Jet flow centrifugal pump structure optimization method
Through simulation and intelligent algorithm optimization, the complex structure and experimental problems of jet centrifugal pumps are solved, and the efficient design of jet centrifugal pumps is realized and the cavitation margin is reduced, providing new structural optimization ideas.
Patent Information
- Application Number
- CN202411164202.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-23
- Publication Date
- 2024-12-03
AI Technical Summary
The structure of the jet centrifugal pump is complex, making it difficult to perform structural optimization analysis and experiments, and the experimental conditions are limited to the intuitive understanding of the flow field state in the pump, resulting in high design and manufacturing costs.
Using simulation experiments, multi-layer artificial neural network approximation model and intelligent algorithm multi-objective optimization calculations are used to improve the efficiency of the jet centrifugal pump and reduce the cavitation allowance through three-dimensional modeling and parameter optimization, and build backpropagation neural network and NSGA. -Ⅱ Fast non-dominant sorting genetic algorithm is designed and verified.
The optimized design of the structure of the jet centrifugal pump is realized, which reduces the design and manufacturing costs, improves efficiency and reduces the cavitation allowance, and provides a new structural optimization idea.
Smart Images

Figure CN119066811A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the technical field of centrifugal pump simulation analysis technology, and in particular to a method for optimizing the structure of a jet centrifugal pump. Background Art
[0002] The jet centrifugal pump is a self-circulating ejector with a venturi tube added to the inlet of an ordinary centrifugal pump, so that the ordinary centrifugal pump can achieve self-priming performance. Its working principle is that a pressurized liquid passes through the nozzle to create low pressure, thereby sucking external liquid into the pump inlet and discharging it through the pump body. Jet centrifugal pumps are widely used in many industries and fields such as pure water systems, rural household well water pumping, floor pipeline pressurization, car spraying, oil tank pumping, fountains, gardening irrigation, etc. due to their strong low-level water extraction capacity and good self-priming performance. With the increase in high-rise buildings, it has been more widely used in residents' lives. Many scholars at home and abroad have studied the various characteristics of jet centrifugal pumps by means of theoretical derivation, experiments or simulations. However, due to the limitations of experimental conditions, it is impossible to intuitively understand the state of the flow field in the centrifugal pump and the force on the pump wall. At the same time, the jet centrifugal pump has a complex structure, and the design and manufacturing costs of the impeller and pump body are high. It is difficult to manufacture a large number of components for structural optimization analysis and experiments. Summary of the invention
[0003] In view of the shortcomings of the prior art, one of the purposes of this application is to provide a method for optimizing the structure of a jet centrifugal pump, which utilizes the internal flow state of the jet centrifugal pump in a simulation experiment to transform actual engineering problems into mathematical and simulation problems, and utilizes a multi-layer artificial neural network approximate model to build and intelligent algorithm multi-objective optimization calculations to achieve the advantages of designing and verifying the pump body design.
[0004] The above-mentioned purpose of the present application is achieved through the following technical solutions:
[0005] A method for optimizing the structure of a jet centrifugal pump is provided. A simulation model is established. A three-dimensional modeling of the pump is performed using three-dimensional software. Parameters are selected. The ejector and impeller of the pump are structurally optimized. The efficiency η and NPSHr of the jet centrifugal pump under the design condition are taken as optimization targets. The efficiency η is improved and the NPSHr is reduced. Parameters η and NPSHr are obtained. A sample library is established. Multiple groups of parameters η and NPSHr are formed into a sample library. A multi-layer artificial neural network is constructed. A back propagation neural network with two hidden layers is used on the traditional BP neural network. A sigmoid function and a linear function are set as activation functions in the hidden layer and the output layer respectively. Six neurons are set in each hidden layer and one neuron is set in the output layer. The sample library data is divided into three groups: a training set is used for neural network training, a validation set is used for model validation, and a test set is used for model testing. After the model training is completed, the mean square error and the average relative error value between the neural network approximate model fitting value and the CFD calculated value are used to evaluate the fitting effect. Solution: The non-dominated optimal solution is obtained by the NSGA-Ⅱ fast non-dominated sorting genetic algorithm.
[0006] In a preferred example, the present application can be further configured as follows: Optimization parameter selection: Select the structural parameters of the ejector and the structural parameters of the centrifugal impeller to optimize the structure of the jet centrifugal pump, take the efficiency η and NPSHr of the jet centrifugal pump under the design conditions as the optimization targets, improve the efficiency η and reduce the NPSHr, and the optimization target efficiency η and NPSHr formula are as follows: Where P u is the effective power; P in is the shaft power; ρ is the density of the pumped fluid; g is the acceleration of gravity; H is the pump head; Q is the pump flow rate; P is the pump inlet pressure; P min is the minimum blade pressure; u is the inlet fluid velocity.
[0007] In a preferred example, the present application can be further configured as follows: Sample library establishment: Using the Latin hypercube method, the injector nozzle diameter D1, throat diameter D2, throat-to-mouth distance L1 and diffuser diffusion angle α, as well as the impeller inlet diameter D3, blade inlet angle β1, impeller outlet diameter D4 and blade outlet angle β2 are designed in the optimization variable region. By remodeling and numerically simulating each scheme, its efficiency η and NPSHr value are obtained.
[0008] In a preferred example, the present application can be further configured as follows: the mathematical expression of the multi-objective optimization problem of the jet centrifugal pump can be described by the following equation:
[0009] and b i Represents the upper and lower limits of the optimization variable value interval.
[0011] In a preferred example, the present application can be further configured as follows: a non-dominated optimal solution is obtained through the NSGA-Ⅱ fast non-dominated sorting genetic algorithm, a suitable optimal solution is selected from the non-dominated solutions in combination with the actual engineering situation, a hydraulic model is established based on this basic data for numerical flow field simulation, and the accuracy of the optimization results is verified. If the error between the simulation verification result and the algorithm solution result is within the set error threshold, the optimization result is considered to be valid, otherwise it is not valid.
[0012] In a preferred example, the present application can be further configured as follows: NSGA-Ⅱ fast non-dominated sorting genetic algorithm includes the following steps:
[0013] a. Initialize the population size to 200, the maximum number of iterations to 3000, the crossover probability to pc0.9, and the counter to gen=1;
[0014] b. Generate 200 individuals that meet the constraints, calculate the values of each objective function of the individuals, and perform non-dominated sorting on the individuals to generate the parent population, calculate the crowding degree for each individual, and solve the selection problem within the same non-dominated level.
[0015] c. Use tournament selection, multi-point crossover, and real-value mutation to perform genetic operations on the parent population to generate a temporary population.
[0016] d. Determine whether the individuals in the temporary population meet the constraints, find out the individuals that do not meet the constraints, generate an equal number of individuals that meet the constraints to replace them, obtain the offspring population, and calculate the objective function values of the individuals in the offspring population.
[0017] e. The parent and child populations are merged and non-dominated sorting is performed. The top 200 individuals are used as the parent population, gen = gen + 1.
[0018] f. If gen≤3000, go to step c, otherwise end.
[0019] In a preferred example, the present application can be further configured as follows: in the establishment of the simulation model, a test bench is built, and the jet centrifugal pump test loop is connected; the head, efficiency and cavitation margin of the test jet centrifugal pump under different working conditions are measured and recorded; the jet centrifugal pump is simulated and analyzed: simulation pre-processing: a three-dimensional model is established according to the basic size parameters of the jet centrifugal pump model, the flow channel model is extracted and meshed; simulation solution: a solver and a turbulence model are selected, fluid parameters and boundary conditions are set, a mesh interface is established, convergence criteria and iterative calculation steps are set, and simulation solution is performed; simulation post-processing: after the calculation results converge, the simulation post-processing software is used to display the fluid velocity, pressure and turbulent kinetic energy cloud map of the specified plane, and the structural parameters that may have a key impact on the efficiency and effective cavitation margin are analyzed based on the internal flow field distribution. The simulation results are statistically compared with the experimental data. If the error between the data exceeds the allowable error range, the above simulation pre-processing and simulation solution steps are repeated, and the mesh division and the setting of the solution conditions are corrected until the error reaches the allowable range.
[0020] The present application has the following advantages: by utilizing the internal flow state of the jet centrifugal pump in the simulation experiment, practical engineering problems are converted into mathematical and simulation problems; by using a multi-layer artificial neural network approximate model and intelligent algorithm multi-objective optimization calculation, a shortcut is found for the optimization design of dimensional parameters, providing a new idea for the structural optimization of the jet centrifugal pump. BRIEF DESCRIPTION OF THE DRAWINGS
[0021] Figure 1 It is a flow chart of the structural optimization method of the jet centrifugal pump of the present application.
[0022] Figure 2 It is a schematic diagram of the optimized variable structure of the nozzle and nozzle of the jet centrifugal pump of the present application.
[0023] Figure 3 It is a schematic diagram of the optimized variable structure of the impeller of the jet centrifugal pump of the present application.
[0024] Figure 4 It is a schematic diagram of the blade optimization variable structure of the impeller of this application.
[0025] Figure 5 This is a schematic diagram of the back propagation multi-layer neural network structure of this application.
[0026] Figure 6 It is the scatter plot matrix of the sample space of the variables optimized in this application.
[0027] Figure 7 This is the flow chart of the NSGA-Ⅱ fast non-dominated sorting genetic algorithm of this application. DETAILED DESCRIPTION
[0028] The following is combined with Figure 1-7This application is described in further detail.
[0029] The present application discloses a method for optimizing the structure of a jet centrifugal pump. A simulation model is established. A three-dimensional modeling of the pump is performed using three-dimensional software. Parameters are selected. The structure of the ejector and impeller of the pump is optimized. The efficiency η and the NPSHr of the jet centrifugal pump under the design working condition are taken as optimization targets. The efficiency η is improved and the NPSHr is reduced. The parameters η and NPSHr are obtained. A sample library is established. Multiple groups of parameters η and NPSHr are formed into a sample library. A multi-layer artificial neural network is constructed: a back propagation neural network with two hidden layers is used on the traditional BP neural network. A sigmoid function and a linear function are respectively set as activation functions in the hidden layer and the output layer. Six neurons are set in each hidden layer and one neuron is set in the output layer. The sample library data is divided into three groups: a training set is used for neural network training, a validation set is used for model validation, and a test set is used for model testing. After the model training is completed, the mean square error and the average relative error value of the neural network approximate model fitting value and the CFD calculated value are used to evaluate the fitting effect. Solution: The non-dominated optimal solution is obtained by the NSGA-Ⅱ fast non-dominated sorting genetic algorithm.
[0030] Specifically, the following steps are included:
[0031] 1. Experimental test on the head, efficiency and NPSH of the jet centrifugal pump:
[0032] (1) Build a test bench and connect the test circuit of the jet centrifugal pump;
[0033] (2) Test the head, efficiency and NPSH of the jet centrifugal pump under different working conditions and record them.
[0034] 2. Simulation analysis of jet centrifugal pump:
[0035] (1) Pre-processing of simulation: Establish a three-dimensional model based on the basic size parameters of the jet centrifugal pump model, extract the flow channel model and perform mesh division;
[0036] (2) Simulation solution: Select the solver and turbulence model. Set the fluid parameters and boundary conditions, and establish the mesh interface. Set the convergence standard and the number of iterative calculation steps, and perform simulation solution.
[0037] (3) Simulation post-processing: After the calculation results converge, use the simulation post-processing software to display the velocity, pressure and turbulent kinetic energy cloud map of the specified plane fluid, and analyze the structural parameters that may have a key impact on efficiency and effective cavitation margin based on the internal flow field distribution. Statistical simulation results are compared with experimental data. If the error between the data is within the allowable error range, the various conditions and meshing methods used in the simulation are considered reliable. If the error between the data exceeds the allowable error range, repeat the above simulation pre-processing and simulation solution steps, correct the meshing and solution condition settings, until the error is within the allowable range.
[0038] 3. Construct a structural optimization approximate model:
[0039] (1) Optimization parameter selection: The ejector and centrifugal impeller are the core flow components of the jet centrifugal pump. The key structural parameters of these two parts are selected to optimize the structure of the jet centrifugal pump. The efficiency η and NPSHr of the jet centrifugal pump under the design conditions are taken as the main optimization targets. The efficiency η and the NPSHr are improved as much as possible while taking into account other performance parameters. The ejector nozzle diameter D1, throat diameter D2, throat-to-mouth distance L1 and diffuser diffusion angle α, as well as the impeller inlet diameter D3, blade inlet angle β1, impeller outlet diameter D4 and blade outlet angle β2 are taken as the main optimization variables, and the upper and lower bounds of the variables are defined. The formulas for optimizing the target efficiency η and NPSHr are as follows:
[0040]
[0041]
[0042] Where P u is the effective power; P in is the shaft power; ρ is the density of the pumped fluid; g is the acceleration of gravity; H is the pump head; Q is the pump flow rate; P is the pump inlet pressure; P min is the minimum blade pressure; u is the inlet fluid velocity.
[0043] (2) Sample library establishment: The Latin hypercube method is used to design 160 sets of initial sample data schemes for the injector nozzle diameter D1, throat diameter D2, throat-to-nozzle distance L1, diffuser diffusion angle α, as well as the impeller inlet diameter D3, blade inlet angle β1, impeller outlet diameter D4, and blade outlet angle β2 in the optimization variable region. By remodeling and numerically simulating each scheme, its efficiency η and NPSHr value are obtained.
[0044] (3) Construction of multi-layer artificial neural network: Due to the complex internal flow characteristics of the jet centrifugal pump, the relationship between the objective function and the design parameters shows strong nonlinearity. In order to achieve better fitting effect, improvements are made on the traditional BP neural network, and a back propagation neural network with two hidden layers is adopted. The sigmoid function and the linear function are set as activation functions in the hidden layer and the output layer respectively. Each hidden layer is set with 6 neurons and the output layer has 1 neuron. The sample data set obtained above is divided into 3 groups: training set (80%, a total of 128 groups, used for neural network training), validation set (10%, a total of 16 groups, used for model verification) and test set (10%, a total of 16 groups, used for model testing). After the model training is completed, the fitting effect is evaluated by the mean square error and average relative error value between the neural network approximate model fitting value and the CFD calculated value.
[0045] 4. Multi-objective optimization calculation:
[0046] (1) The mathematical expression of the multi-objective optimization problem of the jet centrifugal pump can be described by the following equation:
[0047]
[0048] where x i represents the optimization variable, a i and b i Represents the upper and lower limits of the optimization variable value interval.
[0049] (2) Intelligent algorithms commonly used for pump optimization include simulated annealing algorithm (SA), particle swarm algorithm (PSO) and genetic algorithm (NSGA). The structural optimization method for jet centrifugal pumps adopts NSGA-Ⅱ fast non-dominated sorting genetic algorithm. Compared with traditional genetic algorithms, this algorithm introduces elite retention strategy and crowding operator. The initial population, genetic generation and crossover probability are set, and the non-dominated optimal solution is obtained after a complete genetic iterative optimization process.
[0050] Combined with the actual engineering situation, the appropriate optimal solution is selected from the non-dominated solutions, and the hydraulic model is established based on this basic data to perform numerical flow field simulation to verify the accuracy of the optimization results. If the error between the simulation verification result and the algorithm solution result is within the set error threshold, the optimization result is considered to be established, otherwise it is not established.
[0051] In this embodiment, the present application takes a certain type of jet centrifugal pump as the research object. Its overall design hydraulic parameters are: rated flow Q = 2.964m 3 / h, rated head H = 28.51m, speed n = 2849r / min, the variable geometric parameters involved in its structural optimization are shown in the table below.
[0052]
[0053] This embodiment provides a method for optimizing the structure of a jet centrifugal pump based on Fluent. Figure 1 As shown, the following steps are included:
[0054] 1. Build a test bench, connect the test circuit of the jet centrifugal pump, and measure and record the head, efficiency and NPSH value of the jet centrifugal pump under different working conditions.
[0055] 2. Use Fluent software to simulate and analyze the jet centrifugal pump:
[0056] (1) According to the basic size parameters of the jet centrifugal pump model, a three-dimensional model of the jet centrifugal pump is established using Solidworks and output as an igs file;
[0057] (2) The igs file is input into SpaceClaim software for simulation pre-processing, and structures irrelevant to hydraulic performance, such as injection holes and inspection holes, are removed. The connection between the shaft and the impeller is simplified and residual gaps in the assembly are eliminated. The flow structure is retained to the greatest extent. The flow channel file is extracted using the volume extraction function to establish a flow channel model, and the calculation domain is divided into two parts, namely, rotating and stationary parts, and a plane is created for the interface between the dynamic and static areas.
[0058] (3) Import the flow channel model into Fluent Meshing software to divide the mesh. Select the polyhedral mesh, set the maximum and minimum mesh sizes, mesh growth ratio, and number of boundary layer mesh layers. Encrypt the areas where the flow changes dramatically. Set the names of the inlets, outlets, and walls of the model, and output them as mesh files;
[0059] (4) Use Fluent Solution software to perform simulation analysis, select the solver mode and turbulence model. Set the fluid parameters such as fluid density and viscosity, and the unit area conditions including the rotation method, rotation direction and speed. Set the boundary conditions such as inlet flow, outlet pressure and wall conditions, and establish the grid interface. Select the solution method, define the required variable report, set the convergence conditions and iterative calculation steps, and perform initialization and simulation solution.
[0060] (5) After the simulation calculation is completed, the Fluent processing tool is used to check the flow distribution state of the flow field in various locations in the jet centrifugal pump. The simulation results are statistically compared with the experimental data. If the error between the data is within the allowable error range, the various conditions and grid division methods used in the simulation are considered reliable. If the error between the data exceeds the allowable error range, repeat the above steps (2) to (4) to correct the grid division and the setting of the solution conditions until the error is within the allowable range. The allowable error range is within ±5% based on the experimental data.
[0061] 3. Construct a structural optimization approximate model:
[0062] (1) Taking the efficiency η and NPSHr of the jet centrifugal pump under the design condition as the main optimization targets, the efficiency η is improved and the NPSHr is reduced as much as possible while taking into account other performance parameters. The nozzle diameter D1, throat diameter D2, throat-to-nozzle distance L1 and diffuser diffusion angle α of the ejector, as well as the impeller inlet diameter D3, blade inlet angle β1, impeller outlet diameter D4 and blade outlet angle β2 are the main optimization variables. The structural diagram is shown in the figure. Figure 2-4 The value range of the variable is shown in the following table:
[0063]
[0064] The formula for optimizing target efficiency η and NPSHr is as follows:
[0065]
[0066]
[0067] Where P u is the effective power; P in is the shaft power; ρ is the density of the pumped fluid; g is the acceleration of gravity; H is the pump head; Q is the pump flow rate; P is the pump inlet pressure; P min is the minimum blade pressure; u is the inlet fluid velocity.
[0068] (2) Using the Latin hypercube method, 160 sets of initial sample data schemes are designed for the ejector nozzle diameter D1, throat diameter D2, throat-to-nozzle distance L1, diffuser diffusion angle α, as well as the impeller inlet diameter D3, blade inlet angle β1, impeller outlet diameter D4, and blade outlet angle β2 in the optimization variable region, as shown in the following table:
[0069]
[0070]
[0071] Figure 6 The scatter plot matrix of all samples is displayed, which contains two basic graphics: histogram and scatter plot. The histogram located on the diagonal shows the distribution of each variable, while the scatter plots above and below the diagonal show the relationship between the variables. It can be seen that the sample space has sufficient data and the distribution is particularly uniform. By remodeling each solution and Fluent numerical simulation, the efficiency η and NPSHr values of all samples are obtained.
[0072] (3) Using MATLAB, we built a back propagation neural network with two hidden layers as an approximate model. The basic structure of the model is as follows: Figure 5 As shown. The sigmoid function and the linear function are set as activation functions in the hidden layer and the output layer respectively. Each hidden layer is set with 6 neurons and the output layer has 1 neuron. The sample data set obtained above is divided into 3 groups: training set (80%, a total of 128 groups, for neural network training), validation set (10%, a total of 16 groups, for model verification) and test set (10%, a total of 16 groups, for model testing). After the model training is completed, the mean square error and mean relative error value of the neural network approximate model fitting value and the Fluent calculated value are compared to ensure that the fitting effect is good.
[0073] 4. Multi-objective optimization calculation:
[0074] (1) Establish a multivariable and multi-objective mathematical model as follows:
[0075]
[0076] (2) NSGA-Ⅱ fast non-dominated sorting genetic algorithm is used in MATLAB for optimization calculation. The process is as follows: Figure 7 As shown, follow these steps to implement the algorithm:
[0077] a. Initialize the population size to 200, the maximum number of iterations to 3000, and the crossover probability to p c 0.9 and counter
[0078] gen=1;
[0079] b. Generate 200 individuals that meet the constraints, calculate the values of each objective function of the individuals, and perform non-dominated sorting on the individuals to generate the parent population. Calculate the crowding degree for each individual to solve the selection problem within the same non-dominated level.
[0080] c. Use tournament selection, multi-point crossover, and real-valued mutation to perform genetic operations on the parent population to generate a temporary population.
[0081] d. Determine whether the individuals in the temporary population meet the constraints, find out the individuals that do not meet the constraints, generate an equal number of individuals that meet the constraints to replace them, obtain the offspring population, and calculate the objective function values of the individuals in the offspring population.
[0082] e. The parent and child populations are merged and non-dominated sorting is performed. The top 200 individuals are taken as the parent population, gen = gen + 1.
[0083] f. If gen≤3000, go to step c, otherwise end.
[0084] After 3000 generations of genetic iterative optimization, the non-dominated optimal solution was obtained.
[0085] Combined with the actual engineering situation, the appropriate optimal solution is selected from the non-dominated solutions, and the hydraulic model is established based on this basic data to perform numerical flow field simulation to verify the accuracy of the optimization results. If the error between the simulation verification result and the algorithm solution result is within ±5%, the optimization result is considered to be established.
[0086] The implementation principle of this embodiment is: combining experiments with simulations, verifying the correctness of simulation results through experimental data, using simulation data to simulate the internal flow state of the jet centrifugal pump in the experiment, converting actual engineering problems into mathematical and simulation problems, and using multi-layer artificial neural network approximate model construction and intelligent algorithm multi-objective optimization calculation to find a shortcut for dimensional parameter optimization design, providing new ideas for the structural optimization of jet centrifugal pumps.
[0087] The embodiments of this specific implementation method are all preferred embodiments of the present application, and are not intended to limit the protection scope of the present application. Therefore, all equivalent changes made based on the structure, shape, and principle of the present application should be included in the protection scope of the present application.
Claims
1. A method for optimizing the structure of a jet centrifugal pump, characterized in that: The simulation model is established, and the pump is 3D modeled using 3D software. Parameters are selected, and the structure of the ejector and impeller of the pump is optimized. The efficiency η and NPSHr of the jet centrifugal pump under the design conditions are taken as the optimization goals, the efficiency η is improved and the NPSHr is reduced. The parameters η and NPSHr are obtained, and a sample library is established. Multiple groups of parameters η and NPSHr form a sample library. A multi-layer artificial neural network is built: a back propagation neural network with 2 hidden layers is used on the BP neural network. The sigmoid function and the linear function are set as activation functions in the hidden layer and the output layer respectively. Each hidden layer has 6 neurons and the output layer has 1 neuron. The sample library data is divided into 3 groups: the training set is used for neural network training, the validation set is used for model validation, and the test set is used for model testing. After the model training is completed, the mean square error and the average relative error value of the neural network approximate model fitting value and the CFD calculated value are used to evaluate the fitting effect. Solution: The non-dominated optimal solution is obtained by the NSGA-Ⅱ fast non-dominated sorting genetic algorithm.
2. The method for optimizing the structure of a jet centrifugal pump according to claim 1, characterized in that: In parameter selection: select the structural parameters of the ejector and the structural parameters of the centrifugal impeller to optimize the structure of the jet centrifugal pump, take the efficiency η and NPSHr of the jet centrifugal pump under the design conditions as the optimization target, improve the efficiency η and reduce the NPSHr, the optimization target efficiency η and NPSHr formula are as follows: Where P u is the effective power; P in is the shaft power; ρ is the density of the pumped fluid; g is the acceleration due to gravity; H is the pump head; Q is the pump flow rate; P is the pump inlet pressure; P min is the minimum blade pressure; u is the inlet fluid velocity.
3. The method for optimizing the structure of a jet centrifugal pump according to claim 1, characterized in that: In the establishment of the sample library: the Latin hypercube method is used to design X groups of initial sample data schemes for the injector nozzle diameter D1, throat diameter D2, throat-to-nozzle distance L1 and diffuser diffusion angle α, as well as the impeller inlet diameter D3, blade inlet angle β1, impeller outlet diameter D4 and blade outlet angle β2 in the optimization variable region. By remodeling and numerically simulating each scheme, its efficiency η and NPSHr value are obtained.
4. The method for optimizing the structure of a jet centrifugal pump according to claim 3, characterized in that: The mathematical expression of the multi-objective optimization problem of the jet centrifugal pump can be described by the following equation: where x i represents the optimization variable, a i and b i Represents the upper and lower limits of the optimization variable value interval.
5. The method for optimizing the structure of a jet centrifugal pump according to claim 1, characterized in that: In the solution, the non-dominated optimal solution is obtained through the NSGA-Ⅱ fast non-dominated sorting genetic algorithm. The appropriate optimal solution is selected from the non-dominated solutions in combination with the actual engineering situation. Based on this basic data, a hydraulic model is established to carry out numerical flow field simulation and verify the accuracy of the optimization results. If the error between the simulation verification result and the algorithm solution result is within the set error threshold, the optimization result is considered to be valid, otherwise it is not valid.
6. A method for optimizing the structure of a jet centrifugal pump according to claim 1 or 5, characterized in that: The NSGA-Ⅱ fast non-dominated sorting genetic algorithm includes the following steps: a. Initialize the population size to 200, the maximum number of iterations to 3000, the crossover probability to pc0.9, and the counter to gen=1; b. Generate 200 individuals that meet the constraints, calculate the values of each objective function of the individuals, perform non-dominated sorting on the individuals, generate the parent population, calculate the crowding degree for each individual, and solve the selection problem within the same non-dominated level; c. Use tournament selection, multi-point crossover, and real-value mutation to perform genetic operations on the parent population to generate a temporary population; d. Determine whether the individuals in the temporary population meet the constraints, find out the individuals that do not meet the constraints, generate an equal number of individuals that meet the constraints to replace them, obtain the offspring population, and calculate the objective function values of the individuals in the offspring population; e. The parent and child populations are merged and non-dominated sorting is performed. The top 200 individuals are used as the parent population, gen=gen+1. f. If gen≤3000, go to step c, otherwise end.
7. The method for optimizing the structure of a jet centrifugal pump according to claim 1, characterized in that: In the establishment of the simulation model, a test bench is built to connect the test loop of the jet centrifugal pump; the head, efficiency and NPSH of the jet centrifugal pump under different working conditions are measured and recorded; the jet centrifugal pump is simulated and analyzed: Simulation pre-processing: a three-dimensional model is established according to the basic size parameters of the jet centrifugal pump model, the flow channel model is extracted and meshed; simulation solution: the solver and turbulence model are selected, the fluid parameters and boundary conditions are set, the mesh interface is established, the convergence standard and the number of iterative calculation steps are set, and the simulation solution is performed; simulation post-processing: after the calculation results converge, the simulation post-processing software is used to display the fluid velocity, pressure and turbulent kinetic energy cloud map of the specified plane, and the structural parameters that may have a key impact on the efficiency and effective NPSH are analyzed based on the internal flow field distribution. The simulation results are statistically compared with the experimental data. If the error between the data exceeds the allowable error range, the above simulation pre-processing and simulation solution steps are repeated to correct the mesh division and the setting of the solution conditions until the error reaches the allowable range.
Citation Information
Cited By
Ejector optimization method and system based on improved orthogonal test method
CN119885913A
Centrifugal water pump simulation modeling method and model based on fluid network
CN120764082A
