A Simulation and Evaluation Method for Aerodynamic Noise of a Self-Ventilated Motor Fan

Abstract

The present invention relates to the technical field of motor noise, particularly to the technical field of aerodynamic noise of motor fans, and specifically provides a simulation and evaluation method for the aerodynamic noise of a self-ventilated motor fan. The purpose of the present invention is to provide a simulation and evaluation method for the aerodynamic noise of a motor fan with a short calculation period and strong applicability. First, input information such as the original geometric model, rotational speed, and boundary of the motor is obtained, and then a steady-state flow field analysis and data extraction are carried out. The extracted results are used as input parameters for the surrogate model to calculate the sound power level of the aerodynamic noise of the motor centrifugal fan. After obtaining the predicted value of the sound power level, an over-standard risk level assessment is carried out with the standard limit or customer requirements. After the assessment is passed, a prototype is produced and tested. If the assessment fails, the design scheme is adjusted, including adjusting relevant parameters such as the outer diameter of the fan, the number of fan blades, the width of the fan blades, and the shape of the fan blades, and re-analysis and prediction are carried out until the scheme passes the noise risk level assessment. This method is simple and easy to implement, and has a short evaluation period.

Description

Technical Field

[0001] The present invention relates to the technical field of motor noise, particularly to the technical field of aerodynamic noise of fans for self-ventilated motors, and specifically to a simulation and evaluation method for the aerodynamic noise of fans for self-ventilated motors. Background Art

[0002] With the development of technology and the improvement of train speed grades, higher requirements are put forward for the motor speed and power density. The fan speed and power increase accordingly, resulting in the problem that the noise of self-ventilated traction motors often exceeds the standard. Therefore, accurate noise evaluation during the motor design stage has become one of the main issues that designers need to consider.

[0003] According to GB / T 25123.2-2018 and GB / T 25123.4-2015, whether it is an asynchronous traction motor or a permanent magnet traction motor, the noise evaluation is mostly carried out under no-load, and the evaluation parameter is the sound power level. The problem of excessive noise of self-ventilated motors is prominent under high-speed conditions. The noise during high-speed operation mainly comes from the aerodynamic noise generated by the fan, and the electromagnetic noise and mechanical noise are relatively small and can be ignored compared. Therefore, when evaluating the noise of self-ventilated motors, generally only the aerodynamic noise caused by the fan is considered.

[0004] At present, a simple method for evaluating the aerodynamic noise of fans in engineering design is to use a general formula for estimation. In "Analysis and Control of Motor Noise" written by Chen Yongxiao et al. in China, the noise estimation formula for geometrically similar fans is introduced; in "Fan Handbook" edited by Xu Kuichang, the use of the specific A-weighted sound level to solve the noise prediction problem of the same structural form or the same series of ventilators is introduced. The American Society of Heating, Refrigerating and Air-Conditioning Engineers standard "AMCA301-2014" stipulates the use of existing noise test data of fans to calculate the sound power levels of each frequency band at different speeds, outer diameters or operating points of other fans with geometric similarity. The above standards and methods are only applicable to the rapid prediction of the aerodynamic noise of individual fans or have a small scope of application, and the applicability of predicting the sound power level of the aerodynamic noise of the motor fan during the test according to the relevant motor noise test standards cannot be guaranteed.

[0005] For the aerodynamic noise of the motor fan, the method of finite element simulation with mature technology can also be used for calculation. This method requires the establishment of a motor flow field model and an acoustic model, and the implementation of steady-state and transient flow field and acoustic solution calculations, with high accuracy and wide applicability. However, when used for calculating the noise of the motor fan, it takes a long time (generally 2-3 weeks), occupies a large amount of hardware resources, and cannot meet the requirements of a relatively fast calculation cycle during the motor design stage; at the same time, in the actual design process of the motor centrifugal fan, multiple schemes often need to be compared or multiple rounds of optimization are required. The characteristics of long single calculation cycle and low efficiency of finite element simulation slow down the process of motor optimization design and reduce the efficiency.

[0006] In summary, the existing general empirical formulas cannot be reliably applied to the rapid calculation of the aerodynamic noise of motor fans, while the finite element method has excessive requirements for computing resources, resulting in great difficulties in the simulation evaluation and optimization process of the aerodynamic noise of motor centrifugal fans. Therefore, it is necessary to develop a simulation evaluation method for the aerodynamic noise of self-ventilated motor fans with a short calculation period, small error (±3 dBA), and strong applicability. Summary of the Invention

[0007] The purpose of the present invention is to provide a simulation evaluation method for the aerodynamic noise of self-ventilated motor fans with a short calculation period and strong applicability.

[0008] The present invention is implemented by the following technical solutions: A simulation evaluation method for the aerodynamic noise of self-ventilated motor fans includes the following steps:

[0009] 1) Steady-state flow field analysis of the motor and data extraction of the flow field simulation results at various speeds and the corresponding flow field simulation result data: How to perform steady-state flow field analysis of the motor and data extraction belongs to the common knowledge of those skilled in the art. Specifically, the finite volume method is used for the steady-state flow field analysis of the motor, and its calculation process includes the following steps: a. Simplify the motor geometric model and establish a flow field model: The original model has many components. After removing unnecessary features such as rounded corners and chamfers and merging adjacent components, the simplified solid model of the motor includes, but is not limited to, the shaft, fan, rotor core, bars-end rings or permanent magnets, stator core, coils, drive-end cover, non-drive-end cover, and frame, or some of the above models (for example, for a fully enclosed motor, the simplified model only includes the fan, end covers at both ends, and the frame). After obtaining the simplified structural model, extract the fluid domain and appropriately extend the fluid domain at the inlet and outlet. The extension distance is comparable to the motor size. Since the rotor rotates at a certain speed, the part of the fluid domain that encloses the solid rotating region needs to be segmented and given a rotational load. At this time, the fluid domain is generally divided into two parts: a stationary region and a rotating region, and each part contains a certain number of sub-fluid domains; b. After obtaining the flow field model, use unstructured grids to divide the grids: After obtaining the flow field model, use unstructured grids. By specifying the volume grid and surface grid sizes of each part of the computational domain, the division parameters of the inlet and outlet boundary layers, and the encryption parameters at curved or narrow regions, a computational domain grid model with good grid quality can be obtained. The orthogonality of all grids should be greater than 1E-3, and the average value should be greater than 0.75; c. After the grid division is completed, perform the solution settings. After the settings are completed, start the solution until the calculation converges: Specifically, how to perform the solution settings belongs to the common knowledge of those skilled in the art: Specify the material model and turbulence model to be solved (such as ideal gas and k-ε turbulence model), given loads and boundary conditions, mainly including but not limited to speed, inlet and outlet pressure boundaries, etc., specify the solution format (such as coupled solution, second-order accuracy, etc.), and set the convergence judgment criteria (such as the flow rate change rate in one calculation step is less than 1E-4). After the above settings are completed, start the solution until the calculation converges;

[0010] 2) Evaluation of sample data: Given a set of initial sample data consisting of M sample points (sample points are sample points composed of different motors, different fans, or different speeds), each sample point consists of N input parameters and one output parameter. The N input parameters are the speed and the corresponding flow field simulation result data extracted in step 1), and the 1 output parameter is the experimental value of the aerodynamic noise power level of the fan of the same type of motor at the corresponding speed (the experimental value can also be called the test accumulation value, that is, the noise power level value collected during the noise test of the same type of motor fan in the past). After normalizing each input parameter of this set of samples to form an array A, the array A is M rows * N columns. Solve the Euclidean distance between two samples in the array A to form an array B. The maximum Euclidean distance in the N-dimensional space composed of the input variables after normalizing the sample input parameters is The sample evaluation criteria are as follows: In the formula: Mean(B) is the average value of array B, Min(B) is the minimum value of array B, Max(B) is the maximum value of array B. When the sample data does not meet the sample evaluation criteria of formula (1), add, replace or remove sample points until the sample evaluation criteria of formula (1) are met;

[0011] 3) Establish a surrogate model and evaluate it: Select a method for constructing a surrogate model (such as elliptic basis neural network, polynomial response surface, Kriging, support vector regression, radial basis function and its improved methods) and evaluate the constructed surrogate model using the LOO cross-validation method, that is, remove one sample point data, use the remaining sample point data to construct a surrogate model, and then use the model to predict at the removed sample point and calculate the estimated error with the output parameter of this sample point. When the estimated errors of the motor aerodynamic noise at each sample point are all within [-3, 3] dBA and the fitting degree of the final surrogate model should be above 0.9, the surrogate model passes the evaluation; otherwise, reselect other methods to construct the surrogate model;

[0012] 4) Noise evaluation: Substitute the flow field simulation results at each rotational speed extracted in step 1) into the evaluated surrogate model to obtain the predicted value L of the aerodynamic noise sound power level of the motor fan W Then evaluate according to the formula where is the limit value of the aerodynamic noise sound power level of the motor fan at the corresponding rotational speed, △L W is the error of the aerodynamic noise sound power level of the motor fan at the corresponding rotational speed, max(△L W ) is the maximum value of the errors of the aerodynamic noise sound power levels of the motor fan at each rotational speed. When max(△L W ) ≤ 0, the noise evaluation passes; when max(△L W ) > 0, the noise evaluation fails, and the design scheme needs to be improved until the noise evaluation passes.

[0013] The beneficial effects produced by the present invention are as follows: The present invention provides a method for simulating and evaluating the aerodynamic noise of a self-ventilated motor fan, which is particularly applicable to the centrifugal fan of a self-ventilated motor. This method is simple and easy to implement. By combining processes such as the steady-state flow field analysis and calculation of the motor, sample data evaluation, surrogate model establishment and noise evaluation, compared with the currently used finite element method and general empirical formula method respectively, it effectively solves the contradiction between fast calculation and high accuracy, improves both the simulation and evaluation speed of the motor fan noise and maintains the rationality and accuracy of the evaluation results, shortens the simulation and evaluation cycle of the aerodynamic noise of the fan in the design process of the self-ventilated motor (the evaluation of this scheme can be completed in about 1 day), and can be further used to carry out comparison and optimization of multiple fan schemes in the design stage of the motor aerodynamic noise. Description of the Drawings

[0014] Figure 1 is the flow chart for the simulation and evaluation of the aerodynamic noise of a self-ventilated motor fan;

[0015] Figure 2 is the flow chart for the establishment of the surrogate model;

[0016] Figure 3 is the simplified solid domain of the motor;

[0017] Figure 4 is the simplified fluid domain of the motor. Detailed Implementation Manner

[0018] Example 1: A method for simulating and evaluating the aerodynamic noise of a permanent magnet self-ventilated motor centrifugal fan, as Figure 1 shown, includes the following steps:

[0019] 1) Steady-state flow field analysis of the motor and extraction of the flow field simulation result data corresponding to each rotational speed and each rotational speed from the subsequent flow field simulation results: The steady-state flow field analysis of the motor uses the finite volume method, and its calculation process includes the following steps: a. Simplify the motor geometric model and establish a flow field model. After obtaining the simplified structural model, as Figure 3 shown, extract the fluid domain as Figure 4 shown; b. After obtaining the flow field model, use unstructured meshes to divide the meshes: After obtaining the flow field model, use unstructured meshes, and by specifying the volume mesh and surface mesh sizes of each part of the computational domain, the division parameters of the inlet and outlet boundary layers, and the encryption parameters at curved or narrow regions, obtain a computational domain mesh model with good mesh quality. All mesh orthogonality should be greater than 1E-3, and the average value should be greater than 0.75; c. After the mesh division is completed, perform the solution settings. After the settings are completed, start the solution until the calculation converges: Specifically, how to perform the solution settings is well-known to those skilled in the art: Specify the material model and turbulence model for the solution (such as ideal gas and k-ε turbulence model), given loads and boundary conditions, mainly including but not limited to rotational speed, inlet and outlet pressure boundaries, etc., specify the solution format (such as coupled solution, second-order accuracy, etc.), set the convergence judgment criteria (such as the flow rate change rate in one calculation step is less than 1E-4). After the above settings are completed, start the solution until the calculation converges; Through post-processing and related calculations, obtain the flow field simulation result data, that is, the motor cyclic volume flow rate Q m corresponding to each rotational speed ω and each rotational speed ω, the fan cyclic volume flow rate Q fs the increase in total pressure from the fan inlet to the outlet P a the increase in static pressure from the fan inlet to the outlet P s the aerodynamic drag torque M received by the fan during rotation fs the effective total pressure efficiency of the fan in the motor the total pressure efficiency of the fan Among them: extract the motor cyclic volume flow rate Q m When doing so, take the inlet and outlet or the outlet of the motor, with the unit of m 3 / s, extract the fan cyclic volume flow rate Q fs When doing so, an integral cross-section should be intercepted within the fan flow passage, with the unit of m 3 / s, extract the total pressure increase amount P from the fan inlet to the outlet a and the static pressure increase amount P s When doing so, the integral cross-sections at the inlet and outlet should be selected perpendicular to the main flow direction of the air flow, about 2 mm away from the inlet and outlet boundaries of the fan flow passage, with the unit of Pa, extract the aerodynamic resistance moment M received during the rotation of the fan fs When doing so, select all the rotating surfaces of the fan as the integral surfaces, and the moment direction is axial, with the unit of N*m;

[0020] 2) As Figure 2 shown, sample data evaluation: Given a set of initial sample data consisting of 68 sample points (too many data to list, this specific implementation manner is only an example to illustrate this simulation evaluation method, and the actual number of sample points depends on the motor type and the fan type. To ensure the sufficiency of the samples, it is recommended that the number of sample points be greater than 5N), each sample point consists of 8 input parameters and one output parameter. The 8 input parameters are the rotational speed extracted in step 1) and the flow field simulation result data corresponding to the rotational speed, and 1 output parameter is the experimental value of the aerodynamic noise power level of the same type of motor fan at the corresponding rotational speed. After normalizing each input parameter of this set of samples, an array A is formed. Array A is 68 rows * 8 columns. Solve the Euclidean distance between two samples in array A to form an array B (with a length of 2278). The sample evaluation criteria are as follows: In the formula: Mean(B) is the average value of array B, Min(B) is the minimum value of array B, Max(B) is the maximum value of array B. Calculate the relevant parameters according to formula (1), and the calculation results are shown in Table 1. After analysis, among the 68 sample points, 3 sample points do not meet the standard. Remove the three sample points, and the remaining 65 sample points form a new sample for re-evaluation. The evaluation results are shown in Table 2, which meet the standard of formula (1), and a surrogate model can be constructed;

[0021] Table 1 Initial sample evaluation

[0022]

[0023] Table 2 Updated sample evaluation

[0024]

[0025] 3) Establish a surrogate model and evaluate it: Use the elliptic basis neural network model and evaluate the constructed surrogate model using the leave-one-out cross-validation method, that is, remove one sample point data, use the remaining sample point data to construct the surrogate model, and then use this model to predict at the removed sample point and calculate the prediction error with the output parameter of this sample point (the experimental value of the aerodynamic noise sound power level corresponding to this sample point). The prediction errors of the motor aerodynamic noise at each sample point are all within the range of (-2.6, 2.7) dBA. Finally, the goodness of fit of the surrogate model is 0.981, and the surrogate model passes the evaluation;

[0026] 4) Noise evaluation: Substitute the flow field simulation results at each rotational speed extracted in step 1) into the evaluated surrogate model to obtain the predicted value L of the aerodynamic noise sound power level of the motor fan W Then, according to the formula for evaluation, where is the limit value of the aerodynamic noise sound power level of the motor fan at the corresponding rotational speed, △L W is the error of the aerodynamic noise sound power level of the motor fan at the corresponding rotational speed, and max(△L W ) is the maximum value of the aerodynamic noise sound power level errors of the motor fan at each rotational speed. The risk level classification of the motor fan aerodynamic noise exceeding the standard is shown in Table 3,

[0027] Table 3 Risk level classification of motor fan noise exceeding the standard

[0028]

[0029] The evaluation results and test results are shown in Table 4.

[0030] Table 4 Aerodynamic noise evaluation of motor fan (illustrated with three rotational speeds)

[0031]

[0032] Example 2: At the initial stage of the design of a certain self-ventilated motor fan, considering the manufacturing cost of the fan, conical front and rear discs are used, denoted as fan A. The aerodynamic noise of this fan is evaluated and the conclusion is that it fails; the front and rear discs of the fan are readjusted to be arc-shaped front and rear discs, denoted as fan B, and re-evaluated, still fails; the outer diameter and blade width of the fan are adjusted again to design fan C, and this fan passes the rapid evaluation of the aerodynamic noise performance of the fan. Finally, a comparison is made with the experimental results, and the calculation results are listed in Table 5.

[0033] Table 5 Comparison of risk levels of different fans' aerodynamic noise exceeding the standard

[0034] Fan <![CDATA[max(ΔL W ) / dBA]]> Evaluation conclusion Test result A 5.9 Exceed the standard —— B 2.8 High risk of exceeding the standard —— C -1.8 Low risk of exceeding the standard Not exceeding the standard

[0035] All research conducted in accordance with the design process, technical route, and evaluation method described in the patent application scope of the present invention without other creative labor should be included in the scope of this invention patent application. For example:

[0036] 1) Using the technical route of constructing an aerodynamic noise surrogate model for a self-ventilated motor with the flow field input and noise output in the present invention, increasing, decreasing, or replacing the input and output parameters may also achieve the purpose of quickly predicting the motor noise.

[0037] 2) During the steady-state analysis and data extraction of the motor flow field, dividing different meshes, using different turbulence models, and intercepting cross-sections at other positions may also achieve the purpose of obtaining the input parameters of the surrogate model.

[0038] 3) Using different methods (including polynomial response surface, Kriging, support vector regression, radial basis function, and their improved methods, etc.) may also achieve the purpose of constructing a surrogate model.

Claims

1. A method for simulating and evaluating the aerodynamic noise of a self-ventilated motor fan, characterized in that It includes the following steps: 1) Steady-state flow field analysis of the motor and extraction of the flow field simulation results data corresponding to each rotational speed and each rotational speed from the subsequent flow field simulation results; 2) Sample data evaluation: Given a set of initial sample data consisting of M sample points, each sample point is composed of N input parameters and one output parameter. The N input parameters are the rotational speed and the flow field simulation results data corresponding to the rotational speed extracted in step 1), and the 1 output parameter is the experimental value of the aerodynamic noise sound power level of the fan of the same type of motor at the corresponding rotational speed. After normalizing each input parameter of this set of samples to form an array A, the array A is M rows * N columns. Solve the Euclidean distance between every two samples in the array A to form an array B. The maximum Euclidean distance in the N-dimensional space composed of the normalized input variables of the sample is The sample evaluation criteria are as follows: In the formula: Mean(B) is the average value of the array B, Min(B) is the minimum value of the array B, Max(B) is the maximum value of the array B. When the sample data does not meet the sample evaluation criteria of formula (1), add, replace or remove sample points until the sample evaluation criteria of formula (1) are met; 3) Establish a surrogate model and evaluate it: Select a method for constructing a surrogate model and evaluate the constructed surrogate model using the LOO cross-validation method, that is, remove one sample point data, use the remaining sample point data to construct a surrogate model, and then use this model to predict at the removed sample point and calculate the prediction error with the output parameter of this sample point. When the prediction errors of the motor aerodynamic noise at each sample point are all within [-3, 3] dBA and the goodness of fit of the final surrogate model should be above 0.9, the surrogate model passes the evaluation. Otherwise, re-select other methods to construct a surrogate model; 4) Noise evaluation: Substitute the flow field simulation results at each rotational speed extracted in step 1) into the surrogate model that has passed the evaluation to obtain the predicted value L of the aerodynamic noise sound power level of the motor fan W Then according to the formula for evaluation, where is the limit value of the aerodynamic noise sound power level of the motor fan at the corresponding rotational speed, △L W is the error of the aerodynamic noise sound power level of the motor fan at the corresponding rotational speed, max(△L W ) is the maximum value of the aerodynamic noise sound power level errors of the motor fan at each rotational speed. When max(△L W ) ≤ 0, the noise evaluation passes. When max(△L W ) > 0, the noise evaluation fails, and the design scheme is improved until the noise evaluation passes.

2. The method for simulating and evaluating the aerodynamic noise of a self-ventilated motor fan according to claim 1, wherein In step 1), the finite volume method is adopted for the steady-state flow field analysis of the motor, and its calculation process includes the following steps: a. Simplify the geometric model of the motor and establish a flow field model; b. After obtaining the flow field model, use an unstructured grid to divide the grid; c. After the grid division is completed, perform the solution settings. After the settings are completed, start the solution until the calculation converges.

3. A method for simulating and evaluating the aerodynamic noise of a self-ventilated motor fan according to claim 2, characterized in that, In step 2), M > 5N.

4. A method for simulating and evaluating the aerodynamic noise of a self-ventilated motor fan according to claim 3, characterized in that In step 2), N is 8, and the 8 input parameters are the 8 flow field simulation result data extracted in step 1), namely the ω motor speed, whose unit is rad / s, and the motor cyclic volume flow rate Q m , the fan cyclic volume flow rate Q fs , the total pressure increase from the fan inlet to the outlet P a , the static pressure increase from the fan inlet to the outlet P s , the aerodynamic drag torque M exerted on the fan during rotation fs , the effective total pressure efficiency of the fan in the motor The total pressure efficiency of the fan 5. A method for simulating and evaluating the aerodynamic noise of a self-ventilated motor fan according to claim 4, characterized in that, Extract the fan cyclic volume flow rate Q fs When obtaining the data, an integral cross-section should be intercepted within the fan flow path.

6. A method for simulating and evaluating the aerodynamic noise of a self-ventilated motor fan according to claim 5, characterized in that Extract the total pressure increase P from the fan inlet to the outlet a and the static pressure increase P from the fan inlet to the outlet s When doing so, the selected integral cross-sections at the inlet and outlet should be perpendicular to the main flow direction of the air flow.

7. A method for simulating and evaluating the aerodynamic noise of a self-ventilated motor fan according to claim 6, characterized in that, Extract the total pressure increase P from the fan inlet to the outlet a and the static pressure increase P from the fan inlet to the outlet s When doing so, the selected integration cross-sections at the inlet and outlet should be close to the inlet and outlet boundaries of the fan flow channel by 2 mm.

8. A method for simulating and evaluating the aerodynamic noise of a self-ventilated motor fan according to claim 7, characterized in that Motor cyclic volume flow rate Q m It is the cyclic volume flow rate at the inlet or outlet of the motor.

Images