A method for analyzing the time-space characteristics of flow-induced noise source of centrifugal pump / ventilator

Abstract

This invention discloses a method for analyzing the spatiotemporal characteristics of flow-induced noise sources in centrifugal pumps / fans, relating to the field of fluid machinery noise analysis technology. The method includes: S1: Determining the geometric dimensions of the main components of the centrifugal pump / fan, including the impeller, volute, and inlet / outlet pipes; performing a full-channel 3D model of the centrifugal pump / fan based on parametric modeling technology, retaining adjustable interfaces for component structural characteristic parameters; S2: Combining actual operating conditions, obtaining high-precision flow characteristics and noise source results based on the large eddy simulation turbulence model and eddy acoustics theory. This invention, based on eddy acoustics theory and frequency domain orthogonal decomposition, establishes a method for calculating and analyzing the spatiotemporal distribution characteristics and evolution laws of flow-induced noise sources in centrifugal pumps / fans. Its technical universality is reflected in three dimensions: for the discrete noise components unique to rotating machinery, it utilizes modal energy weight analysis obtained from frequency domain orthogonal decomposition calculations to accurately obtain the flow components and noise source structure and their spatiotemporal evolution laws at the dominant frequencies in a wide frequency range.

Description

Technical Field

[0001] This invention relates to the field of fluid machinery noise analysis technology, specifically to a method for analyzing the spatiotemporal characteristics of flow-induced noise sources in centrifugal pumps / fans. Background Technology

[0002] Centrifugal pumps / fans are widely used in nuclear power, ship propulsion, and new energy vehicles, and their flow-induced noise has become one of the bottlenecks restricting equipment performance improvement. Related research shows that the backflow effect at the impeller inlet and the dynamic-static interference in the volute tongue region easily induce unsteady flow phenomena including flow separation and vortex shedding. Simultaneously, the internal flow exhibits strong three-dimensionality and complex characteristics such as the interaction between dynamic and static components, and the vortex structures of different scales increase the difficulty of analyzing the dominant flow frequencies and modal characteristics. Furthermore, there is a strong correlation between flow characteristics and noise source characteristics; a thorough analysis of the flow characteristics and dominant frequency features within centrifugal pumps / fans is a prerequisite for revealing the spatiotemporal characteristics of the noise source.

[0003] Centrifugal pump / fan noise has broadband characteristics. Existing research mainly focuses on the macroscopic analysis of flow and noise characteristics. There is still little research on the microscopic characteristics of the internal flow field and the characteristics of the noise source at the dominant frequency. Furthermore, the understanding of the spatiotemporal characteristics and their relationship between flow characteristics and noise sources needs to be deepened. In response, we propose a method for analyzing the spatiotemporal characteristics of flow-induced noise sources in centrifugal pumps / fans. Summary of the Invention

[0004] To address the aforementioned technical problems, a method for analyzing the spatiotemporal characteristics of flow-induced noise sources in centrifugal pumps / fans is provided. This technical solution resolves the aforementioned technical issues.

[0005] To achieve the above objectives, the technical solution adopted by this invention is: a method for analyzing the spatiotemporal characteristics of flow-induced noise sources in centrifugal pumps / fans, comprising: S1: Determine the geometric dimensions of the main components of the centrifugal pump / fan. The main components include the impeller, volute, and inlet and outlet pipes. Based on parametric modeling technology, perform full-flow-channel 3D modeling of the centrifugal pump / fan to obtain a 3D model, retaining the adjustable interface of the component structural feature parameters; S2: Based on the large eddy simulation turbulence model and combined with actual operating conditions, the flow field numerical calculation is carried out using Fluent software. The flow channel geometry of the three-dimensional model is used as the reference to construct the mesh layer, and the mesh is refined in the high turbulence region of the impeller flow channel and the volute throat. At the same time, the boundary conditions are set to match the actual operating conditions, and the steady-state and transient simulation solutions of the flow field are completed to obtain the basic physical quantity data of the flow field. S3: Based on the acquired flow field fundamental physical quantity data, and using the vortex acoustics theory, the vortex acoustics source terms are solved in real time step by step by the user-defined function UDF, and the binary dataset containing three-dimensional coordinates, velocity, static pressure and vortex acoustic sources is output. S4: Preprocess the binary dataset; structured grid data can be used directly, while unstructured grid data is processed using inverse distance weighting. S5: The preprocessed data stream is analyzed in multiple dimensions using a parallel frequency domain orthogonal decomposition software package. The calculation results are visualized using visualization software as cloud maps, isosurfaces, and dynamic evolution animations. The characteristics of the flow field results and noise sources are analyzed. The spatiotemporal variation law of a single physical quantity at different frequencies and the spatiotemporal variation law and relationship between different physical quantities at the same frequency are obtained. The influence mechanism of transient processes such as flow separation and vortex structure changes on noise sources is revealed.

[0006] Preferably, the parametric modeling technology in step S1 supports the adjustment of key structural parameters such as impeller blade airfoil, volute base circle radius, and tongue clearance. The modeling file is output in STEP / IGES format, and the topological continuity of the geometric model is automatically checked during the modeling process.

[0007] Preferably, in step S2, boundary mass flow inlet boundary conditions and pressure outlet boundary conditions are set in the Fluent software. A multi-reference model is used to process the impeller rotation region, and transient rotor-stator data transfer is set. The calculated convergence residual is set to 5 × 10⁻⁶. -5 The transient solution is obtained using the large eddy simulation method based on the dynamic Smagorinsky-Lilly subgrid model, and the mesh resolution of the near-wall region of the flow field meets the requirement of y+≤1.

[0008] Preferably, step S3 specifically includes: Enable the user-defined function (UDF) and user-defined memory (UDM) modules in Fluent software, and preserve the velocity gradient tensor. UI and xj is used as an intermediate variable; Based on vortex sound source terms The vortex sound source term is defined using the macro DEFINE_EXECUTE_AT_END in UDF. The calculation results are stored in different UDF channels in stages. After each transient flow field time step, the binary result file is automatically exported to the specified calculation directory, and the file is accompanied by timestamp and operating condition identification information.

[0009] Preferably, in step S4, when calculating the weighted distance between spatial points using the inverse distance weighting method, uniform spatial points are created as mapping grids based on the volume of the centrifugal pump / ventilator; the flow field result file and the mapping grid are imported into Python, and the Euclidean distance between the mapping grid and the flow field grid is calculated using the function pydist2; a dynamic spatial near-point search distance L is defined, the value of L is adjusted in real time based on the local density of the flow field grid, all points in the mapping grid that meet the distance requirements with the flow field grid are recorded, and the weighted values ​​are calculated based on the weighting formula. After traversing all mapping points in parallel, the physical quantities of the flow field grid are assigned to the mapping grid. The weighting formula is: in For the k-th mapped grid value, To satisfy the spatial proximity search distance, the distance from the i-th flow field grid point to the mapped grid point, Let be the physical quantity value in the i-th flow field grid point, and n be the number of effective flow field grid points.

[0010] Preferably, in step S5, the multi-dimensional analysis obtains the frequency domain results by Fourier transforming the flow field information of different time series, and then combines the results of any plane or three-dimensional region in the mapping grid of different time series into a multi-dimensional time series containing different physical quantities.

[0011] Where X is the set of results at different times. For the combination of multidimensional physical quantities at time i; The time series is decomposed into M blocks, each containing N time series. Adjacent samples are superimposed with a 50%-70% overlap rate. The resulting time series after block decomposition is:

[0012] in Let be the time series of the m-th block, where s is the non-overlapping step size, 1≤m≤M; The Fast Fourier Transform (FFT) is performed on the block of samples to obtain:

[0013] in Let m be the set of frequencies in the k-th column of the m-th block; The samples calculated by Fourier transform are reordered and combined according to frequency to obtain the full-domain frequency domain dataset:

[0014] The orthogonal decomposition mode values ​​in the frequency domain are calculated using cross spectral density, and the eigenvalues ​​and eigenvectors are obtained through the eigenvalue decomposition formula.

[0015] Where C is the cross spectral density matrix. for The conjugate transpose of , in the complex field Transpose the vector and conjugate each element to obtain eigenvalues ​​and eigenvectors.

[0016] Preferably, the frequency domain orthogonal decomposition software package is computed using the message passing interface (MPI) parallel computing framework.

[0017] Preferably, in step S5, the frequency domain orthogonal decomposition results are imported into visualization post-processing software. The energy spectrum is used to analyze the structural characteristics and spatiotemporal evolution of corresponding physical quantities at different dominant frequencies. Simultaneously, by combining flow field variables and noise sources, the influence mechanism of transient processes such as flow separation and vortex structure changes on noise sources is revealed.

[0018] Preferably, the constructed flow field-sound source spatiotemporal characteristic correlation model employs a multi-dimensional feature fusion algorithm to fuse the spatiotemporal distribution characteristics of the flow field velocity field, pressure field, turbulence field, and noise source. The correlation degree between different flow structures and noise sources is quantified by normalizing the cross-relationships. The correlation coefficient is calculated using the following formula:

[0019] Where R is the normalized cross-correlation coefficient, cov(X,Y) is the covariance between the flow field physical quantity X and the noise source Y, and D(X) and D(Y) are the variances of X and Y, respectively; the contribution level of the flow structure to the noise source is divided based on the magnitude of the correlation coefficient, and the source analysis of the dominant noise source is performed.

[0020] Preferably, the construction steps for the spatiotemporal characteristic correlation model of the flow field and sound source are as follows: Based on the flow channel structure characteristics of the centrifugal pump fan, the entire flow channel is divided into key flow structure units such as the impeller blade leading edge and trailing edge, volute tongue, impeller-volute dynamic and static interference zone, and inlet and outlet diffuser section. The spatiotemporal evolution characteristics of the flow field physical quantities in each unit are extracted. Normalized cross-correlation coefficient is selected as the core correlation evaluation index to quantify the spatiotemporal correlation between a single or multiple flow structure units and the noise source; based on the modal energy contribution coefficient, the proportion of energy transfer from each flow field mode to the noise source mode at different frequencies is characterized. With frequency as the core dimension, the flow field characteristics of each flow structure unit as the input layer, and the spatiotemporal distribution characteristics of noise sources as the output layer, a multi-input-single-output nonlinear correlation model is established, and the least squares method is used to fit the correlation model.

[0021] Preferably, in step S6, the spatiotemporal localization of the noise source and the identification of the dominant frequency are performed from three dimensions: frequency, space and time. The comprehensive contribution of each flow structure unit to the noise source is calculated and classified by model, the contribution of multi-frequency coupling is analyzed, and the contribution results are visualized. By combining the contribution quantification results with the flow field-sound source spatiotemporal data, the core spatial region of the noise source and the moment of strong generation are located sequentially, and bound to the corresponding characteristic frequencies to form a three-dimensional location label. The candidate set of dominant frequencies was initially screened based on the modal energy ratio, and the primary and secondary dominant frequencies were confirmed by combining the contribution of flow structure. The spatiotemporal evolution characteristics of noise sources and the causes of flow structure under the dominant frequencies were analyzed.

[0022] Preferably, the parametric modeling technology in step S1 supports the adjustment of key structural parameters such as impeller blade airfoil, volute base circle radius, and tongue clearance. The modeling file is output in STEP / IGES format, and the topological continuity of the geometric model is automatically checked during the modeling process.

[0023] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention provides a method for analyzing the spatiotemporal characteristics of flow-induced noise sources in centrifugal pumps / fans. Based on eddy acoustics theory and frequency domain orthogonal decomposition, it establishes a method for calculating and analyzing the spatiotemporal distribution characteristics and evolution laws of flow-induced noise sources in centrifugal pumps / fans. Its technical universality is reflected in three dimensions: First, for the discrete noise components unique to rotating machinery, it uses modal energy weighting analysis obtained from frequency domain orthogonal decomposition to accurately obtain the source contribution of the dominant frequency. Second, through eddy dynamics and noise source dynamic analysis, it reveals the influence mechanism of transient processes such as flow separation and eddy structure changes on the noise source. Third, based on high-precision flow field calculations combined with the frequency domain orthogonal decomposition method, it enables spatiotemporal localization of the noise source at a reasonable computational cost. Attached Figure Description

[0024] Figure 1 This is a flowchart of the spatiotemporal characteristic analysis method for flow-induced noise sources from centrifugal pumps / fans in an embodiment of the present invention; Figure 2 This is a schematic diagram of the centrifugal pump / fan structure and a numerical calculation fluid domain provided in the embodiments of the present invention; Figure 3 This is the vortex source term calculation code provided in the embodiments of the present invention; Figure 4 This is a schematic diagram of the frequency domain orthogonal decomposition calculation process in an embodiment of the present invention; Figure 5 These are cloud images of the vortex sound source in the downstream channel at different times in an embodiment of the present invention; Figure 6 The frequency spectrum and homogenized energy of the frequency domain orthogonal decomposition results in this embodiment of the invention; Figure 7These are the first and second mode contour maps of the velocity and noise sources at twice the leaf frequency in this embodiment of the invention; Figure 8 The images show the modal cloud diagrams of the velocity and noise sources at twice the leaf frequency of the original blade and the biomimetic blade in the embodiments of the present invention. Detailed Implementation

[0025] The following description is intended to disclose the invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art.

[0026] Reference Figures 1 to 8 As shown, this embodiment takes a certain type of centrifugal pump as the research object. This centrifugal pump can be used for industrial water transportation, with a rated speed of 1480 r / min, a rated flow rate of 460 m³ / h, and a design head of 31 m. Its main components include an impeller, a volute, and inlet and outlet pipes. The impeller has 6 blades, the volute has a base circle radius of 397 mm, and the inlet and outlet pipes have a diameter of 250 mm. Based on the method described in this invention, the spatiotemporal characteristics of the flow-induced noise source of this centrifugal pump are calculated and analyzed. The specific implementation steps are as follows: S1: Three-dimensional parametric modeling of the entire flow channel of a centrifugal pump First, third-order Bézier curves were used to parametrically model the impeller disk and cover profiles of the centrifugal impeller. Fourth-order Bézier curves were used to parametrically model the impeller blade profiles at different blade heights. Using UG software combined with parametric modeling technology, a full-channel 3D model of the centrifugal pump was created. Key structural parameters, such as the impeller blade profile parameters, were set as adjustable parameters, retaining adjustable interfaces to facilitate subsequent comparative analysis of noise characteristics under different structural parameters. During the modeling process, the software automatically checked the topological continuity of the geometric model to ensure that the model was free of geometric defects (such as overlapping surfaces, gaps, missing chamfers, etc.). After modeling was completed, the model file was output in STEP format (file name: centrifugal_pump_model.STEP) for subsequent flow field numerical calculations.

[0027] S2: Numerical Flow Field Calculation Based on Large Eddy Simulation The STEP format 3D model obtained in S1 was imported into ANSYS ICEM software for mesh generation. A combination of structured and unstructured meshes was used to construct the mesh layered structure of the entire flow channel: the inlet and outlet pipes and impeller were constructed using hexahedral structured meshes, while the volute was constructed using unstructured tetrahedral meshes.

[0028] Mesh refinement was performed in the highly turbulent regions (volute tongue, impeller blade trailing edge, and impeller-volute dynamic-static interference region). The mesh refinement ratio (the ratio of the refined mesh size to the original mesh size) was set to 1:4 to ensure that the mesh resolution in the strong dynamic-static interference region met the computational requirements. Mesh refinement was also performed in the near-wall region of the flow field, controlling the mesh resolution to meet the requirement of y+≤1. The final total number of meshes generated across the entire flow channel was approximately 18 million. Mesh independence checks showed that the mesh met the mesh quality requirements for numerical computation (see...). Figure 2 ).

[0029] The divided mesh was imported into Fluent software and analyzed under the actual operating conditions of the fan (rated speed 1480 r / min, rated flow rate 460 m³ / h, medium is standard liquid water, density 1000 kg / m³, viscosity 0.001003). -5 Pa·s), set the parameters for numerical calculation of the flow field: Boundary conditions: The inlet is set as a mass flow inlet boundary condition, with a mass flow rate of 127.78 kg / s (corresponding to a rated flow rate of 460 m³ / h); the outlet is set as a pressure outlet boundary condition, with an outlet pressure of standard atmosphere (101325 Pa); the wall is set as a no-slip wall boundary condition.

[0030] Rotating region processing: The rotating region of the impeller is processed using a multiple reference frame (MRF) model. The impeller rotation speed is set to 1480 r / min and the rotation direction is clockwise. At the same time, transient rotor-stator data transfer is set up, and the transfer method is interpolation to ensure the continuity of flow field data in the rotor and stator regions.

[0031] Turbulence Model and Solution Settings: Transient solution was performed using the Large Eddy Simulation (LES) method based on the dynamic Smagorinsky-Lilly subgrid model, with the convergence residual set to 5 × 10⁻⁶. -5 The time step is set to 1×10. -4 The maximum number of iterations per time step is 50, and the total computation time step is 10,000 steps, ensuring coverage of 3 complete impeller rotation cycles (each rotation cycle is approximately 0.0405s), thus achieving full development of the flow field and capture of transient characteristics.

[0032] After setting up, start the Fluent software to perform steady-state and transient simulations. First, perform steady-state simulations until convergence, which serves as the initial condition for transient simulations. Then, switch to transient simulation mode to complete the numerical calculation of the transient flow field across the entire flow channel and save the transient flow field data in real time.

[0033] S3: Solving for vortex sound source terms and outputting datasets based on vortex acoustics theory In Fluent software, enable the User-Defined Function (UDF) and User-Defined Memory (UDM) modules, and write a UDF program (written in C language) to solve for the vortex sound source term in real time step by step. The specific steps are as follows: In the UDF program, preserve the velocity gradient tensor. UI and xj is used as an intermediate variable, and velocity gradient data in the transient flow field is extracted using built-in functions in Fluent software. Based on vortex acoustics theory, the macro DEFINE_EXECUTE_AT_END in the UDF is used to define the vortex acoustic source term. The calculation results of the vortex acoustic source term are stored hierarchically in different UDM channels. UDM1 stores three-dimensional coordinate data, UDM2 stores velocity data, UDM3 stores static pressure data, UDM4 stores turbulence intensity data, and UDM5 stores vortex acoustic source data. The specific implementation code is shown below. Figure 3 .

[0034] The automatic data export function is set up so that after each transient flow field time step, the data in the UDM channel is automatically exported as a binary dataset file and saved to the specified calculation directory. The file name format is "physical quantity_timestamp_condition identifier.bin", where the timestamp is the current calculation time step. For example, the condition identifier of the velocity variable in the 1000th time step is "velocity_1000_Q10", which facilitates the search and retrieval of subsequent data.

[0035] After the transient solution is completed, a total of 10,000 time-step binary datasets are output. Each dataset contains complete data on the three-dimensional coordinates, velocity, static pressure, turbulence intensity and vortex sound sources of the entire flow channel at the corresponding time step, providing a foundation for subsequent data preprocessing and analysis.

[0036] S4: Transient flow field data preprocessing Import the binary dataset output from S3 into Python software (version 3.9, equipped with NumPy, Pandas, and SciPy libraries) for data preprocessing. The specific steps are as follows: Grid data classification and processing: Structured grid data of inlet and outlet pipelines can be directly extracted and used without additional processing; unstructured grid data of impeller and volute are processed using the inverse distance weighting method to ensure data uniformity and continuity.

[0037] Mapping mesh construction: Based on the exported relevant physical quantity data files, uniform spatial points of resolution are created as mapping mesh according to the size of the region to be analyzed. To ensure the accuracy of data mapping, the ratio of the minimum flow field mesh size to the mapping mesh size is required to be at least 1:1.

[0038] Distance calculation and data assignment: The Euclidean distance between the mapped grid and the flow field grid is calculated using the pydist2 function in Python; the dynamic spatial near point search distance L is defined, with the initial value of L set to 20 mm. The distance is adjusted in real time based on the local density of the flow field grid (the higher the grid density, the smaller the value of L, with a minimum adjustment of 5 mm). All valid flow field grid points that meet the distance requirement (distance ≤ L) between the mapped grid points and the flow field grid are recorded.

[0039] Weighted calculation: Based on the following weighting formula, the physical quantity value of each mapped grid point is calculated, and all mapped points are traversed in parallel to complete the assignment of the physical quantities of the flow field grid to the mapped grid:

[0040] in, Let d be the value of the k-th mapping grid. i To satisfy the spatial near-point search distance, the distance from the i-th flow field grid point to the mapped grid point is zi, where zi is the physical quantity value in the i-th flow field grid point, and n is the number of effective flow field grid points (n≥5 in this embodiment to ensure calculation accuracy).

[0041] After preprocessing, a standardized mapped grid dataset is output for subsequent multidimensional analysis.

[0042] S5: Multidimensional Analysis and Visualization Based on Parallel Frequency Domain Orthogonal Decomposition The Parallel Frequency Domain Orthogonal Decomposition (POD) software package (written in Python and equipped with the MPI parallel computing framework) was used to perform multi-dimensional analysis on the preprocessed data stream in S4. The calculation process is as follows: Figure 4 The specific steps are as follows: Multidimensional time series construction: The flow field information of different time series is converted into frequency domain results through Fourier transform. The results of the volute tongue section and the impeller blade trailing edge section in the mapped grid of different time series are taken and combined into a multidimensional time series containing four physical quantities: velocity, static pressure, turbulence intensity, and vortex sound source.

[0043] Where X is the set of results at different times, N=5000 (total time step), and xi is the combination of multidimensional physical quantities at time i (containing the values ​​of 4 physical quantities).

[0044] Time series segmentation: The time series X is decomposed into M=10 blocks, each block containing N=500 time series data. Adjacent samples are superimposed with a 50% overlap rate. The segmented time series is as follows:

[0045] Where Xm is the time series of the m-th block, s is the non-overlapping step size (s=200), and 1≤m≤20.

[0046] Frequency domain transformation and construction of the full-domain frequency dataset: Perform a Fast Fourier Transform (FFT) on the samples of each block to obtain the frequency set of each block as follows:

[0047] Where Fm is the frequency set of the m-th block and k-th column; after reordering and combining all samples after Fourier calculation according to frequency order, the full-domain frequency domain dataset is obtained as follows:

[0048] Parallel computing: The frequency domain orthogonal decomposition software package adopts the message passing interface (MPI) parallel computing framework, which is distributed to 8 physical cores for independent computing.

[0049] Eigenvalue and eigenvector calculation: The orthogonal decomposition mode values ​​in the frequency domain are calculated using cross-spectral density. The eigenvalues ​​and eigenvectors are obtained through the following eigenvalue decomposition formula:

[0050] Where C is the cross spectral density matrix, To obtain the eigenvalues ​​and eigenvectors, we take the conjugate transpose of F in the complex field, transpose F, and then conjugate each element.

[0051] After the calculations are completed, the results are imported into Tecplot visualization software to generate contour plots, isosurfaces, and dynamic evolution animations of flow field physical quantities and noise sources. Contour plots are used to display the spatial distribution characteristics of a single physical quantity at different frequencies, such as the noise source distribution characteristics at different flow times (see...). Figure 4 Dynamic evolution animations are used to visually demonstrate the spatiotemporal variation patterns and coupling relationships of multiple physical quantities at the same frequency. Frequency domain orthogonal decomposition spectrum display (see...) Figure 6 Significant peaks are observed at the rotational frequency and twice the blade frequency. Visualization reveals the flow field parameters and spatiotemporal distribution characteristics of noise sources in the region between the volute tongue and the impeller blade trailing edge, and identifies the strip-like and point-like flow structures and noise source structures in this region (see...). Figure 7 By comparing the noise source distribution characteristics of the trailing edge of the original blade and the trailing edge of the biomimetic blade (see...), Figure 8 This method reveals the suppressive effect of the biomimetic trailing edge on noise sources. Construction of a spatiotemporal characteristic correlation model between the flow field and the sound source, and localization and identification of the noise source. Based on the frequency domain orthogonal decomposition results obtained in S5, a spatiotemporal characteristic correlation model of flow field and sound source is constructed to quantify the contribution of flow structure to noise source, and to complete the spatiotemporal localization of noise source and identification of dominant frequency. The specific steps are as follows: Construction of a spatiotemporal characteristic correlation model between flow field and sound source Flow structure unit division: Based on the flow channel structure characteristics of this centrifugal pump fan, the entire flow channel is divided into 4 key flow structure units, namely: impeller blade leading edge and trailing edge unit, volute tongue unit, impeller-volute dynamic and static interference zone unit, and inlet and outlet diffuser section unit. The spatiotemporal evolution characteristics of the flow field velocity field, pressure field, and turbulence field in each unit (such as velocity fluctuation amplitude, pressure pulsation frequency, turbulence intensity peak value, etc.) are extracted.

[0052] Selection and Calculation of Correlation Evaluation Indicators: The normalized cross-correlation coefficient is selected as the core correlation evaluation indicator to quantify the spatiotemporal correlation between a single or multiple flow structure units and the noise source. The correlation coefficient is calculated using the following formula:

[0053] Where R is the normalized cross-correlation coefficient (ranging from -1 to 1), cov(X,Y) is the covariance between the flow field physical quantity X (such as turbulence intensity) and the noise source Y, and D(X) and D(Y) are the variances of X and Y, respectively; the contribution level of the flow structure to the noise source is divided based on the magnitude of the correlation coefficient: R≥0.8 is a high contribution level, 0.5≤R<0.8 is a medium contribution level, and R<0.5 is a low contribution level.

[0054] Correlation Model Establishment and Fitting: A multi-input, single-output nonlinear correlation model is established, using frequency as the core dimension, the flow field characteristics of four flow structure units (velocity fluctuation amplitude, pressure pulsation frequency, and peak turbulence intensity) as the input layer, and the spatiotemporal distribution characteristics of noise sources (noise source intensity and spatial coordinates) as the output layer. The least squares method is used to fit the correlation model. During the fitting process, a first-order Taylor expansion approximation is used to locally linearize the nonlinear model, minimizing the sum of squared residuals.

[0055] in (where is the residual vector of the i-th sample), ensuring model fitting accuracy, with a goodness-of-fit R² ≥ 0.92.

[0056] Spatiotemporal localization of noise sources and identification of dominant frequencies Contribution Grading and Visualization: Using the aforementioned correlation model, the comprehensive contribution of the four flow structure units to the noise source at different frequency components was calculated and graded. The contribution of multi-frequency coupling was analyzed, and the contribution results were visualized using Tecplot software (using a combination of bar charts and contour plots). The results show that the volute tongue unit has the highest comprehensive contribution (average contribution of 0.85), belonging to the high contribution level; the impeller-volute dynamic-static interference zone unit is second (average contribution of 0.72), belonging to the medium-high contribution level; the impeller blade leading and trailing edge units (average contribution of 0.48) and the inlet and outlet diffuser section units (average contribution of 0.35) belong to the low contribution level.

[0057] Spatiotemporal localization of noise source: Combining the contribution quantification results with the flow field-sound source spatiotemporal data, the core spatial region and the moment of strong generation of noise source are located sequentially: The core spatial region of noise source is within 50 mm around the volute tongue (coordinate range: X∈[300,400] mm, Y∈[250,350] mm, Z∈[50,150] mm); The moment of strong generation of noise source is the instant when the impeller blade passes the tongue (a peak occurs every 0.0034 s, matching the impeller rotation period); The core spatial region and the moment of strong generation are bound with the corresponding characteristic frequency (296 Hz) to form a three-dimensional localization label (spatial coordinates + time + frequency).

[0058] Dominant frequency identification: Based on the modal energy percentage obtained in S5, a preliminary candidate set of dominant frequencies was screened (frequencies with modal energy percentage ≥ 5%, including 148Hz, 296Hz, and 444Hz); combined with the contribution of flow structure, the primary and secondary dominant frequencies were confirmed: the primary dominant frequency was 296Hz (modal energy percentage of 32.5%, corresponding to the high contribution of the volute tongue unit), and the secondary dominant frequency was 148Hz (modal energy percentage of 18.3%, corresponding to the contribution of the impeller-volute dynamic-static interference zone unit); the spatiotemporal evolution characteristics of the noise source and the flow structure inducing factors under the dominant frequency were analyzed, and it was found that the noise source under the 296Hz dominant frequency was mainly induced by the strong dynamic-static interference between the volute tongue and the impeller blade trailing edge, and the shedding of turbulent vortices, which is consistent with the visualization results. Through the specific implementation of this embodiment, the spatiotemporal characteristics of the flow-induced noise source of the centrifugal pump / ventilator were successfully calculated and analyzed. The core region, strong generation time and dominant frequency of the noise source were identified, and the contribution of different flow structures to the noise source was quantified. This solved the problem that traditional methods could not accurately capture the spatiotemporal evolution law of the noise source and could not quantify the correlation between flow structure and noise source. It provides accurate data support and technical basis for the noise control and structural optimization of centrifugal pumps / ventilators.

[0059] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention. The scope of protection claimed by the appended claims and their equivalents is defined.

Claims

1. A method for analyzing the spatiotemporal characteristics of flow-induced noise sources from centrifugal pumps / fans, characterized in that, include: S1: Determine the geometric dimensions of the main components of the centrifugal pump / fan. The main components include the impeller, volute, and inlet and outlet pipes. Based on parametric modeling technology, perform full-flow-channel 3D modeling of the centrifugal pump / fan to obtain a 3D model, retaining the adjustable interface of the component structural feature parameters; S2: Based on the large eddy simulation turbulence model and combined with actual operating conditions, the flow field numerical calculation is carried out using Fluent software. The flow channel geometry of the three-dimensional model is used as the reference to construct the mesh layer, and the mesh is refined in the high turbulence region of the impeller flow channel and the volute throat. At the same time, the boundary conditions are set to match the actual operating conditions, and the steady-state and transient simulation solutions of the flow field are completed to obtain the basic physical quantity data of the flow field. S3: Based on the acquired flow field fundamental physical quantity data, and using the vortex acoustics theory, the vortex acoustics source terms are solved in real time step by step by the user-defined function UDF, and the binary dataset containing three-dimensional coordinates, velocity, static pressure and vortex acoustic sources is output. S4: Preprocess the binary dataset; structured grid data can be used directly, while unstructured grid data is processed using inverse distance weighting. S5: The preprocessed data stream is analyzed in multiple dimensions using a parallel frequency domain orthogonal decomposition software package. The calculation results are visualized using visualization software as cloud maps, isosurfaces, and dynamic evolution animations. The characteristics of the flow field results and noise sources are analyzed. The spatiotemporal variation law of a single physical quantity at different frequencies and the spatiotemporal variation law and relationship between different physical quantities at the same frequency are obtained. The influence mechanism of transient processes such as flow separation and vortex structure changes on noise sources is revealed.

2. The method for analyzing the spatiotemporal characteristics of flow-induced noise sources from centrifugal pumps / fans according to claim 1, characterized in that: In step S2, boundary mass flow inlet boundary conditions and pressure outlet boundary conditions are set in the Fluent software. A multi-reference model is used to process the impeller rotation region, and transient rotor-stator data transfer is set. The calculated convergence residual is set to 5 × 10⁻⁶. -5 ; The transient solution is obtained using the large eddy simulation method based on the dynamic Smagorinsky-Lilly subgrid model, and the mesh resolution of the near-wall region of the flow field meets the requirement of y+≤1.

3. The method for analyzing the spatiotemporal characteristics of flow-induced noise sources from centrifugal pumps / fans according to claim 1, characterized in that, Step S3 specifically includes: Enable the user-defined function (UDF) and user-defined memory (UDM) modules in Fluent software, and preserve the velocity gradient tensor. UI and xj is used as an intermediate variable; Based on vortex sound source terms The vortex sound source term is defined using the macro DEFINE_EXECUTE_AT_END in UDF. The calculation results are stored in different UDF channels in stages. After each transient flow field time step, the binary result file is automatically exported to the specified calculation directory, and the file is accompanied by timestamp and operating condition identification information.

4. The method for analyzing the spatiotemporal characteristics of flow-induced noise sources from centrifugal pumps / fans according to claim 1, characterized in that: In step S4, when calculating the weighted distance between spatial points using the inverse distance weighting method, uniform spatial points are created as a mapping grid based on the volume of the centrifugal pump / fan. Import the flow field results file and the mapped mesh into Python. Calculate the Euclidean distance between the mapped mesh and the flow field mesh using the function `pydist2`. Define a dynamic spatial near-point search distance `L`, whose value is adjusted in real-time based on the local density of the flow field mesh. Record all points on the mapped mesh that satisfy the distance requirement with the flow field mesh. Calculate the weighted values ​​using a weighted formula. After traversing all mapped points in parallel, complete the assignment of the flow field mesh physical quantities to the mapped mesh. The weighted formula is: in For the k-th mapped grid value, To satisfy the spatial proximity search distance, the distance from the i-th flow field grid point to the mapped grid point, Let be the physical quantity value in the i-th flow field grid point, and n be the number of effective flow field grid points.

5. The method for analyzing the spatiotemporal characteristics of flow-induced noise sources from centrifugal pumps / fans according to claim 1, characterized in that: In step S5, the multidimensional analysis obtains the frequency domain results by Fourier transforming the flow field information of different time series. The results from any planar or three-dimensional region within the mapped grid of different time series are then combined to form a multidimensional time series containing different physical quantities. Where X is the set of results at different times. For the combination of multidimensional physical quantities at time i; The time series is decomposed into M blocks, each containing N time series. Adjacent samples are superimposed with a 50%-70% overlap rate. The resulting time series after block decomposition is: in Let be the time series of the m-th block, where s is the non-overlapping step size, 1≤m≤M; The Fast Fourier Transform (FFT) is performed on the block of samples to obtain: in Let m be the set of frequencies in the k-th column of the m-th block; The samples calculated by Fourier transform are reordered and combined according to frequency to obtain the full-domain frequency domain dataset: The orthogonal decomposition mode values ​​in the frequency domain are calculated using cross spectral density, and the eigenvalues ​​and eigenvectors are obtained through the eigenvalue decomposition formula. Where C is the cross spectral density matrix. for The conjugate transpose of , in the complex field Transpose the vector and conjugate each element to obtain eigenvalues ​​and eigenvectors.

6. The method for analyzing the spatiotemporal characteristics of flow-induced noise sources from centrifugal pumps / fans according to claim 1, characterized in that: The frequency domain orthogonal decomposition results are imported into visualization post-processing software. Through energy spectrum analysis, the structural characteristics and spatiotemporal evolution of corresponding physical quantities at different dominant frequencies are analyzed. At the same time, combined with flow field variables and noise sources, the influence mechanism of transient processes such as flow separation and vortex structure changes on noise sources is revealed.

7. The method for analyzing the spatiotemporal characteristics of flow-induced noise sources from centrifugal pumps / fans according to claim 1, characterized in that: The frequency domain orthogonal decomposition software package uses the message passing interface MPI parallel computing framework for computation.

8. The method for analyzing the spatiotemporal characteristics of flow-induced noise sources in centrifugal pump fans according to claim 1, characterized in that, Also includes: The constructed spatiotemporal characteristic correlation model of the flow field and sound source is as follows: Based on the flow channel structure characteristics of the centrifugal pump fan, the entire flow channel is divided into key flow structure units such as the impeller blade leading edge and trailing edge, volute tongue, impeller-volute dynamic and static interference zone, and inlet and outlet diffuser section. The spatiotemporal evolution characteristics of the flow field physical quantities in each unit are extracted. Normalized cross-correlation coefficient is selected as the core correlation evaluation index to quantify the spatiotemporal correlation between a single or multiple flow structure units and noise sources. Based on the modal energy contribution coefficient, the proportion of energy transfer from each flow field mode to the noise source mode at different frequencies is characterized; With frequency as the core dimension, the flow field characteristics of each flow structure unit as the input layer, and the spatiotemporal distribution characteristics of noise sources as the output layer, a multi-input-single-output nonlinear correlation model is established, and the least squares method is used to fit the correlation model.

9. The method for analyzing the spatiotemporal characteristics of flow-induced noise sources in a centrifugal pump fan according to claim 8, characterized in that: The spatiotemporal localization and dominant frequency identification of noise sources are performed from three dimensions: frequency, space and time. The comprehensive contribution of each flow structure unit to the noise source is calculated and classified through the model, the contribution of multi-frequency coupling is analyzed, and the contribution results are visualized. By combining the contribution quantification results with the flow field-sound source spatiotemporal data, the core spatial region of the noise source and the moment of strong generation are located sequentially, and bound to the corresponding characteristic frequencies to form a three-dimensional location label. The candidate set of dominant frequencies is initially screened based on the modal energy ratio, and the primary and secondary dominant frequencies are confirmed by combining the contribution of the flow structure. The spatiotemporal evolution characteristics of noise sources and the inducing factors of flow structures under the dominant frequency are analyzed.

10. The method for analyzing the spatiotemporal characteristics of flow-induced noise sources from centrifugal pumps / fans according to claim 1, characterized in that: The parametric modeling technology in step S1 supports the adjustment of key structural parameters such as impeller blade airfoil, volute base circle radius, and tongue clearance. The modeling file is output in STEP / IGES format, and the topological continuity of the geometric model is automatically checked during the modeling process.

Images