A multi-objective optimization method for air compressor impeller based on turbulent pseudo-order regulation

Abstract

The application provides a kind of air compressor impeller multi-objective optimization method based on turbulent flow quasi-order regulation, it is related to the field of air compressor for fuel cell vehicle, and the optimization target of the optimization method is isentropic efficiency and maximum turbulent kinetic energy, and the optimization parameter is blade inlet installation angle β 1A , blade tip clearance C c , blade outlet width b c , blade outlet installation angle β 2A . The optimization method uses Latin hypercube sampling method to obtain several groups of data points of impeller structure, adjusts the impeller parameters according to the data points, and establishes a three-dimensional model. The optimization method carries out CFD simulation analysis on the three-dimensional model, obtains the isentropic efficiency and the maximum turbulent kinetic energy in the impeller corresponding to each air compressor, sets the optimization parameter data points and the corresponding isentropic efficiency and maximum turbulent kinetic energy as a training sample, and constructs a sample set containing multiple training samples. An agent model is constructed by fitting a neural network using the sample set, a multi-objective genetic algorithm is used to perform multi-objective optimization on the agent model, a Pareto frontier solution set is obtained, and the required optimal solution is selected according to the actual requirements.

Description

Technical Field

[0001] This invention relates to the field of air compressor technology for fuel cell vehicles, and in particular to a multi-objective optimization method for air compressor impellers based on turbulent pseudo-sequential control. Background Technology

[0002] In recent years, the government's emphasis on and strong support for the hydrogen energy industry has driven the rapid development of fuel cell vehicles. As a key component of fuel cell vehicles, the air compressor's performance directly affects critical indicators such as fuel cell system efficiency and noise levels. The main component of the air compressor is the impeller, and its performance significantly impacts the overall performance of the air compressor. Optimizing the impeller design can significantly improve the overall performance of the air compressor. However, current impeller optimization designs focus primarily on the relationship between impeller structural parameters and air compressor performance, neglecting to consider how these parameters can also affect the turbulent flow pattern within the flow field. Summary of the Invention

[0003] In view of this, the purpose of this invention is to provide a multi-objective optimization method for air compressor impellers based on turbulent quasi-sequential control. This method analyzes the structural parameters of the air compressor impeller from the perspective of turbulent quasi-sequential control, constructs a surrogate model of the impeller structural parameters, isentropic efficiency, and maximum turbulent kinetic energy through neural network fitting, and combines this with a multi-objective genetic algorithm to perform multi-objective optimization. This method can effectively optimize the structural parameters of the air compressor impeller, improve isentropic efficiency, and reduce turbulent kinetic energy.

[0004] To achieve the above objectives, the present invention adopts the following technical solution: a multi-objective optimization method for air compressor impellers based on turbulent pseudo-sequence control, comprising the following steps:

[0005] Step S1: Construct an initial model of the air compressor, perform CFD simulation analysis on the initial model of the air compressor, and obtain the flow field of the initial model;

[0006] Step S2: Extract the turbulent pseudo-order structure in the flow field using the Q criterion, analyze the influence of the turbulent pseudo-order structure on the isentropic efficiency and the maximum turbulent kinetic energy, and determine the optimization parameters and their value ranges.

[0007] Step S3: Use the Latin hypercube sampling method to obtain several data points for each optimization parameter, and build a three-dimensional model based on all data points; perform CFD simulation analysis on all three-dimensional models to obtain the corresponding isentropic efficiency and maximum turbulent kinetic energy, and set the optimization parameter data points and the corresponding isentropic efficiency and maximum turbulent kinetic energy as a training sample to obtain a sample set containing multiple training samples.

[0008] Step S4: Optimize the parameters of the impeller geometry model. The optimization includes the following steps: S41: Based on the sample set, use a neural network to fit and construct a surrogate model; S42: Combine a multi-objective genetic algorithm to perform multi-objective optimization on the surrogate model, obtain the Pareto front solution set, and select the optimal solution according to actual needs.

[0009] In a preferred embodiment, in step S2, the Q criterion is defined as follows:

[0010]

[0011] In the formula, Ω is the rotation tensor; S is the strain rate tensor; and Q is the vorticity.

[0012] In step S2, the turbulent pseudo-order structure refers to the ordered structure in the irregular flow of the flow field, which manifests in the air compressor as separation vortex, leakage vortex, and wake vortex.

[0013] In a preferred embodiment, in step S2, the formula for calculating the isentropic efficiency of the air compressor is as follows:

[0014]

[0015] In the formula, T 1c * The measured total intake air temperature of the air compressor is expressed in Kelvin (K); Telvin (T). 2c * The measured air output from the air compressor is expressed in K; π c * is the air compressor pressure ratio; k is the ideal gas adiabatic index.

[0016] In a preferred embodiment, in step S2, the turbulent kinetic energy k tke It is an indicator for measuring the development or decline of turbulence, and its calculation formula is as follows:

[0017]

[0018] In the formula, U is the average velocity in m / s; I is the turbulence intensity.

[0019] In step S2, the determined optimization parameter is the blade inlet installation angle β. 1A Blade tip clearance C c Blade exit width b c Blade outlet installation angle β 2A .

[0020] In a preferred embodiment, in step S3, the Latin hypercube sampling method provides a value for each interval (1, 1 / n), (1 / n, 2 / n)...(1-1 / n, 1) and randomly arranges them within n sets of data for x impeller structures.

[0021] In a preferred embodiment, in step S3, the generated sample set includes all combinations of optimized parameters and their corresponding isentropic efficiency and maximum turbulent kinetic energy, wherein the combination of optimized parameters is the blade inlet installation angle β. 1A Blade tip clearance C c Blade exit width b c and blade outlet installation angle β 2A The combination of parameters.

[0022] In a preferred embodiment, in step S41, the sample set is fitted and trained using a neural network toolbox, as follows: S411, with the optimized parameter combination as the independent variable and isentropic efficiency and maximum turbulent kinetic energy as the dependent variables, the data is imported into the neural network toolbox; S412, the proportions of training data and test data are set, and the number of hidden neurons is set; S413, a surrogate model of the air compressor's optimized parameters, isentropic efficiency, and maximum turbulent kinetic energy is trained, and the slope R of the regression line is used as the accuracy evaluation index of the surrogate model.

[0023] In a preferred embodiment, in S413, the accuracy of the surrogate model obtained by the neural network fitting training is determined by the slope R of the regression line; the closer R is to 1, the better the training effect.

[0024] In a preferred embodiment, in step S42, a multi-objective genetic algorithm is used for optimization, and the optimization steps are as follows: S421, adjust the blade inlet installation angle β. 1A Blade tip clearance C c Blade exit width b c and blade outlet installation angle β 2A As the optimization variable, the surrogate model serves as the fitness function; S422, set the constraints on the optimization variables; S423, set the population size, optimal front-end individual coefficient, number of iterations, fitness function value deviation, and the probabilities of crossover and mutation; S424, based on the optimization parameters and the constraints of the fitness function, establish a multi-objective optimization mathematical model; S425, use a multi-objective genetic algorithm to optimize the mathematical model and select the optimal solution from the obtained Pareto front.

[0025] In a preferred embodiment, in S424, the mathematical model is as follows:

[0026]

[0027] In the formula, f1(x) is the isentropic efficiency of the air compressor; f2(x) is the maximum turbulent kinetic energy of the air compressor; β 1A Installation angle for blade inlet; C c b is the blade tip gap; c β is the blade exit width; 2A For the blade outlet installation angle; β 1A1 and β 1A2 These represent the minimum and maximum values ​​of the blade inlet installation angle, C. c1 and C c2 These represent the minimum and maximum values ​​of the blade tip clearance, respectively. c1 and b c2 These represent the minimum and maximum values ​​of the blade exit width, β. 2A1 and β 2A2 These represent the minimum and maximum values ​​of the blade outlet installation angle, respectively.

[0028] Compared with the prior art, the present invention has the following beneficial effects:

[0029] 1. This invention focuses on the main vortex structures in the disordered flow field by adjusting and controlling the pseudo-ordered structure of turbulence. It uses turbulent kinetic energy to characterize the development or decay of turbulence. Compared with optimization methods that only focus on isentropic efficiency and pressure ratio, this invention focuses on isentropic efficiency and the energy of pseudo-ordered structures in the flow field, making the optimization perspective more reasonable.

[0030] 2. The optimization objectives are isentropic efficiency and maximum turbulent kinetic energy. The optimization parameters are blade inlet installation angle β1A, blade tip clearance Cc, blade outlet width bc, and blade outlet installation angle β2A. By constructing a surrogate model and combining it with a multi-objective genetic algorithm, the optimization efficiency is higher when both optimization objectives are addressed simultaneously. Attached Figure Description

[0031] Figure 1 This is a flowchart illustrating the optimized design of a preferred embodiment of the present invention.

[0032] Figure 2 Turbulent kinetic energy at 50% blade height under different tip clearances in preferred embodiments of the present invention: (a) tip clearance = 0.2 mm; (b) tip clearance = 0.3 mm; (c) tip clearance = 0.4 mm;

[0033] Figure 3 Comparison of vorticity at 90% flow direction position under different tip clearances in the preferred embodiment of the present invention: (a) tip clearance = 0.2 mm; (b) tip clearance = 0.3 mm; (c) tip clearance = 0.4 mm;

[0034] Figure 4 Error analysis diagrams of the neural network surrogate model in a preferred embodiment of the present invention: (a) isentropic efficiency; (b) maximum turbulent kinetic energy;

[0035] Figure 5 The Pareto front is optimized for isentropic efficiency and maximum turbulent kinetic energy in a preferred embodiment of the present invention. Detailed Implementation

[0036] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0037] It should be noted that the following detailed descriptions are illustrative and intended to provide further explanation of this application. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.

[0038] It should be noted that the terminology used herein is for the purpose of describing particular implementations only and is not intended to limit the exemplary implementations according to this application; as used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise; furthermore, it should be understood that when the terms “comprising” and / or “including” are used in this specification, they indicate the presence of features, steps, operations, devices, components and / or combinations thereof.

[0039] like Figure 1-5 As shown in the optimized design flowchart, this invention provides a multi-objective optimization method for air compressor impellers based on turbulent pseudo-sequential control, characterized by the following steps:

[0040] S1. Construct an initial model of the air compressor, perform CFD simulation analysis on the initial model, and obtain the flow field of the initial model;

[0041] S2. Extract the turbulent pseudo-order structure in the flow field using the Q criterion, analyze the influence of the turbulent pseudo-order structure on the isentropic efficiency and the maximum turbulent kinetic energy, and determine the optimization parameters and their value range.

[0042] S3. Use the Latin hypercube sampling method to obtain several data points for each optimization parameter, and build a three-dimensional model based on all data points; perform CFD simulation analysis on all three-dimensional models to obtain the corresponding isentropic efficiency and maximum turbulent kinetic energy, and use the optimization parameter data points and the corresponding isentropic efficiency and maximum turbulent kinetic energy as a training sample to obtain a sample set containing multiple training samples.

[0043] S4. Optimize the parameters of the impeller geometry model. The optimization includes the following steps: S41. Based on the sample set, use a neural network to fit and construct a surrogate model; S42. Combine a multi-objective genetic algorithm to perform multi-objective optimization on the surrogate model, obtain the Pareto front solution set, and select the optimal solution according to actual needs.

[0044] This embodiment conducts a simulation study on a certain air compressor.

[0045] First, an initial model of the air compressor is constructed, and CFD simulation analysis is performed on the initial model to obtain the flow field of the initial model. Then, the Q criterion is used to extract the turbulent pseudo-order structure in the flow field, and the influence of the pseudo-order structure on the isentropic efficiency and the maximum turbulent kinetic energy is analyzed.

[0046] like Figure 2 As shown, the turbulent kinetic energy is highest in the region of interference between the leakage flow and the wake vortex. As the airflow moves downstream from the trailing edge, the turbulent kinetic energy generally increases first and then decreases. This is because the pseudo-ordered structure generated by the interaction between the endwall vortex and the leakage vortex increases the turbulent kinetic energy, reaching its maximum in the region of interference between the leakage vortex and the wake vortex. The vortex breaking caused by the interference gradually reduces the turbulent kinetic energy, dissipating it into a disordered vortex structure. Furthermore, the interaction between vortex systems exacerbates the unsteady flow losses, which intensify with increasing tip clearance. Therefore, changes in the tip leakage vortex, wake vortex, and separated vortex are important reasons for the decrease in isentropic efficiency and the increase in turbulent kinetic energy.

[0047] Depend on Figure 3 It is known that the vorticity is concentrated at the blade tip. The blade tip leakage vortex mainly originates from leakage vortex 1 formed by the shearing of the leakage flow at the blade tip, and leakage vortex 2 formed by the mixing of the mainstream and the leakage flow. As the blade tip clearance increases, the leakage flow rate increases, the volume of leakage vortex 1 increases, and it compresses leakage vortex 2, reducing the effective flow area and increasing flow losses. In summary, considering that a reasonable blade tip clearance can achieve the control of the pseudo-sequential structure and reduce turbulent dissipation losses, the optimization range of the blade tip clearance is narrowed to 0.25–0.35 mm.

[0048] Based on this, the blade inlet installation angle, blade tip clearance, blade outlet width, and blade outlet installation angle were determined as optimization parameters, with their respective optimization ranges being 30°~35°, 0.25mm~0.35mm, 3.8mm~5.4mm, and 45°~60°.

[0049] Several sets of data points for the impeller structure were obtained using the Latin hypercube sampling method. Impeller parameters were adjusted based on these data points to establish a three-dimensional model. CFD simulation analysis was then performed to obtain the isentropic efficiency and maximum turbulent kinetic energy within the impeller for each air compressor. The optimized parameter data points and their corresponding isentropic efficiency and maximum turbulent kinetic energy were used as training samples, resulting in a sample set containing multiple training samples. The sample set is shown in Table 1.

[0050] Table 1

[0051]

[0052]

[0053] The sample set was fitted and trained using the Neural Network Toolbox, with the optimized parameter combination as the independent variable and isentropic efficiency and maximum turbulent kinetic energy as the dependent variables. The data was imported into the Neural Network Toolbox. The training and test data ratios were set to 85% and 15%, respectively, and the number of hidden neurons was set to 10. A surrogate model for the optimized air compressor parameters, isentropic efficiency, and maximum turbulent kinetic energy was trained. The slope R of the regression line was used as the accuracy evaluation index for the surrogate model. Figure 4 As shown, the training effect is good.

[0054] Install the blade inlet angle β 1A Blade tip clearance C c Blade exit width b c and blade outlet installation angle β 2A The surrogate model serves as the fitness function, serving as the optimization variable. Constraints are set for the optimization variables: the optimization ranges for the blade inlet angle, tip clearance, outlet width, and outlet angle are 30°–35°, 0.25mm–0.35mm, 3.8mm–5.4mm, and 45°–60°, respectively. The population size is set to 100, the optimal front-end individual coefficient to 0.3, the number of iterations to 150, the fitness function value deviation to 1e-5, and the probabilities of crossover and mutation to 0.8 and 0.1, respectively. Based on the constraints of the optimization parameters and the fitness function, a multi-objective optimization mathematical model is established. A multi-objective genetic algorithm is used to optimize the mathematical model, yielding the Pareto front, such as... Figure 5 As shown in the figure, the x and y axes represent isentropic efficiency and maximum turbulent kinetic energy, respectively. It can be seen from the figure that, in general, the increase in isentropic efficiency leads to an increase in maximum turbulent kinetic energy.

[0055] In this embodiment, the mathematical model is as follows:

[0056]

[0057] In the formula, f1(x) is the isentropic efficiency of the air compressor; f2(x) is the maximum turbulent kinetic energy of the air compressor; β 1A Installation angle for blade inlet; C c b is the blade tip gap; c β is the blade exit width; 2A For the blade outlet installation angle; β 1A1 and β 1A2 These represent the minimum and maximum values ​​of the blade inlet installation angle, C. c1 and C c2 These represent the minimum and maximum values ​​of the blade tip clearance, respectively. c1 and b c2 These represent the minimum and maximum values ​​of the blade exit width, β. 2A1 and β 2A2 These represent the minimum and maximum values ​​of the blade exit installation angle, respectively. In this embodiment, β1A1 Take 30°, β 1A2 Take 35°, C c1 Take 0.25mm, C c2 Take 0.35mm, b c1 Take 3.8mm, b c2 Take 5.4mm, β 2A1 Take 45°, β 2A2 Take 60°.

[0058] CFD simulation was used to verify the isentropic efficiency and maximum turbulent kinetic energy of the optimized air compressor. The impeller structural parameters after optimization are shown in Table 2. The optimized isentropic efficiency is 81.21%, and the maximum turbulent kinetic energy is 2587.44 m³ / s. 2 / s 2 Compared to before optimization, the isentropic efficiency improved by 1.61%, and the maximum turbulent kinetic energy decreased by 418.03 m. 2 / s 2 .

[0059] Table 2 Comparison of parameters before and after optimization

[0060]

[0061] It should also be noted that modifications or substitutions made by those skilled in the art to the foregoing embodiments do not depart from the scope of the technical solutions of the embodiments of the present invention; similarly, meaningless alterations or refinements made to the main design concept and spirit of the present invention that solve the same technical problem as the present invention should be included within the protection scope of the present invention.

Claims

1. A method for multi-objective optimization of an air compressor impeller based on turbulent pseudo-order regulation, characterized in that, The method comprises the following steps: Step S1, constructing an initial model of an air compressor, performing CFD simulation analysis on the initial model of the air compressor, and obtaining a flow field of the initial model; Step S2, extracting a turbulent quasi-order structure in the flow field by Q criterion, analyzing the influence of the turbulent quasi-order structure on isentropic efficiency and maximum turbulent kinetic energy, determining an optimization parameter and a value interval; Step S3, using a Latin hypercube sampling method to obtain a plurality of data points about each optimization parameter, establishing a three-dimensional model according to all the data points, performing CFD simulation analysis on all the three-dimensional models, obtaining corresponding isentropic efficiency and maximum turbulent kinetic energy, setting the optimization parameter data points and the corresponding isentropic efficiency and maximum turbulent kinetic energy as a training sample, and obtaining a sample set containing a plurality of training samples; Step S4, performing parameter optimization on a blade wheel geometric model, the optimization comprising the following steps: S41, based on the sample set, using a neural network fitting to construct a surrogate model; S42, jointly using a multi-objective genetic algorithm to perform multi-objective optimization on the surrogate model, obtaining a Pareto front solution set, and selecting a required optimal solution according to actual requirements; In the step S2, the Q criterion definition formula is as follows: (1) In the formula, Ω is a rotation tensor; S is a strain rate tensor; and Q is a vorticity; In the step S2, the turbulent quasi-order structure refers to an ordered structure in irregular flow in a flow field, which in an air compressor is in the form of a separation vortex, a leakage vortex and a wake vortex; In the step S41, the sample set is fitted and trained by using a neural network toolbox, and the steps are as follows: S411, taking an optimization parameter combination as an independent variable and isentropic efficiency and maximum turbulent kinetic energy as dependent variables, importing data into the neural network toolbox; S412, setting a training data proportion, a test data proportion and a number of hidden neurons; S413, training to obtain a surrogate model of air compressor optimization parameters and isentropic efficiency and maximum turbulent kinetic energy, and using a regression straight line slope R as a surrogate model precision evaluation index; In the step S42, the multi-objective genetic algorithm is used for optimization, and the optimization steps are as follows: S421, taking the blade inlet installation angle β 1A , the blade tip clearance C c , the blade outlet width b c and the blade outlet installation angle β 2A as optimization variables, and taking the agent model as the fitness function; S422, setting the constraint conditions of the optimization variables; S423, setting the population size, the optimal front individual coefficient, the iteration number, the fitness function value deviation, the probabilities of crossover and variation; S424, establishing a mathematical model of multi-objective optimization based on the optimization parameters and the constraint conditions of the fitness function; and S425, adopting the multi-objective genetic algorithm to optimize the mathematical model, and selecting the optimal solution in the derived Pareto front. In the S424, the mathematical model is as follows: (4) In the formula, f1(x) is the isentropic efficiency of air compressor; f2(x) is the maximum turbulent kinetic energy of air compressor; β 1A is the blade inlet installation angle; C c is the tip clearance; b c is the blade outlet width; β 2A is the blade outlet installation angle; β 1A1 and β 1A2 are respectively the minimum and maximum values of the blade inlet installation angle, C c1 and C c2 are respectively the minimum and maximum values of the tip clearance, b c1 and b c2 are respectively the minimum and maximum values of the blade outlet width, β 2A1 and β 2A2 are respectively the minimum and maximum values of the blade outlet installation angle.

2. The air compressor blade multi-objective optimization method based on turbulent quasi-order regulation according to claim 1, characterized in that: In the step S2, the isentropic efficiency calculation formula of the air compressor is as follows: (2) wherein Tin is the measured total temperature of the air compressor inlet, in K; Tout is the measured total temperature of the air compressor outlet, in K; k is the ideal gas adiabatic index.

3. The method according to claim 1, wherein the method is characterized by: In the step S2, the turbulent kinetic energy is an index that measures the development or decay of turbulence, and is calculated as shown below. (3) In the formula, U is an average speed, in units of m / s; and I is a turbulent intensity; In the step S2, the determined optimization parameters are the blade inlet installation angle β 1A , the blade tip clearance C c , the blade outlet width b c , the blade outlet installation angle β 2A .

4. The air compressor blade multi-objective optimization method based on turbulent quasi-order regulation according to claim 1, characterized in that: In the step S3, the Latin hypercube sampling method provides a value for each interval (1, 1 / n), (1 / n, 2 / n) … (1-1 / n, 1) in n groups of data of x blade structures and randomly arranges the values.

5. The air compressor blade multi-objective optimization method based on turbulent quasi-order regulation according to claim 1, characterized in that: In the step S3, the generated sample set comprises combinations of all optimization parameters and the corresponding isentropic efficiency and maximum turbulent kinetic energy, wherein the combinations of optimization parameters are combinations of the parameter set of the blade inlet installation angle β 1A , the blade tip clearance C c , the blade outlet width b c and the blade outlet installation angle β 2A .

6. The air compressor blade multi-objective optimization method based on turbulent quasi-order regulation according to claim 1, characterized in that: In the S413, the precision of the surrogate model obtained by neural network fitting and training is judged by the regression straight line slope R, and the closer R is to 1, the better the training effect is.

Images