valve port size, orifice structure size, or other structural parameters;
[0011] Multiple inlet pressure conditions are set according to the actual working range of the solenoid valve. The inlet pressure set is represented as follows:
[0012]
[0013] in, Let represent the inlet pressure of the j-th group, and n represent the number of inlet pressure conditions.
[0014] This invention combines the structural parameters and operating condition parameters of the solenoid valve into a joint model. Structural features such as step height, column height, platform height, and orifice size, along with operating condition information such as inlet pressure, outlet pressure, and differential pressure, are used as inputs to the prediction model to establish a mapping relationship between structural parameters, operating condition parameters, and outlet flow rate.
[0015] Preferably, the combination of structural parameters in step two can be generated through manual design, full factorial experimental design, orthogonal experimental design, uniform design, and Latin hypercube sampling. When generating the structural scheme, the geometric rationality of each parameter combination is checked to eliminate invalid structures that have structural interference, closed flow channels, insufficient wall thickness, or do not meet processing constraints. For the i-th solenoid valve structural sample, multiple sets of inlet pressure conditions are set respectively.
[0016] The i-th structural sample and the j-th inlet pressure together form a set of structural-operating condition parameter combinations:
[0017]
[0018] The corresponding model input feature vector is represented as:
[0019]
[0020] By combining each structural scheme with multiple inlet pressure conditions, a set of parameter combinations covering different structural forms and different inlet pressure conditions is formed.
[0021] Preferably, the simulation in step three includes establishing a fluid computational domain, meshing, setting fluid properties and boundary conditions, performing steady-state flow field solving, and extracting steady-state outlet flow rate; for the i-th solenoid valve structural sample and the j-th inlet pressure, the corresponding outlet mass flow rate or outlet volumetric flow rate is extracted; when the CFD simulation output is mass flow rate, it is converted to volumetric flow rate according to the fluid density.
[0022]
[0023] in, This represents the steady-state outlet mass flow rate of the i-th solenoid valve structural sample under the j-th inlet pressure; Indicates fluid density; This represents the corresponding steady-state outlet volumetric flow rate;
[0024] To reduce simulation iteration fluctuations, the average outlet flow rate over K consecutive iterations after convergence can be used as the final flow rate label.
[0025]
[0026] in, This represents the outlet flow rate of the i-th structural sample at the k-th stable iteration step under the j-th inlet pressure.
[0027] The final dataset consists of structural parameters, inlet pressure, and steady-state outlet flow rate. .
[0028] This invention establishes a data sample library for model training by acquiring flow data under different combinations of solenoid valve structural parameters and corresponding operating conditions, enabling existing performance data to be used for rapid flow prediction of subsequent new structural schemes.
[0029] Preferably, in step four, the input features are standardized to eliminate the influence of different parameter units and numerical ranges:
[0030]
[0031] in, and Let represent the mean and standard deviation of the j-th feature in the training set, respectively;
[0032] The processed data is divided into training set, validation set and test set; preferably, the data is grouped according to the solenoid valve structure sample so that samples of different inlet pressures corresponding to the same structure are located in only the same data subset.
[0033] Preferably, the mapping relationship established by each model in step five is expressed as follows:
[0034] Where Z represents the standardized fusion feature. Indicates predicted flow;
[0035] The candidate models were evaluated using validation and test sets, with mean absolute error (MAE), root mean square error (RMSE), and coefficient of determination (R²) used as evaluation metrics.
[0036]
[0037]
[0038]
[0039] Based on a comprehensive comparison of the prediction accuracy, stability, and generalization ability of the models, the SVR model and the Ridge model were selected as the basic prediction branches of the weighted ensemble model.
[0040] This invention employs a prediction method that combines support vector regression (SVR) and ridge regression models. SVR is used to describe the nonlinear relationship between structural parameters, operating parameters, and flow rate, while ridge regression is used to constrain the overall trend of change. The final flow rate prediction result is obtained through model fusion.
[0041] Preferably, in step six, the SVR model is used to learn the complex nonlinear relationship between the solenoid valve's structural parameters, inlet pressure, and outlet flow rate, and its prediction result is as follows:
[0042]
[0043] The Ridge model is used to learn the overall trend of outlet flow rate as a function of structural parameters and inlet pressure. Its prediction results are as follows:
[0044]
[0045] The prediction results from the two models are then weighted and fused:
[0046] in:
[0047]
[0048]
[0049] The fusion weights can be determined through traversal search, cross-validation, or minimizing the prediction error on the validation set.
[0050]
[0051] By combining the nonlinear fitting capability of SVR with the trend stabilization capability of Ridge, the accuracy and generalization ability of traffic prediction results are improved.
[0052] Preferably, in step seven, the fused features are input into the SVR model and the Ridge model respectively, and then the final traffic prediction value is calculated based on the determined fusion weights:
[0053]
[0054] Output the steady-state outlet flow rate of the target solenoid valve structure under the corresponding inlet pressure condition.
[0055] This invention adjusts the weights of the output results of different prediction models based on the prediction errors in the validation data, so that the fusion model can obtain more stable prediction performance according to the characteristics of the actual data.
[0056] Based on the internal flow mechanism of the solenoid valve, the original structural parameters are extended to construct integrated features such as height difference, size ratio, and effective flow area that can reflect the flow characteristics, thereby improving the model's ability to express complex flow relationships.
[0057] Compared with the prior art, the present invention has the following beneficial effects:
[0058] (1) It reduces the cost of repetitive simulation calculations in the flow analysis process of solenoid valves.
[0059] This invention establishes a data sample database relating solenoid valve structural parameters, operating condition parameters, and outlet flow rate. This database stores flow information under different structural schemes and operating conditions in a unified manner, and uses a trained prediction model to replace some repetitive fluid simulation calculations. When new solenoid valve structural and operating parameters are input, the corresponding flow prediction results can be obtained without re-executing the complete geometric modeling, mesh generation, and flow field solution process. This reduces the number of CFD simulation calls during multi-scheme design and improves the efficiency of flow performance evaluation.
[0060] (2) It improves the model’s ability to describe the coupling effect between structural parameters and operating conditions.
[0061] This invention integrates solenoid valve structural parameters such as step height, column height, platform height, orifice structure size, and height difference with operating condition parameters such as inlet pressure, outlet pressure, and pressure difference to construct a unified input feature space. Compared to prediction methods that only consider a single structural factor or a single operating state, this method can simultaneously reflect the comprehensive impact of geometric structure changes and flow condition changes on outlet flow rate, improving the model's adaptability to flow rate variations under different structural schemes and operating conditions.
[0062] (3) Improved the stability and accuracy of flow prediction under limited sample conditions.
[0063] This invention employs a dual-model fusion prediction method consisting of a support vector regression (SVR) model and a ridge regression model. The SVR model learns the complex nonlinear relationship between structural and operational parameters and flow rate, while the ridge regression model constrains flow rate variation trends, improving model output stability. By weighted fusion of the prediction results from the two models and determining the fusion weights based on validation errors, the prediction fluctuations caused by the limited sample size of a single prediction model can be reduced, thus improving the model's generalization ability under limited data conditions.
[0064] (4) Improved the data utilization efficiency of the solenoid valve structure optimization design.
[0065] This invention utilizes flow data obtained from CFD simulations or experiments to establish a reusable data-driven model, enabling existing structural and operational samples to be used for performance prediction of new structures. Compared to the traditional method of re-analyzing performance for each designed structure, this invention can quickly evaluate the flow variation trend under different combinations of structural parameters, providing data support for adjusting solenoid valve structural parameters and optimizing performance. Attached Figure Description
[0066] Figure 1 This is a flowchart illustrating the specific implementation process of the present invention. Detailed Implementation
[0067] The present invention will be further described below with reference to the embodiments illustrated in the accompanying drawings:
[0068] like Figure 1 As shown, this embodiment discloses an intelligent flow prediction method for solenoid valves based on the fusion of structural parameters and operating conditions. The method includes the following steps: Step 1: Determine the structural parameters and operating conditions of the solenoid valve. First, a parametric three-dimensional model of the solenoid valve is established. Based on the internal flow channel structure and design requirements of the solenoid valve, the key structural parameters affecting the outlet flow rate are determined. The structural parameters include valve port size parameters, step height parameters, column height parameters, platform height parameters, and height difference parameters between different regions. The structural parameters also include the number and diameter of flow channel holes, flow channel width, length, and other dimensional parameters affecting the effective flow area.
[0069] Define the structural parameter vector of the i-th solenoid valve prototype as:
[0070]
[0071] in:
[0072] This represents the step height of the i-th solenoid valve sample;
[0073] ~ Indicates the column height parameters for different regions;
[0074] Indicates the platform height parameter;
[0075] This indicates the valve port size, orifice structure size, or other structural parameters.
[0076] Multiple inlet pressure conditions are set according to the actual working range of the solenoid valve. The inlet pressure set is represented as follows:
[0077]
[0078] in, Let represent the inlet pressure of the j-th group, and n represent the number of inlet pressure conditions.
[0079] The exit boundary conditions remain consistent across all simulation groups and are not used as parameters of varying operating conditions or as input features for machine learning models.
[0080] Step 2: Generate parameter combinations corresponding to the inlet pressure of the structure; generate multiple sets of solenoid valve structure schemes based on the allowable value range of each structural parameter.
[0081] Structural parameter combinations can be generated through manual design, full factorial experimental design, orthogonal experimental design, uniform design, Latin hypercube sampling, or other parameter sampling methods.
[0082] When generating structural schemes, the geometric rationality of each parameter combination is checked to eliminate invalid structures that have structural interference, closed flow channels, insufficient wall thickness, or do not meet processing constraints.
[0083] For the i-th solenoid valve structural prototype, multiple sets of inlet pressure conditions are set. The i-th structural prototype and the j-th set of inlet pressures together form a set of structural-operating condition parameter combinations:
[0084]
[0085] The corresponding model input feature vector is represented as:
[0086]
[0087] By combining each structural scheme with multiple inlet pressure conditions, a set of parameter combinations covering different structural forms and different inlet pressure conditions is formed.
[0088] Step 3: Conduct CFD simulation and extract steady-state outlet flow rate; perform computational fluid dynamics simulation for each set of structure-inlet pressure parameter combinations to obtain the steady-state outlet flow rate under the corresponding conditions.
[0089] (1) Establishing the fluid computation domain
[0090] The fluid computation domain of the internal flow channel is extracted based on the parametric 3D model of the solenoid valve.
[0091] For different combinations of structural parameters, the geometric model of the solenoid valve is modified according to the corresponding parameter values to generate the corresponding internal fluid domain.
[0092] (2) Perform grid division
[0093] The fluid computational domain is meshed, and the mesh is refined in areas such as valve orifices, steps, channels, local contraction regions, local expansion regions, and regions with significant flow changes.
[0094] Mesh independence analysis is used to determine the mesh size that meets the computational accuracy requirements, thereby reducing the impact of mesh size on the outlet flow calculation results.
[0095] (3) Set fluid properties and boundary conditions
[0096] The fluid density, dynamic viscosity, and other physical property parameters are set according to the actual working medium of the solenoid valve.
[0097] The inlet boundary is set as a pressure inlet with a pressure value of the corresponding inlet pressure Pj; the outlet boundary adopts a uniform fixed condition; and the flow channel wall is set as a non-slip wall.
[0098] For all samples, except for the inlet pressure and solenoid valve structural parameters, other fluid properties and boundary conditions are kept consistent to ensure comparability between different simulation samples.
[0099] (4) Perform steady-state flow field solution
[0100] Based on the internal flow state of the solenoid valve, an appropriate flow model is selected, and the pressure field, velocity field, and flow rate in the fluid domain are solved in steady-state iterative solution.
[0101] During the simulation, the residuals of the continuity equation, momentum equation, inlet mass flow rate, and outlet mass flow rate are monitored.
[0102] When the residual reaches the preset convergence condition and the outlet flow rate tends to stabilize within several consecutive iterations, the current simulation is determined to have reached a steady-state convergence state.
[0103] (5) Extract steady-state outlet flow rate
[0104] For the i-th solenoid valve structural sample and the j-th inlet pressure, extract the corresponding outlet mass flow rate or outlet volume flow rate.
[0105] When the CFD simulation output is mass flow rate, it is converted to volumetric flow rate based on the fluid density:
[0106]
[0107] in:
[0108] This represents the steady-state outlet mass flow rate of the i-th solenoid valve structural sample under the j-th inlet pressure;
[0109] Indicates fluid density;
[0110] This represents the corresponding steady-state outlet volumetric flow rate.
[0111] To reduce simulation iteration fluctuations, the average outlet flow rate over K consecutive iterations after convergence can be used as the final flow rate label.
[0112]
[0113] in, This represents the outlet flow rate of the i-th structural sample at the j-th inlet pressure and the k-th stable iteration step.
[0114] If the simulation fails to converge, or the inlet and outlet mass flow rates do not meet the mass conservation requirements, the corresponding data will be marked as anomalies, and the geometric model, mesh quality, boundary conditions, and solution settings will be re-examined.
[0115] The final dataset consists of structural parameters, inlet pressure, and steady-state outlet flow rate.
[0116]
[0117] That is: Solenoid valve structural parameters + inlet pressure → steady-state outlet flow rate.
[0118] Step 4: Construct fused features and partition the dataset
[0119] Based on the flow mechanism of solenoid valves, height differences, size ratios, effective flow areas, and pressure-structure interaction characteristics can also be constructed. For example:
[0120]
[0121]
[0122]
[0123]
[0124] Where n represents the number of flow channel holes, and φ represents the hole diameter. Indicates the effective circulation area.
[0125] To eliminate the influence of different parameter units and numerical ranges, the input features are standardized:
[0126]
[0127] in, and Let represent the mean and standard deviation of the j-th feature in the training set, respectively.
[0128] The processed data is divided into training, validation, and test sets. Preferably, the data is grouped according to the solenoid valve structural samples, ensuring that samples with different inlet pressures corresponding to the same structure reside only in the same subset of data.
[0129] Step 5: Establish and screen traffic prediction models
[0130] Using the fusion characteristics of the standardized structure corresponding to the inlet pressure as input and the CFD steady-state outlet flow rate as output, various regression prediction models are established, including:
[0131] Ridge, Lasso, and ElasticNet models;
[0132] SVR and Kernel Ridge models;
[0133] Random Forest, Extra Trees, and Gradient Boosting models;
[0134] MLP (Multilayer Perceptron) model.
[0135] The mapping relationships established by each model are represented as follows:
[0136]
[0137] Where Z represents the standardized fusion feature. This indicates the predicted flow rate.
[0138] The candidate models were evaluated using validation and test sets, with mean absolute error (MAE), root mean square error (RMSE), and coefficient of determination (R²) used as evaluation metrics.
[0139]
[0140]
[0141]
[0142] Based on a comprehensive comparison of the prediction accuracy, stability, and generalization ability of the models, the SVR model and the Ridge model were selected as the basic prediction branches of the weighted ensemble model.
[0143] Step 6: Construct the SVR-Ridge weighted ensemble model
[0144] The same fusion features were input into the SVR model and the Ridge model, respectively.
[0145] The SVR model is used to learn the complex nonlinear relationship between the structural parameters of the solenoid valve, the inlet pressure, and the outlet flow rate. Its prediction results are as follows:
[0146]
[0147] The Ridge model is used to learn the overall trend of outlet flow rate as a function of structural parameters and inlet pressure. Its prediction results are as follows:
[0148]
[0149] The prediction results from the two models are then weighted and fused:
[0150]
[0151] in:
[0152]
[0153]
[0154] The fusion weights can be determined through traversal search, cross-validation, or minimizing the prediction error on the validation set.
[0155]
[0156] By combining the nonlinear fitting capability of SVR with the trend stabilization capability of Ridge, the accuracy and generalization ability of traffic prediction results are improved.
[0157] Step 7: Output the flow prediction results
[0158] For the new solenoid valve structure to be predicted, its key structural parameters and target inlet pressure are obtained, and fusion features are constructed and standardized in the same way as in the training phase.
[0159] The fused features are then input into the SVR model and the Ridge model respectively, and the final traffic prediction value is calculated based on the determined fusion weights:
[0160]
[0161] Output the steady-state outlet flow rate of the target solenoid valve structure under the corresponding inlet pressure condition.
[0162] For the same solenoid valve structure, multiple sets of inlet pressures can be continuously input to obtain flow prediction results under different inlet pressures, and form an inlet pressure-outlet flow curve, providing a basis for solenoid valve structure design, performance evaluation and parameter optimization.
[0163] The specific embodiments described herein are merely illustrative of the spirit of the invention. Those skilled in the art to which this invention pertains may make various modifications or additions to the described specific embodiments or use similar methods to replace them, without departing from the spirit of the invention or exceeding the scope defined by the appended claims.
Claims
1. A method for intelligent prediction of flow rate of a solenoid valve based on the fusion of structural parameters and operating conditions, characterized in that, The process includes the following steps: Step 1: Determine the structural parameters and operating parameters of the solenoid valve; establish a parametric 3D model of the solenoid valve, and determine the key structural parameters affecting the outlet flow rate based on the internal flow channel structure and design requirements of the solenoid valve; Step 2: Generate parameter combinations for the inlet pressure corresponding to the structure, and generate multiple solenoid valve structural schemes based on the allowable value range of each structural parameter. The structural parameter combinations are generated through parameter sampling methods; Step 3: Conduct CFD simulation and extract steady-state outlet flow rate; Perform computational fluid dynamics simulation for the parameter combinations corresponding to the inlet pressure of each structure to obtain the steady-state outlet flow rate under the corresponding conditions; Step 4: Construct fusion features and divide the dataset. Based on the flow mechanism of the solenoid valve, construct interactive features of height difference, size ratio, effective flow area, and pressure corresponding to the structure; Step 5: Establish and screen flow prediction models; Using the standardized fusion features corresponding to the inlet pressure of the structure as input and the CFD steady-state outlet flow rate as output, establish multiple regression prediction models respectively; Step 6: Construct an SVR-Ridge weighted ensemble model, inputting the same fusion features into the SVR model and the Ridge model respectively; Step 7: Output the flow prediction results. For the new solenoid valve structure to be predicted, obtain its key structural parameters and target inlet pressure, and construct fusion features and perform standardization processing according to the same method as in the training phase.
2. The intelligent flow prediction method for solenoid valves based on the fusion of structural parameters and operating conditions according to claim 1, characterized in that, The structural parameters in step one include valve port size parameters, step height parameters, column height parameters, platform height parameters, and height difference parameters between different regions; the structural parameters also include the number and diameter of flow channel holes, flow channel width, length, and other dimensional parameters affecting the effective flow area; the structural parameter vector of the i-th solenoid valve prototype is defined as: in: This represents the step height of the i-th solenoid valve sample; ~ Indicates the column height parameters for different regions; Indicates the platform height parameter; Indicates valve port size, orifice structure size, or other structural parameters; Multiple inlet pressure conditions are set according to the actual working range of the solenoid valve. The inlet pressure set is represented as follows: in, Let represent the inlet pressure of the j-th group, and n represent the number of inlet pressure conditions.
3. The intelligent flow prediction method for solenoid valves based on the fusion of structural parameters and operating conditions according to claim 1, characterized in that, The combination of structural parameters in step two can be generated through manual design, full factorial experimental design, orthogonal experimental design, uniform design and Latin hypercube sampling. When generating the structural scheme, the geometric rationality of each parameter combination is checked to eliminate invalid structures with structural interference, closed flow channels, insufficient wall thickness or failure to meet processing constraints. For the i-th solenoid valve structural sample, multiple sets of inlet pressure conditions are set respectively. The i-th structural sample and the j-th inlet pressure together form a set of structural-operating condition parameter combinations: The corresponding model input feature vector is represented as: By combining each structural scheme with multiple inlet pressure conditions, a set of parameter combinations covering different structural forms and different inlet pressure conditions is formed.
4. The intelligent flow prediction method for solenoid valves based on the fusion of structural parameters and operating conditions according to claim 1, characterized in that, The simulation in step three includes establishing the fluid computational domain, mesh generation, setting fluid properties and boundary conditions, performing steady-state flow field solving, and extracting the steady-state outlet flow rate; for the i-th solenoid valve structural sample and the j-th inlet pressure, the corresponding outlet mass flow rate or outlet volumetric flow rate is extracted; when the CFD simulation output is mass flow rate, it is converted to volumetric flow rate according to the fluid density. in, This represents the steady-state outlet mass flow rate of the i-th solenoid valve structural sample under the j-th inlet pressure; Indicates fluid density; This represents the corresponding steady-state outlet volumetric flow rate; To reduce simulation iteration fluctuations, the average outlet flow rate over K consecutive iterations after convergence can be used as the final flow rate label. in, This represents the outlet flow rate of the i-th structural sample at the k-th stable iteration step under the j-th inlet pressure. The final dataset consists of structural parameters, inlet pressure, and steady-state outlet flow rate. .
5. The intelligent flow prediction method for solenoid valves based on the fusion of structural parameters and operating conditions according to claim 1, characterized in that, In step four, to eliminate the influence of different parameter units and numerical ranges, the input features are standardized: in, and Let represent the mean and standard deviation of the j-th feature in the training set, respectively; The processed data is divided into training set, validation set and test set; preferably, the data is grouped according to the solenoid valve structure sample so that samples of different inlet pressures corresponding to the same structure are located in only the same data subset.
6. The intelligent flow prediction method for solenoid valves based on the fusion of structural parameters and operating conditions according to claim 1, characterized in that, The mapping relationships established by each model in step five are expressed as follows: Where Z represents the standardized fusion feature. Indicates predicted flow; The candidate models were evaluated using validation and test sets, with mean absolute error (MAE), root mean square error (RMSE), and coefficient of determination (R²) used as evaluation metrics. Based on a comprehensive comparison of the prediction accuracy, stability, and generalization ability of the models, the SVR model and the Ridge model were selected as the basic prediction branches of the weighted ensemble model.
7. The intelligent flow prediction method for solenoid valves based on the fusion of structural parameters and operating conditions according to claim 1, characterized in that, In step six, the SVR model is used to learn the complex nonlinear relationship between the solenoid valve's structural parameters, inlet pressure, and outlet flow rate. Its prediction results are as follows: The Ridge model is used to learn the overall trend of outlet flow rate as a function of structural parameters and inlet pressure. Its prediction results are as follows: The prediction results from the two models are then weighted and fused: in: The fusion weights can be determined through traversal search, cross-validation, or minimizing the prediction error on the validation set. By combining the nonlinear fitting capability of SVR with the trend stabilization capability of Ridge, the accuracy and generalization ability of traffic prediction results are improved.
8. The intelligent flow prediction method for solenoid valves based on the fusion of structural parameters and operating conditions according to claim 1, characterized in that, In step seven, the fused features are input into the SVR model and the Ridge model respectively, and then the final traffic prediction value is calculated based on the determined fusion weights: Output the steady-state outlet flow rate of the target solenoid valve structure under the corresponding inlet pressure condition.
Citation Information
Patent Citations
Valve flow coefficient convenient calculation method based on CFD simulation
CN111767663A
