Axial flow compressor rapid prediction method based on self-adaptive correction

Abstract

An axial flow compressor rapid prediction method based on self-adaptive correction comprises the following steps that a one-dimensional performance rapid prediction model of an axial flow compressor is constructed based on compressor geometric parameters and a one-dimensional average pitch diameter theory; based on a one-dimensional performance rapid prediction model, taking a relative error between a calculation result and experimental performance as an optimization target, introducing a correction factor into a lag angle and total pressure loss empirical model as an optimization parameter, and carrying out optimization of the correction factor; establishing a mapping relation among the compressor inlet speed coefficient, the inlet relative Mach number and the optimal correction factor under different working conditions; based on an existing compressor experimental performance result, a correction factor optimization method is adopted, a mapping relation between the multiple compressor inlet working condition parameters and the optimal correction factor is constructed, and a data set is formed; a BP neural network is utilized to train a data set, so that the data set has generalization ability; and calling the trained BP neural network in the one-dimensional performance rapid prediction model to realize adaptive correction of the empirical model and rapid prediction of the performance.

Description

Technical Field

[0001] The present invention belongs to the field of fluid machinery and relates to a fast prediction method for axial flow compressors based on adaptive correction. Background Technique

[0002] Axial flow compressors are key components in thermal equipment such as aeroengines and gas turbines. Their performance prediction is of great significance for optimal design and engineering applications. Currently, the performance prediction methods for axial flow compressors include one-dimensional, two-dimensional, and three-dimensional CFD numerical simulations from the perspective of dimensions. Among them, although three-dimensional calculations can accurately capture complex flow structures, they have high computational costs and strong dependencies on mesh generation and turbulence models. The patent with the publication number CN 117494566 A proposes that the full three-dimensional numerical simulation of the compressor based on CFD is extremely time-consuming and not suitable for the dynamic real-time calculation of the engine's overall model; two-dimensional calculations also have high computational complexities. The paper titled "Review of Geometric Aerodynamic Performance Optimization Design Methods for Compressor Blades" proposes that when conducting the optimization design of compressors, one-dimensional calculation results are usually used as the basis for two-dimensional calculations. Thus, it can be seen that although one-dimensional methods have undergone a large amount of model simplification, they play an irreplaceable role in the fields of preliminary design optimization, performance fast prediction, and real-time data simulation of compressors.

[0003] In the one-dimensional calculation of compressors, the mean mean diameter method is widely used in engineering problems due to its clear physical concept. The core of this method lies in the introduction of empirical models, which are derived from a large number of cascade experiments. Therefore, the accuracy of the models is related to the types of experimental cascades. With the development of turbomachinery, the shapes of compressor blades are diverse. If calculated directly according to the results of empirical models, there may be large deviations. And the method of manually adjusting model parameters to correct errors lacks a systematic optimization mechanism and also increases the computational cost consumption and uncertainty. Therefore, it is necessary to develop a method for automatically correcting empirical models based on experimental characteristic targets, and this correction can be accurately applied to the entire operating process of the compressor. This is of great significance for the accuracy of the fast prediction of compressor performance. Summary of the Invention

[0004] The present invention provides a fast prediction method for axial flow compressors based on adaptive correction, aiming to solve the problems of fixed empirical parameters, poor applicability, and low calculation accuracy in existing axial flow compressor performance prediction methods. This method combines an optimization algorithm with neural network modeling to achieve the adaptive adjustment of the parameters of the one-dimensional mean mean diameter model, improving the accuracy and generalization ability of axial flow compressor performance prediction.

[0005] The technical solution of the present invention:

[0006] A fast prediction method for axial flow compressors based on adaptive correction is as follows:

[0007] S1. Based on the compressor geometric parameters and the one-dimensional mean pitch line theory, construct a one-dimensional performance rapid prediction model for the axial compressor;

[0008] S11. To construct a one-dimensional performance rapid prediction model for the axial compressor, first determine three types of input parameters required, namely compressor geometric parameters, compressor inlet aerodynamic parameters, and physical property constants;

[0009] The compressor geometric parameters include the leading edge blade profile angle of each blade row, the trailing edge blade profile angle of each blade row, the blade setting angle of each blade row, the leading edge thickness of each blade row, the maximum thickness of each blade row, the chord length of each blade row,

[0010] the blade solidity of each blade row, the hub and tip radii at the inlet and outlet;

[0011] where the subscript t represents the tip, h represents the hub, 1 represents the inlet of the blade row, and 2 represents the outlet of the blade row; The compressor inlet aerodynamic parameters include the total temperature at the compressor inlet, the total pressure at the compressor inlet, and the absolute flow angle

[0012] at the compressor inlet;

[0013] The physical property constants include the atmospheric total pressure under standard conditions,

[0014] the atmospheric total temperature

[0015] under standard conditions,

[0016] the ideal gas constant of air,

[0017] and the air dynamic viscosity

[0018] Introduce the actual incidence angle correction See formula (4), three-dimensional effect correction See formula (5), axial velocity ratio correction See formula (6), add all corrections to the designed stagger angle to obtain the stagger angle of the blade row ;

[0019] (4)

[0020] (5)

[0021] (6)

[0022] (7)

[0023] Among them, represents the blade bend angle; represents the airfoil correction factor; represents the correction factor when the ratio of the maximum blade thickness to the chord length is not 10%; represents the designed incidence angle of the standard NACA-65 airfoil when the bend angle is 0°; n represents the correction factor for the bend angle; represents the designed stagger angle of the standard NACA-65 airfoil when the bend angle is 0°; m represents the correction factor for the bend angle; i represents the actual incidence angle; represents the relative Mach number of the airflow; represents the axial inlet velocity of the blade row; represents the axial outlet velocity of the blade row;

[0024] After obtaining the stagger angle of the blade row, carry out the calculation of the velocity triangle at the outlet of the blade row; first calculate the total pressure loss , which is obtained by summing the airfoil loss , the shock loss , the secondary flow loss and the off-design condition loss , see formula (8):

[0025] (8)

[0026] Adopt the initial value iteration method, and give the initial value of the outlet axial velocity as ; use the total pressure loss and the Euler work equation to obtain the total temperature and total pressure at the outlet of the blade row, and replace with for the next iteration until and The residual is less than 1e-5, which is regarded as cyclic convergence and used as the axial velocity at the blade row outlet. Then, the velocity triangle, total temperature, and total pressure at the blade row outlet are determined. Here, the superscript 0 represents the initial value, and 1 to x represent the number of iterations.

[0027] The Euler work equation is shown in Equation (8):

[0028] (9)

[0029] Among them, represents the tangential velocity of the air inlet of the blade row; represents the tangential velocity of the air outlet of the blade row; represents the circumferential velocity of the air inlet of the blade row; represents the circumferential velocity of the air outlet of the blade row; cp represents the specific heat capacity at constant pressure of air; represents the total temperature of the air inlet of the blade row; represents the total temperature of the air outlet of the blade row;

[0030] The above calculations are successively applied to the rotor blade row and stator blade row in a single stage of the compressor to obtain the total pressure ratio, efficiency characteristics, and inlet and outlet aerodynamic parameters of a single stage of the axial compressor.

[0031] S13. After completing the performance calculation of a single stage of the axial compressor, determine whether the calculated inlet flow rate and actual rotational speed correspond to the stall zone operating conditions. When the inequality (10) is satisfied, it indicates that the compressor is operating in a stalled state under this condition, and no further calculations are performed. Otherwise, continue with the subsequent calculations;

[0032] (10)

[0033] Among them, represents the static pressure of the air inlet of the blade row; represents the static pressure of the air outlet of the blade row; represents the total pressure of the air inlet of the blade row; represents the total pressure of the air outlet of the blade row;

[0034] S14. After confirming that the current operating condition has not entered the stall zone, input the aerodynamic parameters at the outlet of the previous stage stator as the inlet parameters for the next stage rotor; according to the calculation method for the performance of a single stage of the axial compressor in step S12, calculate the performance stage by stage, and finally obtain the total pressure ratio and isentropic efficiency characteristics of the compressor corresponding to the actual inlet flow rate and actual rotational speed to be calculated; the total pressure ratio and isentropic efficiency characteristics are calculated according to Equations (11)-(12):

[0035] (11)

[0036] (12)

[0037] Among them, is the total pressure at the compressor outlet, is the total temperature at the compressor outlet, and k is the adiabatic index of air;

[0038] S2. Based on the one-dimensional performance rapid prediction model proposed in S1, with the relative error between the calculation result and the experimental performance as the optimization objective, a correction factor is introduced into the empirical model of the stagger angle and the empirical model of the total pressure loss as the optimization parameter, and the optimization of the correction factor is carried out. Finally, the mapping relationship between the inlet velocity coefficient, the inlet relative Mach number and the optimal correction factor of the compressor under different working conditions is established;

[0039] S21. Define the optimization objective function to characterize the relative error between the pressure ratio and efficiency results of the one-dimensional performance rapid prediction model and the experimental values under the conditions of the same inlet flow rate and actual rotational speed. The optimization objective is the minimum value of F; the specific definition is as follows:

[0040] (13)

[0041] where the subscript 1D represents the compressor performance obtained by the rapid prediction method; the subscript exp represents the compressor performance known from experiments; and represent the weight factor to balance the importance of the indicators, and take 0.5;

[0042] S22. To avoid the optimization falling into a local optimal solution, a regularization constraint is imposed on the optimization objective function, and a regularization strength factor λ is introduced into the optimization objective function. The constrained optimization objective function is defined as follows:

[0043] (14)

[0044] where λ is the regularization factor and takes 0.05, y is the number of correction factors participating in the optimization, is a certain correction factor parameter in the current optimization iteration, is the initial value of this correction factor;

[0045] S23. Introduce a correction factor into the empirical model of the stagger angle and the empirical model of the total pressure loss, so as to correct the stagger angle and the total pressure loss in the rapid prediction method of compressor performance as shown in formulas (15)-(16); is the variable parameter to be optimized. Among them, the change range of is 0.5 - 1.5;

[0046] (15)

[0047] (16)

[0048] S24. Based on the one-dimensional performance fast prediction model established in step S1, with the goal of minimizing the optimization objective function in S22, use the particle swarm optimization algorithm to optimize the optimization variables determined in S23; obtain an optimal set of correction factors , and thus establish the mapping relationship between the compressor inlet operating condition parameters and the optimal correction factors; the compressor inlet operating condition parameters include the inlet velocity coefficient and the inlet relative Mach number; the parameter settings of the particle swarm optimization include the number of particles, the maximum number of iterations, the error convergence threshold, and the maximum movement speed; establish the standard particle swarm optimization velocity-position update relationship as follows:

[0049] (17)

[0050] (18)

[0051] Among them, is the updated particle movement speed, is the particle movement speed before update, the inertia weight is set to 0.8, , the learning factors are both set to 2, r1 and r2 are random numbers in the interval [0,1], is the historical optimal position of the current particle, is the global optimal position of all particles, is the updated particle position, is the particle position before update;

[0052] S3. Based on the existing compressor experimental performance results, adopt the correction factor optimization method in S2, construct the mapping relationships between multiple compressor inlet operating condition parameters and the optimal correction factors and form a data set; use the BP neural network to train the data set to make it have generalization ability; call the trained BP neural network in the one-dimensional performance fast prediction model to realize the adaptive correction of the empirical model and the fast prediction of performance;

[0053] In the mapping data set of the compressor inlet operating condition parameters and the optimal correction factors, the input variables are defined as the inlet velocity coefficient and the inlet relative Mach number, and the output variable is the corresponding optimal correction factor; set the learning rate parameter, batch size, and number of training rounds of the exponential decay strategy in the BP neural network, and use Dropout regularization to prevent overfitting; adopt the Adam optimizer to update the weights, and the loss function selects the root mean square error, and the calculation formula is as follows:

[0054] (19)

[0055] Among them, N is the total number of sample points, is the target value of the i-th sample, is the predicted value of the i-th sample;

[0056] After completing the training of the data set, the BP network is transferred into the one-dimensional performance rapid prediction model established in step S1 to obtain the optimal correction factor corresponding to the inlet operating conditions and introduce the lag angle empirical model and the total pressure loss empirical model for calculation.

[0057] Technical effects of the present invention:

[0058] (1) Improved prediction accuracy: Compared with the unoptimized traditional one-dimensional model, the optimized model has reduced prediction errors under different working conditions and improved calculation accuracy.

[0059] (2) Enhanced applicability: The optimization method is used to dynamically adjust the empirical parameters to make the model applicable to different working conditions, thereby improving its adaptability.

[0060] (3) Improve computing efficiency: Avoid repeated optimization calculations. The BP neural network can quickly provide optimization parameters and achieve real-time performance prediction.

[0061] The invention is applicable to the performance evaluation of axial flow compressors, the design optimization of gas turbines and aeroengines, and has high engineering application value. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 It is a flowchart of the method of the present invention;

[0063] Figure 2 It is a schematic diagram of the calculation framework of the one-dimensional average median diameter method for an axial flow compressor;

[0064] Figure 3 It is a flow chart of particle swarm optimization (PSO) algorithm;

[0065] Figure 4 It is a schematic diagram of the BP neural network training framework;

[0066] Figure 5 This is a performance prediction and experimental comparison chart of a three-stage axial flow compressor, where (a) is the total pressure ratio-flow rate, and (b) is the isentropic efficiency-flow rate. DETAILED DESCRIPTION

[0067] The specific implementation of the present invention is further described below in conjunction with the accompanying drawings and technical solutions.

[0068] Example

[0069] A fast prediction method for an axial flow compressor based on adaptive correction, the method can be implemented by an electronic device, the electronic device can be a terminal or a server. Figure 1Flow chart of a rapid performance prediction method for axial flow compressors based on adaptive correction. The processing flow of this method may include the following steps:

[0070] S1. Based on the geometric parameters of the compressor and the one-dimensional mean diameter theory, construct a one-dimensional rapid performance prediction model for the axial flow compressor.

[0071] Optionally, S1 specifically includes:

[0072] S11. To construct a one-dimensional rapid performance prediction model for the axial flow compressor, first determine three types of input parameters required, namely the geometric parameters of the compressor, the inlet aerodynamic parameters of the compressor, and the physical property constants;

[0073] The geometric parameters of the compressor include the leading edge blade profile angle of each blade row , the trailing edge blade profile angle , the blade setting angle , the leading edge thickness of the blade , the maximum thickness of the blade , the chord length of the blade , the blade solidity , the hub and tip radii at the inlet and outlet , , , ; where the subscript t represents the tip, h represents the hub, 1 represents the inlet of the blade row, and 2 represents the outlet of the blade row;

[0074] The inlet aerodynamic parameters of the compressor include the total temperature at the inlet of the compressor , the total pressure at the inlet of the compressor and the absolute flow angle of the air at the inlet of the compressor ;

[0075] The physical property constants include the atmospheric total pressure under standard conditions , the atmospheric total temperature under standard conditions , the ideal gas constant of air and the air dynamic viscosity under conditions;

[0076] S12. To obtain the total pressure ratio, efficiency characteristics, and aerodynamic parameters of a single stage of the compressor, first establish the velocity triangle at the inlet of the blade row; the parameters constituting the velocity triangle at the inlet of the blade row include the absolute inlet velocity of the air , the inlet axial velocity , the inlet tangential velocity , the inlet relative velocity , the inlet circumferential velocity , the absolute inlet flow angle , the relative inlet flow angle ;

[0077] Obtain the stagger angle by combining the design incidence angle and the stagger angle empirical model;

[0078] First, obtain the design incidence angle , as shown in formulas (1)-(2);

[0079] (1)

[0080] (2)

[0081] Then, design the stagger angle , as shown in formula (3):

[0082] (3)

[0083] Introduce the actual incidence angle correction See formula (4), the three-dimensional effect correction See formula (5), the axial velocity ratio correction See formula (6), and add all the corrections to the design stagger angle to obtain the stagger angle of the blade row ;

[0084] (4)

[0085] (5)

[0086] (6)

[0087] (7)

[0088] Among them, represents the blade camber angle; represents the airfoil correction factor; represents the correction factor when the ratio of the maximum blade thickness to the chord length is not 10%; represents the design incidence angle of the standard NACA-65 airfoil when the camber angle is 0°; n represents the correction factor for the camber angle; represents the design stagger angle of the standard NACA-65 airfoil when the camber angle is 0°; m represents the correction factor for the camber angle; i represents the actual incidence angle; represents the relative Mach number of the airflow;

[0089] After obtaining the stagger angle of the blade row, carry out the calculation of the velocity triangle at the outlet of the blade row. The parameters constituting the velocity triangle at the outlet of the blade row include the absolute outlet velocity of the airflow , the outlet axial velocity , the outlet tangential velocity , the outlet relative velocity , the circumferential velocity at the outlet , the absolute flow angle at the outlet , the relative flow angle at the outlet ; Calculate the total pressure loss See Equation (8). The total pressure loss is divided into 4 terms, consisting of the profile loss , the shock loss , the secondary flow loss and the off-design condition loss and is obtained by summing them up. See Equations (9)-(12):

[0090] (8)

[0091] (9)

[0092] (10)

[0093] (11)

[0094] (12)

[0095] Among them, represents the reduced momentum thickness; represents the blade shape factor; k is the air adiabatic index; represents the Mach number at the inlet of the blade row; represents the Mach number at the throat of the flow passage; represents the blade lift coefficient;

[0096] Adopt the initial value iteration method, and the initial value of the axial velocity at the outlet is given as ; Use the total pressure loss and the Euler work equation to obtain the total temperature and total pressure at the outlet of the blade row, and replace with for the next iteration until and The residuals of are less than 1e-5, which is regarded as the convergence of the loop and used as the axial velocity at the outlet of the blade row. Then, the velocity triangle and the total temperature and total pressure at the outlet of the blade row are determined; among them, the superscript 0 represents the initial value; 1~x represents the number of iterations;

[0097] The Euler work equation is shown in Equation (13):

[0098] (13)

[0099] Among them, cp represents the specific heat capacity at constant pressure of air; represents the total temperature of the air inlet of the blade row; represents the total temperature of the air outlet of the blade row;

[0100] Apply the above calculations to the rotor blade row and stator blade row in a single stage of the compressor successively to obtain the total pressure ratio, efficiency characteristics, and inlet and outlet aerodynamic parameters of a single stage of the axial compressor;

[0101] S13. After completing the performance calculation of a single stage of the axial compressor, determine whether the calculated inlet flow rate and actual rotational speed correspond to the stall zone operating conditions. When the inequality (14) is satisfied, it indicates that the compressor has stalled under this operating condition, and no further calculations are performed. Otherwise, continue with the subsequent calculations;

[0102] (14)

[0103] Among them, represents the static pressure at the inlet of the blade row; represents the static pressure at the outlet of the blade row; represents the total pressure at the inlet of the blade row; represents the total pressure at the outlet of the blade row;

[0104] S14. After confirming that the current operating condition has not entered the stall zone, input the aerodynamic parameters at the outlet of the previous stage stator as the inlet parameters of the next stage rotor; calculate the performance stage by stage according to the calculation method of the single stage performance of the axial compressor in step S12, and finally obtain the total pressure ratio and isentropic efficiency characteristics of the compressor corresponding to the actual inlet flow rate and actual rotational speed to be calculated; the total pressure ratio and isentropic efficiency characteristics are calculated according to formulas (15)-(16):

[0105] (15)

[0106] (16)

[0107] Among them, is the total pressure at the outlet of the compressor, is the total temperature at the outlet of the compressor, and k is the air adiabatic index.

[0108] S2. Based on the one-dimensional performance rapid prediction model proposed in S1, with the relative error between the calculation results and the experimental performance as the optimization objective, introduce correction factors as optimization parameters in the trailing angle empirical model and the total pressure loss empirical model, conduct the optimization of the correction factors, and finally establish the mapping relationship between the inlet velocity coefficient, inlet relative Mach number of the compressor and the optimal correction factor under different operating conditions;

[0109] Optionally, the S2 specifically includes:

[0110] S21. Define the optimization objective function, which characterizes the relative error between the pressure ratio and efficiency results of the one-dimensional performance rapid prediction model and the experimental values under the conditions of the same inlet flow rate and actual rotational speed. The optimization objective is the minimum value of F; the specific definition is as shown in formula (17):

[0111] (17)

[0112] Among them, the subscript 1D represents the compressor performance obtained by the fast prediction method; the subscript exp represents the compressor performance known from experiments; and represent the weight factors to balance the importance of indicators; and In this example, the same value of 0.5 is selected, that is, the influence of the errors of the pressure ratio and efficiency on the target is considered according to the same weight.

[0113] S22. To avoid the optimization falling into a local optimal solution, a regularization constraint is imposed on the optimization objective function, and a regularization strength factor λ is introduced into the optimization objective function. The constrained optimization objective function is defined as shown in Equation (18):

[0114] (18)

[0115] Among them, λ is the regularization factor and takes 0.05, y is the number of correction factors participating in the optimization, is a certain correction factor parameter in the current optimization iteration, is the initial value of this correction factor;

[0116] S23. A correction factor is introduced into the stagger angle empirical model and the total pressure loss empirical model to correct the stagger angle and total pressure loss in the fast prediction method of compressor performance as shown in Equations (19)-(20); is the variable parameter to be optimized. Among them, the variation range of is 0.5 - 1.5; the variation range of is 0.5 - 3;

[0117] (19)

[0118] (20)

[0119] S24. Based on the one-dimensional performance fast prediction model established in step S1, with the goal of minimizing the optimization objective function in S22, the particle swarm optimization algorithm is used to optimize the optimization variables determined in S23; the optimal set of correction factors is obtained, and the mapping relationship between the compressor inlet condition parameters and the optimal correction factors is established; the compressor inlet condition parameters include the inlet velocity coefficient and the inlet relative Mach number; the parameter settings of the particle swarm optimization include the number of particles, the maximum number of iterations, the error convergence threshold, and the maximum movement speed; the standard particle swarm optimization velocity-position update relationship is established as follows:

[0120] (21)

[0121] (22)

[0122] Wherein, is the updated particle movement speed, is the particle movement speed before update, is the inertia weight set to 0.8, 、 are the learning factors both set to 2, r1, r2 are random numbers in the interval [0, 1], is the historical optimal position of the current particle, is the global optimal position of all particles, is the updated particle position, is the particle position before update;

[0123] In this example, the number of particles in the particle swarm algorithm is set to 40, the maximum number of iterations is limited to 150, the error convergence threshold is set to 10e - 3, and the maximum movement speed is limited to 0.15.

[0124] S3. Based on the existing experimental performance results of the compressor, adopt the correction factor optimization method in S2 to construct the mapping relationship between multiple compressor inlet condition parameters and the optimal correction factor and form a data set; use the BP neural network to train the data set to make it have generalization ability; call the trained BP neural network in the one - dimensional performance rapid prediction model to realize the adaptive correction of the empirical model and the rapid prediction of performance;

[0125] In the mapping data set of the compressor inlet condition parameters and the optimal correction factor, the input variables are defined as the inlet velocity coefficient and the inlet relative Mach number, and the output variable is the corresponding optimal correction factor; set the learning rate parameter, batch size, and number of training epochs with an exponential decay strategy in the BP neural network, and use Dropout regularization to prevent overfitting; use the Adam optimizer to update the weights, and the loss function is selected as the root mean square error, and the calculation formula is as follows:

[0126] (23)

[0127] Wherein, N is the total number of sample points, is the target value of the i - th sample, is the predicted value of the i - th sample;

[0128] In this example, the initial value of the BP neural network training is set to 0.01 and gradually decreased using the exponential decay strategy; the batch size is set to 32 or 64 to balance computational efficiency and stability; the number of training epochs is set to 500 epochs.

[0129] Those of ordinary skill in the art can understand that all or part of the steps of implementing the above embodiments can be completed by hardware circuits, firmware programs or software instructions. The program can be stored in a computer-readable storage medium, which may include, but is not limited to, read-only memory (ROM), random access memory (RAM), CD-ROM, magnetic tape, floppy disk, optical storage device or other media that can store instructions.

[0130] To verify the effectiveness of the rapid prediction method for axial flow compressors based on adaptive correction, a certain 3-stage transonic axial flow compressor was selected for calculation. First, based on the geometric parameters of the compressor and the one-dimensional mean pitch-line theory, a one-dimensional performance rapid prediction model of the 3-stage compressor was constructed. The experimental data under the conditions of 100%, 90%, and 80% of the design speed were selected as the experimental performance. Taking the relative error between the calculation result and the experimental performance as the optimization goal, correction factors were introduced into the empirical model of the stagger angle and the empirical model of the total pressure loss as optimization parameters, and the particle swarm algorithm was used to optimize the correction factors, and the mapping relationship between the inlet velocity coefficient, inlet relative Mach number and the optimal correction factor of the compressor under different conditions was established. Multiple mapping relationships between the inlet condition parameters of the compressor and the optimal correction factor were constructed to form a data set; the BP neural network was used to train the data set, the trained BP neural network was called in the one-dimensional performance rapid prediction model, and the total pressure ratio and isentropic efficiency characteristics of the 3-stage compressor under the condition of 70% of the design speed were calculated.

[0131] Figure 5 The total pressure ratio-flow rate and isentropic efficiency-flow rate characteristics obtained by directly using the one-dimensional mean pitch-line method and the rapid prediction method for axial flow compressors based on adaptive correction were compared, and at the same time, they were compared with the experimental data. The performance of the compressor obtained directly by using the one-dimensional mean pitch-line method was generally higher than the experimental points, the maximum error of the total pressure ratio was 14.9%, and the maximum error of the isentropic efficiency was 12.6; by using the prediction method proposed in the present invention, the maximum error of the total pressure ratio can be reduced to 3.17%, and the maximum error of the isentropic efficiency can be reduced to 3.73%. There is a significant improvement in the performance prediction accuracy.

Claims

1. A fast prediction method for axial flow compressors based on adaptive correction, characterized in that The steps are as follows: S1. Based on the compressor geometric parameters and the one-dimensional mean pitch-line theory, construct a one-dimensional performance rapid prediction model for the axial-flow compressor; S11. To construct a one-dimensional performance rapid prediction model for the axial-flow compressor, first determine three types of input parameters required, namely compressor geometric parameters, compressor inlet aerodynamic parameters, and physical property constants; S12. To obtain the total pressure ratio, efficiency characteristics, and aerodynamic parameters of a single stage of the compressor, first establish the velocity triangle at the inlet of the blade row, and combine the design incidence angle and the empirical model of the trailing angle to obtain the trailing angle; S13. After completing the calculation of the performance of a single stage of the axial-flow compressor, judge whether the calculated inlet flow rate and actual rotational speed correspond to the stall zone conditions. When the inequality (10) is satisfied, it indicates that the compressor is operating in a stalled state under this condition, and no further calculations are performed. Otherwise, continue with the subsequent calculations; （10） Among them, represents the inlet static pressure of the blade row; represents the outlet static pressure of the blade row; represents the inlet total pressure of the blade row; represents the outlet total pressure of the blade row; represents the maximum blade thickness; represents the blade chord length; represents the difference between the leading-edge blade profile angle and the trailing-edge blade profile angle; represents the blade solidity; represents the blade installation angle; S14. After confirming that the current condition has not entered the stall zone, use the aerodynamic parameters at the outlet of the upstream stator as the inlet parameters for the downstream rotor; according to the calculation method for the performance of a single stage of the axial-flow compressor in step S12, calculate the performance stage by stage, and finally obtain the total pressure ratio and isentropic efficiency characteristics of the compressor corresponding to the actual inlet flow rate and actual rotational speed to be calculated; the total pressure ratio and isentropic efficiency characteristics are calculated according to formulas (11)-(12): （11） （12） Among them, is the total pressure at the compressor inlet; is the total pressure at the compressor outlet, is the total temperature at the compressor inlet; is the total temperature at the compressor outlet, and k is the adiabatic index of air; S2. Based on the one-dimensional performance rapid prediction model proposed in S1, with the relative error between the calculation result and the experimental performance as the optimization objective, introduce correction factors as optimization parameters in the empirical model of the trailing angle and the empirical model of the total pressure loss, conduct the optimization of the correction factors, and finally establish the mapping relationship between the inlet velocity coefficient, inlet relative Mach number, and the optimal correction factor of the compressor under different conditions; S3. Based on the existing experimental performance results of the compressor, use the correction factor optimization method in S2 to construct the mapping relationship between multiple compressor inlet condition parameters and the optimal correction factor and form a data set; use the BP neural network to train the data set to make it have the generalization ability; call the trained BP neural network in the one-dimensional performance rapid prediction model to realize the adaptive correction of the empirical model and the rapid prediction of the performance.

2. The rapid prediction method for the axial-flow compressor based on adaptive correction according to claim 1, characterized in that S11. The three types of input parameters are specifically as follows: The compressor geometric parameters include the leading edge blade profile angles of each blade row , the trailing edge blade profile angles , the blade setting angles , the leading edge blade thickness , the maximum blade thickness , the blade chord length , the blade solidity , the hub and shroud radii at the inlet and outlet , , , ; where the subscript t represents the shroud, h represents the hub, 1 represents the inlet of the blade row, and 2 represents the outlet of the blade row The aerodynamic parameters at the compressor inlet include the total temperature at the compressor inlet , the total pressure at the compressor inlet and the absolute flow angle at the compressor inlet ; The physical property constants include the total atmospheric pressure under standard conditions 、the total atmospheric temperature under standard conditions 、the ideal gas constant of air and the aerodynamic viscosity of air under conditions 3. The rapid prediction method for the axial-flow compressor based on adaptive correction according to claim 1, characterized in that S12. The specific process of obtaining the trailing angle is as follows: First, obtain the design angle of attack , as shown in formulas (1)-(2); （1） （2） Redesigned trailing angle , as shown in Equation (3): （3） Introduce the actual incidence angle correction See Equation (4), three-dimensional effect correction See Equation (5), axial velocity ratio correction See Equation (6), add all corrections to the designed stagger angle to obtain the stagger angle of the blade row ; （4） （5） （6） （7） Among them, represents the blade camber angle; represents the airfoil correction factor; represents the correction factor when the ratio of the maximum blade thickness to the chord length is not 10%; represents the design incidence angle of the standard NACA-65 airfoil when the camber angle is 0°; n represents the correction factor for the camber angle; represents the design stagger angle of the standard NACA-65 airfoil when the camber angle is 0°; m represents the correction factor for the camber angle; i represents the actual incidence angle; represents the relative Mach number of the airflow; represents the axial velocity of the inlet airflow of the blade row; represents the axial velocity of the outlet airflow of the blade; After obtaining the trailing angle of the blade row, the velocity triangle calculation at the outlet of the blade row is carried out; first, the total pressure loss is calculated , which is obtained by summing the profile loss , shock loss , secondary flow loss and off-design condition loss , as shown in formula (8): （8） Using the initial value iteration method, the initial value of the outlet axial velocity is given as ; the total temperature and total pressure at the outlet of the blade row are obtained by using the total pressure loss and the Euler work equation, and is replaced with for the next iteration until and the residuals of are less than 1e-5, which is regarded as the loop convergence and used as the outlet axial velocity of the blade row, and then the velocity triangle and the total temperature and total pressure at the outlet of the blade row are determined; where the superscript 0 represents the initial value; 1~x represents the number of iterations; The Euler work equation is shown in formula (8): （9） Among them, represents the tangential inlet velocity of the blade row; represents the tangential outlet velocity of the blade row; represents the circumferential inlet velocity of the blade row; represents the circumferential outlet velocity of the blade row; cp represents the specific heat capacity at constant pressure of air; represents the total inlet temperature of the blade row; represents the total outlet temperature of the blade row; Apply the above calculations to the rotor blade row and stator blade row in a single stage of the compressor in sequence to obtain the total pressure ratio, efficiency characteristics, and inlet and outlet aerodynamic parameters of a single stage of the axial-flow compressor.

4. The rapid prediction method for the axial-flow compressor based on adaptive correction according to claim 1, characterized in that S2. The specific implementation process is as follows: S21. Define the optimization objective function, which characterizes the relative error between the pressure ratio and efficiency results of the one-dimensional performance rapid prediction model and the experimental values under the conditions of the same inlet flow rate and actual rotational speed. The optimization objective is the minimum value of F; the specific definition is as follows: （13） where the subscript 1D represents the compressor performance obtained by the fast prediction method; the subscript exp represents the compressor performance known from experiments; and represents the weight factor to balance the importance of the indicators, taking 0.5; S22. To avoid the optimization from falling into a local optimal solution, the optimization objective function is regularized. A regularization strength factor λ is introduced into the optimization objective function, and the constrained optimization objective function is defined as follows: as follows: （14） Among them, λ is the regularization factor taking 0.05, y is the number of correction factors participating in the optimization, is a certain correction factor parameter in the current optimization iteration, is the initial value of this correction factor; S23. Introduce correction factors into the incidence angle empirical model and the total pressure loss empirical model , so as to correct the incidence angle and total pressure loss in the compressor performance rapid prediction method as shown in formulas (15)-(16); are the variable parameters to be optimized, where has a variation range of 0.5-1.5; has a variation range of 0.5-3; （15） （16） S24. Based on the one-dimensional performance rapid prediction model established in step S1, with the goal of minimizing the optimization objective function in S22, the particle swarm optimization algorithm is used to optimize the optimization variables determined in S23; obtain the optimal set of correction factors , thereby establishing the mapping relationship between the compressor inlet operating condition parameters and the optimal correction factors; the compressor inlet operating condition parameters include the inlet velocity coefficient and the inlet relative Mach number; the parameter settings of the particle swarm optimization include the number of particles, the maximum number of iterations, the error convergence threshold, and the maximum movement speed; establish the standard particle swarm optimization velocity-position update relationship as follows: （17） （18） Among them, is the updated particle motion speed, is the particle motion speed before update, the inertia weight is set to 0.8, 、 the learning factors are both set to 2, r1 and r2 are random numbers in the interval [0, 1], is the historical optimal position of the current particle, is the global optimal position of all particles, is the updated particle position, is the particle position before update.

5. The rapid prediction method for the axial-flow compressor based on adaptive correction according to claim 1, characterized in that The specific implementation process of S3. is as follows: In the mapping data set of compressor inlet operating parameters and optimal correction factors, the input variables are defined as the inlet speed coefficient and the inlet relative Mach number, and the output variable is the corresponding optimal correction factor; the learning rate parameter, batch size, and training rounds of the exponential decay strategy are set in the BP neural network, and Dropout regularization is used to prevent overfitting; the Adam optimizer is used for weight update, and the root mean square error is used as the loss function. The calculation formula is as follows: （19） where N is the total number of sample points, is the target value of the i-th sample, is the predicted value of the i-th sample; After completing the training of the data set, the BP network is transferred into the one-dimensional performance rapid prediction model established in step S1 to obtain the optimal correction factor corresponding to the inlet operating conditions and introduce the lag angle empirical model and the total pressure loss empirical model for calculation.

Images