Methods and systems for optimizing the aeroacoustic performance of cooling fan systems

Abstract

This invention discloses a method and system for optimizing the aeroacoustic performance of a cooling fan system, relating to the field of data processing technology. The method includes: real-time acquisition of multi-dimensional dynamic data during the operation of the cooling fan system; construction of aerodynamic state feature vectors and identification of flow topology states; establishment of a mapping relationship between flow topology units and sound source contributions; calculation of sound source contribution weights to extract dominant sound source types; analysis of their aerodynamic excitation spectrum distribution characteristics; calculation of spectral coupling risk values ​​using acoustic modal risk channels; and finally, output of an optimized speed control strategy. This invention solves the technical problems of existing technologies, such as difficulty in adapting to complex dynamic conditions, inability to accurately identify and quantify multi-source aerodynamic noise sources in real time, and lack of dynamic monitoring of multi-physics field coupling resonance risks. It achieves the technical effects of improving the adaptive optimization capability of the fan system's aeroacoustic performance under dynamic operating conditions, significantly reducing broadband and discrete noise, and enhancing system operational reliability.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, specifically to a method and system for optimizing the aeroacoustic performance of a cooling fan system. Background Technology

[0002] As a core component of thermal management, cooling fan systems are widely used in many fields such as automotive engine cooling, electronic equipment heat dissipation, air conditioning ventilation, and industrial machinery. Cooling fan systems face increasingly severe performance challenges. Traditional optimization design of fan systems mainly focuses on their aerodynamic performance, such as air volume, air pressure, and efficiency. However, in practical applications, especially in scenarios with strict noise control requirements such as passenger cars, data centers, and precision instruments, the aerodynamic noise generated by the fan during operation has become a key constraint affecting user experience, environmental comfort, and even system reliability. Aerodynamic noise is mainly generated by the interaction between fan blades and air during rotation. This includes rotational noise caused by the periodic sweeping of air by the blades, and broadband eddy noise caused by turbulent boundary layers and eddy shedding. Existing fan noise reduction methods mostly focus on optimizing a single physical field, such as changing the flow field characteristics by improving the blade profile, adding a flow guide structure, or modifying the installation angle, in order to reduce noise. However, cooling fans often face complex dynamic environments in actual operation, such as real-time changes in speed with load, inlet flow field distortion, and system impedance fluctuations. Dynamic factors can significantly change the internal flow topology of the fan, leading to drastic changes in the sound source characteristics. Static design is difficult to maintain low noise performance across the entire operating range. Furthermore, it is difficult to accurately construct the mapping relationship between the flow structure and far-field / near-field noise during operation, and it is impossible to quantify the contribution weight of different flow units to the overall noise in real time, resulting in a lack of targeted noise reduction measures. In addition, there is a lack of dynamic monitoring and early warning mechanisms for the risk of multi-physical field coupling between convection, solid-state, and acoustic fields.

[0003] Therefore, current technologies have technical problems such as difficulty in adapting to complex dynamic working conditions, inability to accurately identify and quantify multi-source aerodynamic noise sources in real time, and lack of dynamic monitoring of multi-physics field coupling resonance risks. Summary of the Invention

[0004] This application provides a method and system for optimizing the aeroacoustic performance of a cooling fan system, which solves the technical problems in the prior art that make it difficult to adapt to complex dynamic working conditions, cannot accurately identify and quantify the sources of multi-source aerodynamic noise in real time, and lacks dynamic monitoring of the risk of multi-physics field coupling resonance. It achieves the technical effects of improving the aeroacoustic performance of the fan system under dynamic operating conditions, significantly reducing broadband and discrete noise, and enhancing the reliability of system operation.

[0005] This application provides a method for optimizing the aeroacoustic performance of a cooling fan system. The method includes: during the operation of the cooling fan system, real-time acquisition of multi-dimensional dynamic data characterizing the fan's operating state, including speed signals, load change signals, inlet and outlet pressure differences, vibration signals, and broadband noise spectrum data; constructing an aerodynamic state feature vector based on the multi-dimensional dynamic data; performing flow topology state identification on the aerodynamic state feature vector to construct the flow structure distribution result under the current operating state; constructing a mapping relationship between flow topology units and sound source contributions based on the flow structure distribution result, and calculating the sound source contribution weights of different sound source types under the current operating state; extracting the dominant sound source type using the sound source contribution weights, and analyzing the aerodynamic excitation spectrum distribution characteristics corresponding to the dominant sound source type; constructing an acoustic modal risk channel for the cooling fan system, and calculating a spectral coupling risk value using the acoustic modal risk channel and the aerodynamic excitation spectrum distribution characteristics; and outputting an optimized speed control strategy based on the spectral coupling risk value.

[0006] In a possible implementation, the spectral coupling risk value is calculated using the acoustic modal risk channel and the aerodynamic excitation spectral distribution characteristics, including: performing band energy decomposition on the aerodynamic excitation spectral distribution characteristics to extract the energy density and frequency drift trend parameters of each discrete frequency band; analyzing the pre-stored system inherent modal frequency ranges, modal damping factors, and modal amplification coefficients in the acoustic modal risk channel to construct a modal sensitivity distribution curve; performing matching analysis between the center frequencies of each discrete frequency band and the modal sensitivity distribution curve to calculate the frequency overlap; inputting the energy density into the modal amplification coefficient to calculate the modal amplification energy value; calculating the frequency convergence factor based on the frequency drift trend parameters, whereby the frequency convergence factor characterizes the evolution rate of the driving excitation frequency approaching the modal sensitivity range; and calculating the spectral coupling risk value based on the frequency overlap, modal amplification energy value, and frequency convergence factor.

[0007] In a possible implementation, constructing an aerodynamic state feature vector based on the multidimensional dynamic data includes: performing time synchronization alignment processing on the speed signal, load change signal, inlet and outlet pressure difference signal, vibration signal, and broadband noise spectrum data to construct a unified sampling time sequence; extracting time-domain statistical features, frequency-domain energy distribution features, and phase correlation features based on the unified sampling time sequence to generate a multidimensional original feature set; performing correlation constraint mapping processing on the multidimensional original feature set to construct a coupled feature subspace characterizing the coupling relationship between speed disturbance and pressure difference response, and the coupling relationship between vibration response and noise energy; performing feature compression and normalization processing on the coupled feature subspace to output the aerodynamic state feature vector.

[0008] In a possible implementation, constructing an aerodynamic state feature vector based on the multidimensional dynamic data includes: performing time synchronization alignment processing on the speed signal, load change signal, inlet and outlet pressure difference signal, vibration signal, and broadband noise spectrum data to construct a unified sampling time sequence; extracting time-domain statistical features, frequency-domain energy distribution features, and phase correlation features based on the unified sampling time sequence to generate a multidimensional original feature set; performing correlation constraint mapping processing on the multidimensional original feature set to construct a coupled feature subspace characterizing the coupling relationship between speed disturbance and pressure difference response, and the coupling relationship between vibration response and noise energy; performing feature compression and normalization processing on the coupled feature subspace to output the aerodynamic state feature vector.

[0009] In a possible implementation, flow topology state identification is performed on the aerodynamic state feature vector to construct the flow structure distribution result under the current operating state. This includes: inputting the aerodynamic state feature vector into a pre-constructed flow topology discrimination model, which establishes a flow state mapping relationship based on the feature evolution patterns corresponding to different flow structures; using the flow topology discrimination model to perform feature pattern matching of the aerodynamic state feature vector to identify the feature response intensity corresponding to the tip leakage vortex enhancement state, shear layer stabilization state, and local separation recirculation state; calculating the topology proportion parameter based on the feature response intensity corresponding to each type of flow structure; and constructing the flow structure distribution result under the current operating state based on the topology proportion parameter.

[0010] In a possible implementation, the construction of the flow topology discrimination model includes: establishing a sample mapping dataset between aerodynamic state feature vectors and flow topology based on multidimensional dynamic data under historical operating conditions and corresponding flow structure calibration results; extracting feature evolution trajectory parameters characterizing different flow structure states from the sample mapping dataset, the feature evolution trajectory parameters including speed perturbation response gradient, pressure difference fluctuation amplitude growth rate, vibration frequency band offset, and noise spectrum energy concentration change rate; constructing a multidimensional discrimination boundary channel based on the feature evolution trajectory parameters, the multidimensional discrimination boundary channel being used to distinguish between tip leakage vortex enhancement-dominated state, shear layer instability-dominated state, and local separation backflow-dominated state; and constructing a flow topology discrimination model based on the multidimensional discrimination boundary channel.

[0011] In a possible implementation, the optimized speed control strategy is output based on the spectrum coupling risk value, including: determining whether the spectrum coupling risk value is within a preset risk level range; when the spectrum coupling risk value is within the preset risk level range, generating a corresponding speed offset; and adding the speed offset to the current operating speed to perform speed control.

[0012] In a possible implementation, before outputting the optimized speed control strategy, the following steps are included: determining the operating condition type of the cooling fan system based on the current load, inlet and outlet pressure difference, and vibration characteristics; configuring a speed offset coefficient based on the operating condition type; and using the speed offset coefficient to generate compensation for the optimized speed control strategy.

[0013] In a possible implementation, after executing the optimized speed control strategy, the data is continuously updated with multi-dimensional dynamic data to establish an updated incremental dataset. The updated incremental dataset is then used to verify and evaluate the optimized speed control strategy, and verification and evaluation feedback is established. Based on the verification and evaluation feedback, the optimized speed control strategy is corrected in real time.

[0014] This application also provides an aeroacoustic performance optimization system for a cooling fan system. The system includes: a data acquisition module for real-time acquisition of multi-dimensional dynamic data characterizing the fan's operating state during system operation, including speed signals, load change signals, inlet and outlet pressure differences, vibration signals, and broadband noise spectrum data; a state recognition module for constructing an aerodynamic state feature vector based on the multi-dimensional dynamic data, performing flow topology state recognition on the aerodynamic state feature vector, and constructing the flow structure distribution result under the current operating state; a weight calculation module for constructing a mapping relationship between flow topology units and sound source contributions based on the flow structure distribution result, and calculating the sound source contribution weights of different sound source types under the current operating state; a spectrum feature analysis module for extracting the dominant sound source type using the sound source contribution weights, and analyzing the aerodynamic excitation spectrum distribution characteristics corresponding to the dominant sound source type; a coupling risk calculation module for constructing an acoustic modal risk channel for the cooling fan system, and calculating a spectrum coupling risk value using the acoustic modal risk channel and the aerodynamic excitation spectrum distribution characteristics; and a control strategy output module for outputting an optimized speed control strategy based on the spectrum coupling risk value.

[0015] This application proposes a method and system for optimizing the aeroacoustic performance of a cooling fan system. The method involves real-time acquisition of multi-dimensional dynamic data during system operation; construction of aerodynamic state feature vectors and identification of flow topology; establishment of a mapping relationship between flow topology units and sound source contributions; calculation of sound source contribution weights to extract dominant sound source types; analysis of their aerodynamic excitation spectrum distribution characteristics; and calculation of spectral coupling risk values ​​using acoustic modal risk channels. Finally, an optimized speed control strategy is output. This addresses the technical problems of existing technologies, such as difficulty in adapting to complex dynamic conditions, inability to accurately identify and quantify multi-source aerodynamic noise in real time, and lack of dynamic monitoring for multi-physics field coupling resonance risks. The method achieves the technical effects of improving the adaptive optimization capability of the fan system's aeroacoustic performance under dynamic operating conditions, significantly reducing broadband and discrete noise, and enhancing system operational reliability. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of this disclosure, the accompanying drawings of the embodiments of this disclosure will be briefly described below. Flowcharts are used in this application to illustrate the operations performed by the system according to the embodiments of this application. It should be understood that the preceding or following operations are not necessarily performed precisely in sequence. Instead, various steps can be processed in reverse order or simultaneously as needed. Furthermore, other operations can be added to these processes, or one or more steps can be removed from these processes.

[0017] Figure 1 A schematic diagram of the aeroacoustic performance optimization method for a cooling fan system provided in an embodiment of this application.

[0018] Figure 2 A schematic diagram of the aeroacoustic performance optimization system structure of the cooling fan system provided in the embodiments of this application.

[0019] Figure labeling: Data acquisition module 10, state recognition module 20, weight calculation module 30, spectrum feature analysis module 40, coupling risk calculation module 50, control strategy output module 60. Detailed Implementation

[0020] To further illustrate the technical means and effects adopted by the present invention in order to achieve the intended purpose, the following detailed description is provided in conjunction with the accompanying drawings and preferred embodiments, based on the specific implementation methods, structures, features and effects of the present invention.

[0021] This application provides a method for optimizing the aeroacoustic performance of a cooling fan system, such as... Figure 1 As shown, the method includes:

[0022] Step S100: During the operation of the cooling fan system, multi-dimensional dynamic data characterizing the fan's operating status are collected in real time. The multi-dimensional dynamic data includes speed signal, load change signal, inlet and outlet pressure difference, vibration signal, and broadband noise spectrum data.

[0023] Preferably, the system collects multi-dimensional dynamic data characterizing the fan's operating status in real time during the cooling fan system's operation. This includes speed signals, load change signals, inlet and outlet pressure differences, vibration signals, and broadband noise spectrum data. Specifically, the speed signal refers to the number of revolutions per minute of the fan impeller, acquired through a Hall sensor, photoelectric encoder, or built-in feedback signal of the motor controller, used to determine the fan's basic operating conditions. The load change signal refers to the power or current change driving the fan motor, reflecting the aerodynamic load borne by the fan. When the radiator is blocked, causing an increase in system resistance or a change in air density, the load fluctuates accordingly, measured by a current / power sensor or torque sensor in the motor driver, used to capture sudden changes in external operating conditions. The inlet and outlet pressure difference refers to the static pressure difference between the fan inlet and outlet, measured by installing micro differential pressure transmitters at the fan inlet and outlet. The vibration signal is used to measure the fan's ability to overcome system resistance. Vibration signal refers to the acceleration / velocity fluctuations generated by aerodynamic imbalance, turbulence excitation, or mechanical resonance during the operation of the fan blades, motor, or mounting structure. It is acquired by placing acceleration sensors on the fan motor housing, bearing housing, or mounting bracket to reflect the strength of fluid-structure interaction. Wideband noise spectrum data refers to the distribution of sound pressure level radiated into the air by the fan during operation, including discrete noise from the blade passing frequency and its harmonics, as well as wideband noise caused by turbulence. It is acquired by placing free-field microphones at a certain distance (usually 1 meter) from the fan inlet or outlet and performing real-time spectrum analysis using a data acquisition card. It serves as the final performance evaluation indicator. The energy distribution of different frequency bands is used to infer the dominant sound source type. For example, low-frequency wideband bulges usually correspond to tip leakage vortices, and mid-frequency narrowband peaks usually correspond to wake shedding.

[0024] Step S200: Construct an aerodynamic state feature vector based on the multidimensional dynamic data, perform flow topology state identification on the aerodynamic state feature vector, and construct the flow structure distribution result under the current operating state.

[0025] Step S200 further includes performing time synchronization alignment processing on the speed signal, load change signal, inlet and outlet pressure difference signal, vibration signal, and broadband noise spectrum data to construct a unified sampling time sequence; extracting time-domain statistical features, frequency-domain energy distribution features, and phase correlation features based on the unified sampling time sequence to generate a multi-dimensional original feature set; performing correlation constraint mapping processing on the multi-dimensional original feature set to construct a coupled feature subspace characterizing the coupling relationship between speed disturbance and pressure difference response, and the coupling relationship between vibration response and noise energy; performing feature compression and normalization processing on the coupled feature subspace to output an aerodynamic state feature vector.

[0026] Preferably, the physical characteristics of the speed signal, load change signal, inlet and outlet pressure difference signal, vibration signal, and broadband noise spectrum data are different from those of the sensor, and their sampling frequencies and trigger times are not completely consistent. Time synchronization and alignment processing is performed on all signals and broadband noise spectrum data. Specifically, all signals are resampled to a common frequency based on the signal with the highest sampling frequency, eliminating signal transmission delays caused by different sensor installation positions, such as vibration sensors being attached to the housing and microphones being suspended in the air. Then, the speed value, load value, pressure difference value, vibration waveform segment, and noise spectrum segment collected at the same time are packaged into a data frame to construct a unified sampling time sequence.

[0027] Preferably, time-domain statistical features, frequency-domain energy distribution features, and phase correlation features are extracted from the unified sampling time sequence. The time-domain statistical features are calculated directly based on the waveform's time axis. For vibration / noise / pressure difference pulsation signals, the mean, root mean square, kurtosis, and peak factor are calculated, representing the DC component, energy magnitude, presence of an impact component, and the ratio of the maximum value to the effective value, respectively, to determine if there is a transient strong excitation. For speed and load signals, the mean, rate of change, and variance are calculated. The frequency-domain energy distribution features are obtained by converting the signal to the frequency axis using Fourier transform. For narrowband features, the amplitude at the blade passage frequency and its harmonics is extracted. For broadband features, the total energy in specific frequency bands such as the low-frequency turbulent noise band (0-500Hz) and the mid-frequency eddy noise band (500-1500Hz) is calculated. Phase correlation features refer to the order of signals. For example, the cross-correlation function is used to check whether the noise signal appears after the vibration signal occurs and to determine the transmission path. The phase angle between pressure difference pulsation and speed fluctuation is calculated to determine whether it is "synchronous" or "lagging". Finally, the time-domain statistical features, frequency-domain energy distribution features and phase correlation features are integrated to generate a multi-dimensional original feature set.

[0028] Preferably, the original multidimensional feature set may contain a large amount of redundancy. For example, the energy of vibration at 100Hz and the energy of noise at 100Hz may be highly correlated. Correlation constraint mapping is performed on these features, that is, physical prior knowledge is introduced to constrain the direction of feature fusion. Coupled feature subspaces representing the coupling relationship between rotational speed disturbance and pressure difference response, and the coupling relationship between vibration response and noise energy are constructed respectively. The input features include rotational speed disturbance characteristics such as rotational speed fluctuation rate and acceleration, and pressure difference response characteristics such as pressure difference mean, pulsation amplitude, and response delay. Canonical correlation analysis (CCA) is used to find a linear combination that satisfies the relationship between the two... The correlation coefficient of the input is the largest, thus determining the strongest coupling mode of pressure difference fluctuation caused by speed fluctuation. This coupling feature subspace represents the aerodynamic load response characteristics of the fan. Similarly, vibration features such as the vibration acceleration level of the input casing and the vibration amplitude at a specific frequency, and noise features such as the far-field sound pressure level and broadband energy, are used. Canonical correlation analysis (CCA) is used to find the quantitative relationship between structural vibration and noise radiation. This coupling feature subspace represents the sound radiation efficiency of the fan. If the vibration is large but the noise is small, it indicates that the structural sound radiation efficiency is low. If the vibration is small but the noise is large, it indicates that the aerodynamics directly radiates sound, such as eddy current shedding. Finally, the coupling feature subspace is compressed by principal component analysis or autoencoder to remove noise. Then, the ratio of its mean deviation to standard deviation is calculated by Z-score normalization. The compressed features are scaled to the same order of magnitude, such as the [0, 1] interval or a mean of 0 and a variance of 1, to prevent features of large magnitude from dominating features of small magnitude. Finally, the aerodynamic state feature vector is output.

[0029] Furthermore, step S200 also includes inputting the aerodynamic state feature vector into a pre-constructed flow topology discrimination model, wherein the flow topology discrimination model establishes a flow state mapping relationship based on the feature evolution patterns corresponding to different flow structures; using the flow topology discrimination model to perform feature pattern matching of the aerodynamic state feature vector, identifying the feature response intensity corresponding to the tip leakage vortex enhancement state, shear layer instability state, and local separation backflow state; calculating the topology proportion parameter based on the feature response intensity corresponding to various flow structures, and constructing the flow structure distribution result under the current operating state based on the topology proportion parameter.

[0030] Preferably, the aerodynamic state feature vector is input into a pre-constructed flow topology discrimination model and compared with the flow state mapping relationship established during its training based on the feature evolution patterns corresponding to different flow structures. Feature pattern matching of the aerodynamic state feature vector is performed to identify the corresponding states of enhanced tip leakage vortex, shear layer instability, and local separation and recirculation. Specifically, the enhanced tip leakage vortex state is due to the pressure difference on both sides of the tip clearance, where airflow is entrained from the pressure surface to the suction surface to form a vortex. Its characteristics include a broadband bulge in the vibration signal that is independent of the blade passage frequency, an increase in the energy concentration of the noise spectrum in a specific mid-frequency band, and an increase in pressure difference pulsation. The shear layer instability state is caused by the airflow separating... The interface between high-speed and low-speed fluids near the point of origin is unstable, generating large-scale vortex structures. These are characterized by significant peaks in vibrations and noise at specific frequencies, which may drift with changes in flow velocity. Local separation and recirculation occurs when airflow separates at the blade suction surface or hub, forming a recirculation zone. This can lead to rotational stall or surge, characterized by low-frequency, high-amplitude fluctuations in vibrations and noise. The flow topology discrimination model outputs characteristic response intensities for three states based on feature pattern matching results: tip leakage vortex response intensity, shear layer instability response intensity, and local separation and recirculation response intensity. These intensities represent the confidence level of the flow topology discrimination model in determining that the current state belongs to that category. The characteristic response intensities corresponding to each flow structure are normalized to a sum of 1, and then the topological proportion parameter for each flow structure is calculated, ultimately generating the flow structure distribution results under the current operating state.

[0031] Furthermore, step S200 also includes: establishing a sample mapping dataset between aerodynamic state feature vectors and flow topology based on multidimensional dynamic data under historical operating conditions and corresponding flow structure calibration results; extracting feature evolution trajectory parameters characterizing different flow structure states from the sample mapping dataset, the feature evolution trajectory parameters including speed disturbance response gradient, pressure difference fluctuation amplitude growth rate, vibration frequency band offset, and noise spectrum energy concentration change rate; constructing a multidimensional discrimination boundary channel based on the feature evolution trajectory parameters, the multidimensional discrimination boundary channel being used to distinguish between tip leakage vortex enhancement-dominated state, shear layer instability-dominated state, and local separation backflow-dominated state; and constructing a flow topology discrimination model based on the multidimensional discrimination boundary channel.

[0032] Preferably, multi-dimensional dynamic data from a large number of historical operating conditions are collected as input data, and the corresponding flow topology is calibrated through particle image velocimetry (PIV), computational fluid dynamics (CFD) simulation, or expert annotation. The calibration results of the flow structure are used as labels. For example, when PIV shows obvious and strong vortices at the blade tip, the input data at this time is labeled as the blade tip leakage vortex enhancement state, thereby establishing a sample mapping dataset between aerodynamic state feature vectors and flow topology. Then, feature evolution trajectory parameters representing different flow structure states are extracted from the sample mapping dataset, including speed disturbance response gradient, pressure difference fluctuation amplitude growth rate, vibration frequency band offset, and noise spectrum energy concentration change rate. Among them, the speed disturbance response gradient represents the degree of change of a certain feature response when the speed changes slightly. A large gradient indicates sensitivity to speed changes and may be close to the instability boundary. The pressure difference fluctuation amplitude growth rate represents the rate at which the pressure difference pulsation increases over time. The vibration frequency band offset represents the frequency drift of vibration energy in the main frequency band. The noise spectrum energy concentration change rate represents the concentration trend of noise energy in a specific frequency band.

[0033] Preferably, the three flow states have different distribution characteristics in the evolution parameter space. Specifically, in the tip leakage vortex enhancement state, the rotational speed disturbance response gradient is moderate, the pressure difference fluctuation growth rate increases, the vibration frequency band shifts slightly downward, and the noise energy concentration changes towards the mid-frequency range. In the shear layer instability state, the rotational speed disturbance response gradient is high, the pressure difference fluctuation growth rate oscillates, the vibration frequency band shifts periodically, and the noise energy concentration changes towards the narrow band. In the local separation and recirculation state, the rotational speed disturbance response gradient is low, the pressure difference fluctuation growth rate increases sharply, the vibration frequency band shifts significantly downward, and the noise energy concentration changes towards the low-frequency range. A multidimensional discriminative boundary channel based on feature evolution trajectory parameters is constructed using Support Vector Machine (SVM). This involves finding the optimal hyperplane in the enhanced feature space to maximize the sample margin between different classes. For nonlinearly separable cases, a kernel function is used to map to a high-dimensional space. The kernel function is a radial basis function with a kernel parameter of 0.1, a penalty coefficient of 10, and balanced class weights. The multidimensional discriminative boundary channel distinguishes between the dominant states of tip leakage vortex enhancement, shear layer instability, and local separation and backflow. Each dimension of the channel corresponds to a feature evolution parameter, and its boundary region corresponds to the critical region transitioning from one flow structure state to another. When the real-time aerodynamic state feature vector falls into a certain region, it is determined to be the corresponding flow state; falling into the boundary region indicates a state transition. Finally, the trained multidimensional discriminative boundary channel is encapsulated into a flow topology discriminative model, which can output the dominant state of the flow structure and its corresponding response intensity based on the real-time aerodynamic state feature vector and its evolution parameters.

[0034] Step S300: Construct a mapping relationship between flow topology units and sound source contributions based on the flow structure distribution results, and calculate the sound source contribution weights of different sound source types in the current operating state.

[0035] Preferably, the aerodynamic noise of the fan is generated by different flow structures such as vortices, separation zones, and shear layers as independent sound source units, each radiating noise and then superimposed. Specifically, aeroacoustic theory (such as the Lighthill analogy and the FW-H equation) is first used to establish a basic mapping relationship between the flow structure and sound radiation. Then, using the flow structure distribution results as input and the actual contributions of various sound sources calibrated by sound arrays or sound source separation (such as beamforming and acoustic holography) as output, a regression model is used to learn the mapping relationship. The loss function is defined as minimizing the error between the predicted sound source contribution and the calibrated sound source contribution, thereby correcting the basic mapping relationship to adapt to complex working conditions. Different sound source types include leakage vortex noise, shear layer noise, and separation flow noise corresponding to the flow topology unit. Each flow topology unit may generate multiple sound source types. Assuming that each flow unit contributes to each sound source, but with different contribution coefficients, the transfer matrix obtained based on historical data calibration is as follows:

[0036]

[0037] Among them, tip leakage vortices contribute 85% to leakage vortex noise, 10% to shear layer noise, and 5% to separated flow noise, and so on. The flow structure distribution at the current moment is obtained, and its product with the calibration transfer matrix is ​​calculated to determine the initial contribution weight. At the same time, the intensity difference of different flow units is considered. For example, the tip leakage vortex has a large proportion and high intensity, while the separated backflow intensity is weak. Intensity correction is introduced and normalized to obtain the final sound source contribution weight, which represents the proportion of different types of noise in the total noise under the current state. Then, broadband noise spectrum data is collected simultaneously, and the noise components are separated using prior knowledge and the energy proportion of each component of the measured noise is calculated. The calculation error is checked to ensure the accuracy of the sound source contribution weight.

[0038] Step S400: Extract the dominant sound source type using the sound source contribution weight, and analyze the aerodynamic excitation spectrum distribution characteristics corresponding to the dominant sound source type.

[0039] Preferably, the dominant sound source type under the current operating condition is identified from multiple sound sources, and the aerodynamic excitation spectrum distribution characteristics of this sound source are analyzed in depth. Specifically, based on the sound source contribution weight, the sources are arranged according to their weight, and the sound source type with the largest contribution weight is extracted as the dominant sound source type, representing the most important noise generation mechanism under the current operating condition. If the contribution weights of two sound sources are similar, it is determined to be a dual-dominant mode, and the spectrum characteristics of both sound sources need to be analyzed simultaneously. If the maximum sound source contribution weight is lower than a preset threshold of 0.4, it is determined to be a dispersed mode, and comprehensive spectrum analysis is required. Aerodynamic excitation refers to the unsteady pressure pulsation generated by the flow structure and acting on the blades or fluid, which is usually characterized by the pressure pulsation spectrum or equivalent sound source spectrum. Among them, broadband noise spectrum data directly reflects the spectrum of radiated noise, and vibration signal spectrum reflects the structural response. The system directly reflects aerodynamic excitation, with the pressure difference pulsation spectrum directly reflecting the aerodynamic excitation force. The flow topology-related spectrum is the sound source spectrum calculated based on the flow structure, thereby constructing a complete aerodynamic excitation spectrum distribution characteristics corresponding to the dominant sound source type. Then, for the dominant sound source type, the spectral component contributed by the sound source is separated from the total noise spectrum. Using the coherence function of the vibration signal and the noise signal, the spectral components strongly correlated with the dominant flow unit are extracted. Then, key feature parameters are extracted from the separated aerodynamic excitation spectrum, including global features such as total excitation energy, peak frequency, and peak amplitude of the sound source, energy features of low, medium, and high frequency bands, morphological features such as spectral centroid and bandwidth, and evolutionary features such as frequency drift rate and energy change rate. Finally, the dominant sound source type identifier, aerodynamic excitation spectrum distribution characteristics, and key spectral feature parameters are output and used as input data for calculating the spectral coupling risk value.

[0040] Step S500: Construct an acoustic modal risk channel for the cooling fan system, and calculate the spectral coupling risk value using the acoustic modal risk channel and the aerodynamic excitation spectrum distribution characteristics.

[0041] Preferably, acoustic modes refer to the inherent resonance characteristics of the acoustic space constituted by the fan system, including blades, volute, ducts, and cavities. When sound waves propagate in a closed or semi-closed space, standing waves are formed at certain specific frequencies, leading to a sharp amplification of sound pressure. Then, the modal parameters of the cooling fan system are calculated using acoustic theory, including the natural frequency range (the frequency range where resonance may occur, used to delineate the risk frequency band), the modal damping factor (the system's ability to dissipate vibrational energy, determining the height and width of the resonance peak), and the modal amplification factor (the factor by which sound pressure is amplified at resonance, used to quantify the severity of the risk). The relationship between the modal amplification factor and the modal damping factor is as follows: Based on the inherent frequency range, modal damping factor, and modal amplification factor, a complete acoustic modal risk channel is constructed, which is a high-risk area divided on the frequency axis. When the frequency of aerodynamic excitation falls into these areas, it may induce strong acoustic resonance, leading to a sharp amplification of noise or even structural damage.

[0042] Furthermore, step S500 also includes: performing frequency band energy decomposition on the aerodynamic excitation spectrum distribution characteristics to extract the energy density and frequency drift trend parameters of each discrete frequency band; analyzing the pre-stored system inherent modal frequency range, modal damping factor, and modal amplification coefficient in the acoustic modal risk channel to construct a modal sensitivity distribution curve; performing matching analysis between the center frequency of each discrete frequency band and the modal sensitivity distribution curve to calculate the frequency overlap; inputting the energy density into the modal amplification coefficient to calculate the modal amplification energy value; calculating the frequency convergence factor based on the frequency drift trend parameters, the frequency convergence factor being used to characterize the evolution rate of the driving excitation frequency approaching the modal sensitivity range; and calculating the spectral coupling risk value based on the frequency overlap, modal amplification energy value, and frequency convergence factor.

[0043] Preferably, a constant percentage bandwidth division (e.g., 1 / 3 octave) is used to decompose the aerodynamic excitation spectrum distribution characteristics into frequency band energy, resulting in multiple continuous discretized frequency bands. The energy density and frequency drift trend parameters of each discrete frequency band are extracted, i.e., the rate of change of the center frequency of the frequency band over time, to capture the evolution trend of the excitation frequency over time. Then, pre-stored system inherent modal frequency ranges, modal damping factors, and modal amplification coefficients are extracted from the acoustic modal risk channel. A single-order modal sensitivity distribution curve is constructed using a normal distribution function to describe the sensitivity distribution of each mode. Considering the superposition effect of multiple modes, the total sensitivity curve is taken as the envelope of the sensitivity of each mode. The center frequency of each discrete frequency band is matched with the modal sensitivity distribution curve to calculate... Frequency overlap refers to the degree to which the excitation frequency falls within the modal sensitivity range. Complete overlap occurs when the excitation frequency falls exactly at the center of the most sensitive mode; complete non-overlap occurs when the excitation frequency is far from all modes. Partial overlap, the degree of which is determined by modal sensitivity, is calculated by inputting the energy density into the modal amplification factor. Multiplying the energy density of each frequency band by the corresponding modal amplification factor yields the modal amplification energy value, representing the equivalent excitation energy after considering resonance amplification. This reflects the degree to which resonance exacerbates noise. If the excitation frequency is far from the mode, the modal amplification energy is approximately equal to the original energy; if the excitation frequency is close to the modal center, the amplified energy is much greater than the original energy. The frequency convergence factor represents the rate and direction of the excitation frequency's approach to the modal sensitivity range. It is calculated based on the frequency drift trend parameter. ,in, As the frequency convergence factor, For frequency drift trend parameters, These are normalization coefficients used to map the rate of change to the interval [0, 1]. +1 indicates a approaching mode. 1 indicates a mode far removed from the target mode, and 0 indicates no change.

[0044] Furthermore, step S500 also includes constructing the frequency overlap, modal amplification energy value, and frequency convergence factor of each discrete frequency band into a coupling strength factor for the corresponding frequency band; applying a damping correction weight to the coupling strength factor, wherein the damping correction weight is calculated based on the modal damping factor; weighting and summing the coupling strength factors of each frequency band after damping correction according to the frequency band energy ratio to construct an instantaneous coupling risk value; and performing time integration or moving average processing on the instantaneous coupling risk value within a preset time window to generate a spectral coupling risk value.

[0045] Preferably, the frequency overlap, modal amplification energy value, and frequency convergence factor of each discrete frequency band are integrated into a comprehensive risk value. This includes constructing a coupling strength factor for each frequency band and attenuating the coupling strength factor according to the modal damping factor, i.e., applying a damping correction weight to the coupling strength factor. The damping correction weight is calculated based on the attenuation of the modal damping factor. Finally, the coupling strength factors of each frequency band after damping correction are weighted and summed, where the weight is the energy proportion of each frequency band. An instantaneous coupling risk value is constructed, and the instantaneous coupling risk value is processed by time integration or moving average within a preset time window to eliminate instantaneous fluctuations and generate a spectral coupling risk value to accurately assess the current degree of danger and future danger trends.

[0046] Step S600: Optimize the rotational speed control strategy based on the spectral coupling risk value.

[0047] Step S600 further includes determining whether the spectrum coupling risk value is within a preset risk level range; when the spectrum coupling risk value is within the preset risk level range, generating a corresponding speed offset; and superimposing the speed offset onto the current operating speed to perform speed control.

[0048] Preferably, based on the calculated spectral coupling risk value, the severity of the current risk is determined, and a corresponding speed adjustment strategy is generated. This proactively avoids acoustic resonance risk by changing the fan speed, achieving noise reduction optimization. Specifically, based on historical engineering experience data and safety margins, the spectral coupling risk value is divided into three levels: low risk (e.g., 0~0.3), where the excitation frequency is far from the mode or the energy is extremely low; medium risk (e.g., 0.3~0.7), where there is some overlap but strong resonance has not yet been triggered; and high risk (e.g., 0.7~1), where the frequencies highly overlap and the energy is high, resulting in a high resonance risk. The calculated spectral coupling risk value is compared with a preset risk level range to determine the current risk level. When the risk value falls within the preset medium / high risk level range, a corresponding speed offset is generated based on the risk level and the current operating condition. The amount of speed adjustment indicates acceleration or deceleration. For high-risk levels, a preset base speed offset of -150 rpm is used to reduce the speed and change the excitation frequency, moving it away from the mode. For medium-risk levels, a preset base speed offset of -50 rpm is used to slightly reduce the speed and observe changes in risk. Based on the offset, the speed is proportionally adjusted according to the current risk value. The calculated speed offset is then superimposed on the current operating speed to obtain the target speed, and a speed control command is generated and output to the fan controller for execution. After speed adjustment, multi-dimensional dynamic data continues to be collected, and the risk value is recalculated to form a closed-loop control. If the risk value drops to a safe range, the new speed is maintained. Finally, the generated speed control command is encapsulated into a complete optimized speed control strategy, output to the actuator, and the strategy information is recorded, including at least the speed strategy identifier, target speed, adjustment amount, speed change rate, expected effect, effective time, and validity period.

[0049] Furthermore, before outputting the optimized speed control strategy in step S500, the process includes determining the operating condition type of the cooling fan system based on the current load, inlet and outlet pressure difference, and vibration characteristics; configuring a speed offset coefficient based on the operating condition type; and using the speed offset coefficient to generate compensation for the optimized speed control strategy.

[0050] Preferably, based on the current load, inlet / outlet pressure difference, and vibration characteristics, a multi-parameter fusion judgment is employed. This involves comprehensively considering load characteristics, speed characteristics, pressure difference characteristics, and vibration characteristics to determine the operating condition type of the cooling fan system. This includes, but is not limited to, high-load steady-state conditions, low-load steady-state conditions, transient acceleration conditions, transient deceleration conditions, critical stall conditions, and start / stop conditions. The aerodynamic characteristics and acoustic performance of the system differ significantly under different operating conditions, requiring different speed control strategies. The speed deviation coefficient is a dimensionless multiplier factor used to correct the base speed deviation. Based on theoretical analysis and extensive experimental data, a correlation between operating condition type and speed deviation is established. The mapping relationship of the shift coefficients is used to obtain the working condition-speed offset coefficient mapping table. Then, according to the identified working condition type, the corresponding speed offset coefficient is queried from the preset working condition-speed offset coefficient mapping table. Among them, high load steady-state working condition is 0.5~0.8, low load steady-state working condition is 1.0~1.2, transient acceleration working condition is 0.3~0.5, transient deceleration working condition is 0.3~0.5, critical stall working condition is 1.2~1.5, and start / stop working condition is 0. Then, the configured speed offset coefficient is applied to the speed offset calculation to perform working condition adaptive compensation on the optimized speed control strategy, and output the compensated target speed and the corresponding optimized speed control strategy.

[0051] Furthermore, step S500 also includes, after executing the optimized speed control strategy, continuously executing data updates of multi-dimensional dynamic data, establishing an updated incremental dataset, using the updated incremental dataset to verify and evaluate the optimized speed control strategy, establishing verification and evaluation feedback, and performing real-time correction of the optimized speed control strategy based on the verification and evaluation feedback.

[0052] Preferably, an optimized speed control strategy is implemented, and its effectiveness is continuously tracked and evaluated. A feedback correction mechanism is established to continuously improve the system's optimization capability over time. Specifically, after implementing the optimized speed control strategy, multi-dimensional dynamic data is continuously collected to construct an updated incremental dataset, including the time range of data updates and the content of data updates within each time window. Then, using the updated incremental dataset, the effectiveness of the implemented optimized speed control strategy is verified and quantitatively evaluated based on multiple dimensions such as risk reduction effect, acoustic improvement effect, heat dissipation impact, stability, and response time. The corresponding evaluation indicators are risk reduction rate, noise... The evaluation indicators, including reduction rate, airflow change rate, risk fluctuation coefficient, and time for risk to decrease to a safe level, are weighted and integrated to obtain a comprehensive evaluation value. The corresponding weight coefficients are 0.4, 0.3, 0.2, 0.05, and 0.05, respectively. The strategy effect is then divided into different levels to generate verification evaluation feedback. Finally, based on the verification evaluation feedback, the current optimized speed control strategy is dynamically adjusted, such as correcting the base offset, adjusting the operating condition compensation coefficient, and adjusting the risk level threshold. This improves the aeroacoustic performance adaptive optimization capability of the fan system under dynamic operating conditions, significantly reduces broadband and discrete noise, and enhances the system's operational reliability.

[0053] In the above text, refer to Figure 1 A method for optimizing the aeroacoustic performance of a cooling fan system according to an embodiment of the present invention is described in detail. Next, reference will be made to... Figure 2 A system for optimizing the aeroacoustic performance of a cooling fan system according to an embodiment of the present invention is described.

[0054] The aeroacoustic performance optimization system for a cooling fan system according to embodiments of the present invention addresses the technical problems in the prior art, such as difficulty in adapting to complex dynamic operating conditions, inability to accurately identify and quantify multi-source aerodynamic noise sources in real time, and lack of dynamic monitoring of multi-physics field coupling resonance risks. It achieves the technical effects of improving the aeroacoustic performance adaptive optimization capability of the fan system under dynamic operating conditions, significantly reducing broadband and discrete noise, and enhancing system operational reliability. Figure 2 As shown, the aeroacoustic performance optimization system for the cooling fan system includes: a data acquisition module 10, a state recognition module 20, a weight calculation module 30, a spectrum feature analysis module 40, a coupling risk calculation module 50, and a control strategy output module 60.

[0055] The data acquisition module 10 is used to acquire multi-dimensional dynamic data characterizing the fan's operating state in real time during the operation of the cooling fan system. The multi-dimensional dynamic data includes speed signals, load change signals, inlet and outlet pressure differences, vibration signals, and broadband noise spectrum data. The state recognition module 20 is used to construct an aerodynamic state feature vector based on the multi-dimensional dynamic data, perform flow topology state recognition on the aerodynamic state feature vector, and construct the flow structure distribution result under the current operating state. The weight calculation module 30 is used to construct the mapping relationship between flow topology units and sound source contributions based on the flow structure distribution result, and calculate the sound source contribution weights of different sound source types under the current operating state. The spectrum feature analysis module 40 is used to extract the dominant sound source type using the sound source contribution weights and analyze the aerodynamic excitation spectrum distribution characteristics corresponding to the dominant sound source type. The coupling risk calculation module 50 is used to construct an acoustic modal risk channel for the cooling fan system and calculate the spectrum coupling risk value using the acoustic modal risk channel and the aerodynamic excitation spectrum distribution characteristics. The control strategy output module 60 is used to output an optimized speed control strategy based on the spectrum coupling risk value.

[0056] The specific configuration of the coupling risk calculation module 50 will be described in detail below. The coupling risk calculation module 50 further includes: performing band energy decomposition on the aerodynamic excitation spectrum distribution characteristics to extract the energy density and frequency drift trend parameters of each discrete frequency band; analyzing the pre-stored system inherent modal frequency range, modal damping factor, and modal amplification coefficient in the acoustic modal risk channel to construct a modal sensitivity distribution curve; performing matching analysis between the center frequency of each discrete frequency band and the modal sensitivity distribution curve to calculate the frequency coincidence; inputting the energy density into the modal amplification coefficient to calculate the modal amplification energy value; calculating the frequency convergence factor based on the frequency drift trend parameters, whereby the frequency convergence factor characterizes the evolution rate of the driving excitation frequency approaching the modal sensitivity range; and calculating the spectral coupling risk value based on the frequency coincidence, modal amplification energy value, and frequency convergence factor.

[0057] The specific configuration of the coupling risk calculation module 50 will be described in detail below. The coupling risk calculation module 50 further includes: constructing coupling strength factors for each discrete frequency band by taking the frequency overlap, modal amplification energy value, and frequency convergence factor; applying damping correction weights to the coupling strength factors, wherein the damping correction weights are attenuated based on the modal damping factor; weighting and summing the damped coupling strength factors of each frequency band according to the frequency band energy ratio to construct an instantaneous coupling risk value; and performing time integration or moving average processing on the instantaneous coupling risk value within a preset time window to generate a spectral coupling risk value.

[0058] The specific configuration of the state recognition module 20 will be described in detail below. The state recognition module 20 further includes: performing time synchronization alignment processing on the speed signal, load change signal, inlet / outlet pressure difference signal, vibration signal, and broadband noise spectrum data to construct a unified sampling time sequence; extracting time-domain statistical features, frequency-domain energy distribution features, and phase correlation features based on the unified sampling time sequence to generate a multi-dimensional original feature set; performing correlation constraint mapping processing on the multi-dimensional original feature set to construct a coupled feature subspace characterizing the coupling relationship between speed disturbance and pressure difference response, and the coupling relationship between vibration response and noise energy; performing feature compression and normalization processing on the coupled feature subspace to output an aerodynamic state feature vector.

[0059] The specific configuration of the state recognition module 20 will be described in detail below. The state recognition module 20 further includes: inputting the aerodynamic state feature vector into a pre-constructed flow topology discrimination model, which establishes a flow state mapping relationship based on the feature evolution patterns corresponding to different flow structures; using the flow topology discrimination model to perform feature pattern matching of the aerodynamic state feature vector, identifying the feature response intensity corresponding to the tip leakage vortex enhancement state, shear layer instability state, and local separation recirculation state; calculating the topology proportion parameter based on the feature response intensity corresponding to various flow structures, and constructing the flow structure distribution result under the current operating state based on the topology proportion parameter.

[0060] The specific configuration of the state recognition module 20 will be described in detail below. The state recognition module 20 further includes: establishing a sample mapping dataset between aerodynamic state feature vectors and flow topology based on multi-dimensional dynamic data under historical operating conditions and corresponding flow structure calibration results; extracting characteristic evolution trajectory parameters representing different flow structure states from the sample mapping dataset, wherein the characteristic evolution trajectory parameters include the speed disturbance response gradient, the pressure difference fluctuation amplitude growth rate, the vibration frequency band offset, and the noise spectrum energy concentration change rate; constructing a multi-dimensional discrimination boundary channel based on the characteristic evolution trajectory parameters, wherein the multi-dimensional discrimination boundary channel is used to distinguish between the tip leakage vortex enhancement-dominated state, the shear layer instability-dominated state, and the local separation backflow-dominated state; and constructing a flow topology discrimination model based on the multi-dimensional discrimination boundary channel.

[0061] The specific configuration of the control strategy output module 60 will be described in detail below. The control strategy output module 60 further includes: determining whether the spectrum coupling risk value is within a preset risk level range; when the spectrum coupling risk value is within the preset risk level range, generating a corresponding speed offset; and adding the speed offset to the current operating speed to perform speed control.

[0062] The specific configuration of the control strategy output module 60 will be described in detail below. The control strategy output module 60 further includes: determining the operating condition type of the cooling fan system based on the current load, inlet / outlet pressure difference, and vibration characteristics; configuring a speed deviation coefficient based on the operating condition type; and using the speed deviation coefficient to generate and compensate for an optimized speed control strategy.

[0063] The specific configuration of the control strategy output module 60 will be described in detail below. The control strategy output module 60 further includes: after executing the optimized speed control strategy, continuously updating multi-dimensional dynamic data, establishing an updated incremental dataset, using the updated incremental dataset to verify and evaluate the optimized speed control strategy, establishing verification and evaluation feedback, and performing real-time correction of the optimized speed control strategy based on the verification and evaluation feedback.

[0064] The aeroacoustic performance optimization system for the cooling fan system provided in this embodiment of the invention can execute the aeroacoustic performance optimization method for the cooling fan system provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the method.

[0065] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A method for optimizing the aeroacoustic performance of a cooling fan system, characterized in that, The method includes: During the operation of the cooling fan system, multi-dimensional dynamic data characterizing the fan's operating status are collected in real time. The multi-dimensional dynamic data includes speed signal, load change signal, inlet and outlet pressure difference, vibration signal, and broadband noise spectrum data. Based on the multidimensional dynamic data, an aerodynamic state feature vector is constructed, and flow topology state identification is performed on the aerodynamic state feature vector to construct the flow structure distribution result under the current operating state. Based on the flow structure distribution results, a mapping relationship between flow topology units and sound source contributions is constructed, and the sound source contribution weights of different sound source types in the current operating state are calculated. The dominant sound source type is extracted using the sound source contribution weight, and the aerodynamic excitation spectrum distribution characteristics corresponding to the dominant sound source type are analyzed. An acoustic modal risk channel for the cooling fan system is constructed, and the spectral coupling risk value is calculated using the acoustic modal risk channel and the aerodynamic excitation spectral distribution characteristics. The optimized speed control strategy is output based on the aforementioned spectrum coupling risk value; The calculation of the spectral coupling risk value using the acoustic modal risk channel and the aerodynamic excitation spectral distribution characteristics includes: The frequency band energy is decomposed based on the aerodynamic excitation spectrum distribution characteristics to extract the energy density and frequency drift trend parameters of each discrete frequency band; The pre-stored system inherent modal frequency range, modal damping factor, and modal amplification coefficient in the acoustic modal risk channel are analyzed to construct a modal sensitivity distribution curve; The center frequency of each discrete frequency band is matched with the modal sensitivity distribution curve to calculate the frequency overlap. The energy density is input into the modal amplification factor to calculate the modal amplification energy value; The frequency approach factor is calculated based on the frequency drift trend parameter. The frequency approach factor is used to characterize the evolution rate of the driving excitation frequency approaching the modal sensitive range. The spectral coupling risk value is calculated based on the frequency overlap, modal amplification energy value, and frequency convergence factor.

2. The method for optimizing the aeroacoustic performance of a cooling fan system as described in claim 1, characterized in that, The spectral coupling risk value is calculated based on the frequency overlap, modal amplification energy value, and frequency convergence factor, including: The frequency overlap, modal amplification energy value, and frequency convergence factor of each discrete frequency band are used to construct the coupling strength factor of the corresponding frequency band. A damping correction weight is applied to the coupling strength factor, and the damping correction weight is calculated based on the attenuation of the modal damping factor; The instantaneous coupling risk value is constructed by weighting and summing the coupling strength factors of each frequency band after damping correction according to the frequency band energy ratio. Within a preset time window, the instantaneous coupling risk value is processed by time integration or moving average to generate a spectral coupling risk value.

3. The method for optimizing the aeroacoustic performance of a cooling fan system as described in claim 1, characterized in that, Based on the aforementioned multidimensional dynamic data, an aerodynamic state feature vector is constructed, including: The rotation speed signal, load change signal, inlet and outlet pressure difference signal, vibration signal, and broadband noise spectrum data are time-synchronized and aligned to construct a unified sampling time sequence. Based on the unified sampling time series, time-domain statistical features, frequency-domain energy distribution features, and phase correlation features are extracted respectively to generate a multi-dimensional original feature set; The multidimensional original feature set is subjected to correlation constraint mapping processing to construct a coupled feature subspace characterizing the coupling relationship between rotational speed disturbance and differential pressure response, and the coupling relationship between vibration response and noise energy. Perform feature compression and normalization processing on the coupled feature subspace to output aerodynamic state feature vectors.

4. The method for optimizing the aeroacoustic performance of a cooling fan system as described in claim 3, characterized in that, Perform flow topology state identification on the aerodynamic state feature vector to construct the flow structure distribution result under the current operating state, including: The aerodynamic state feature vector is input into a pre-constructed flow topology discrimination model, which establishes a flow state mapping relationship based on the feature evolution patterns corresponding to different flow structures. The flow topology discrimination model is used to perform feature pattern matching of aerodynamic state feature vectors to identify the feature response intensity corresponding to the blade tip leakage vortex enhancement state, shear layer instability state, and local separation backflow state. The topology proportion parameter is calculated based on the characteristic response intensity corresponding to various flow structures, and the flow structure distribution result under the current operating state is constructed based on the topology proportion parameter.

5. The method for optimizing the aeroacoustic performance of a cooling fan system as described in claim 4, characterized in that, The construction of the flow topology discrimination model includes: A sample mapping dataset between aerodynamic state feature vectors and flow topology is established based on multidimensional dynamic data under historical operating conditions and corresponding flow structure calibration results. The characteristic evolution trajectory parameters representing different flow structure states are extracted from the sample mapping dataset. These characteristic evolution trajectory parameters include the speed perturbation response gradient, the growth rate of pressure difference fluctuation amplitude, the vibration frequency band offset, and the rate of change of noise spectrum energy concentration. A multidimensional discrimination boundary channel based on feature evolution trajectory parameters is constructed. The multidimensional discrimination boundary channel is used to distinguish between the tip leakage vortex enhancement dominant state, the shear layer instability dominant state, and the local separation backflow dominant state. A flow topology discrimination model is constructed based on the multidimensional discrimination boundary channel.

6. The method for optimizing the aeroacoustic performance of a cooling fan system as described in claim 1, characterized in that, The optimized speed control strategy is output based on the aforementioned spectral coupling risk value, including: Determine whether the spectral coupling risk value is within a preset risk level range; When the spectral coupling risk value is within a preset risk level range, a corresponding rotational speed offset is generated; The speed offset is added to the current operating speed to perform speed control.

7. The method for optimizing the aeroacoustic performance of a cooling fan system as described in claim 1, characterized in that, Before optimizing the speed control strategy, the following is also included: Determine the operating condition of the cooling fan system based on the current load, inlet and outlet pressure difference, and vibration characteristics. Configure the speed offset coefficient according to the operating condition type; The aforementioned speed offset coefficient is used to optimize the generation and compensation of the speed control strategy.

8. The method for optimizing the aeroacoustic performance of a cooling fan system as described in claim 1, characterized in that, After implementing the optimized speed control strategy, the system continuously updates the multi-dimensional dynamic data, establishes an incremental update dataset, uses the incremental update dataset to verify and evaluate the optimized speed control strategy, establishes verification and evaluation feedback, and performs real-time correction of the optimized speed control strategy based on the verification and evaluation feedback.

9. An aeroacoustic performance optimization system for a cooling fan system, characterized in that, The system is used to implement the aeroacoustic performance optimization method for the cooling fan system according to any one of claims 1 to 8, the system comprising: The data acquisition module is used to collect multi-dimensional dynamic data that characterizes the operating status of the fan in real time during the operation of the cooling fan system. The multi-dimensional dynamic data includes speed signal, load change signal, inlet and outlet pressure difference, vibration signal and broadband noise spectrum data. The state recognition module is used to construct an aerodynamic state feature vector based on the multidimensional dynamic data, perform flow topology state recognition on the aerodynamic state feature vector, and construct the flow structure distribution result under the current operating state. The weight calculation module is used to construct the mapping relationship between the flow topology unit and the sound source contribution based on the flow structure distribution result, and to calculate the sound source contribution weight of different sound source types in the current operating state. The spectrum feature analysis module is used to extract the dominant sound source type using the sound source contribution weight, and to analyze the aerodynamic excitation spectrum distribution characteristics corresponding to the dominant sound source type. The coupling risk calculation module is used to construct the acoustic modal risk channel of the cooling fan system and calculate the spectral coupling risk value using the acoustic modal risk channel and the aerodynamic excitation spectrum distribution characteristics. The control strategy output module is used to output an optimized speed control strategy based on the spectrum coupling risk value.

Images