Sound absorption equivalent evaluation method and device based on regression model and bayesian optimization, and medium

Abstract

The application provides a sound absorption equivalent evaluation method and device based on a regression model and Bayesian optimization, and a medium. The method comprises the following steps: obtaining an original frequency spectrum of a target device without installing a sound absorption structure, and a real frequency spectrum of the target device with the sound absorption structure; constructing multiple groups of initial filter parameters, and performing convolution on the original frequency spectrum by using each group of filters to obtain corresponding equivalent frequency spectra; calculating real RMS difference values of each equivalent frequency spectrum and the real frequency spectrum in multiple frequency bands, taking the initial filter parameters as input and the real RMS difference values in the multiple frequency bands as labels, and training a regression model by using a loss function; taking the trained regression model as a proxy model, iteratively generating new filter parameters in a search space by using a Bayesian optimization algorithm, and obtaining optimal filter parameters that minimize a target function. The application realizes quantitative description of the real noise reduction capability of the sound absorption structure.

Description

Technical Field

[0001] This application relates to the field of acoustic testing technology, specifically to a method, device, and medium for evaluating the sound absorption equivalence based on regression models and Bayesian optimization. Background Technology

[0002] Currently, the performance evaluation of sound-absorbing materials typically employs the reverberation chamber method or standing wave tube method to test the sound absorption coefficient curve, characterizing its sound absorption performance at different frequencies. However, the sound absorption coefficient, as a physical property of the material itself, cannot fully reflect the noise reduction effect of the structure on the actual noise source in specific application scenarios (such as the amount of cotton, the installation location of the cotton, the spectral characteristics of the sound source, etc.).

[0003] Therefore, while traditional sound absorption coefficient curves have some guiding significance, they deviate from the noise reduction effect under actual working conditions and cannot serve as the sole basis for the design and optimization of sound absorption structures. Currently, there is a lack of a universal method to map the actual noise reduction effect of sound absorption structures to the frequency domain, making it difficult to effectively fit and visualize the performance of complex sound absorption structures. Summary of the Invention

[0004] Therefore, it is necessary to provide a sound absorption equivalence assessment method, computer equipment, and storage medium based on regression models and Bayesian optimization to address the aforementioned technical problems.

[0005] In a first aspect, embodiments of the present invention propose a sound absorption equivalence evaluation method based on a regression model and Bayesian optimization, the method comprising: Obtain the original spectrum of the target device without the sound-absorbing structure installed, and the actual spectrum of the target device with the sound-absorbing structure installed; Multiple sets of initial filter parameters are constructed, and the original spectrum is convolved using each set of filters to obtain the corresponding equivalent spectrum; Calculate the true RMS difference between each equivalent spectrum and the true spectrum in multiple frequency bands, and use the initial filter parameters as input, the true RMS difference in multiple frequency bands as labels, and train a regression model using a loss function; The trained regression model is used as a surrogate model, and the Bayesian optimization algorithm is used to iteratively generate new filter parameters in the search space to obtain the optimal filter parameters that minimize the objective function.

[0006] In some embodiments, the multiple sets of initial filter parameters are constructed based on the characteristics of the sound-absorbing structure and / or the characteristics of the true spectrum.

[0007] In some embodiments, the loss function is constructed using the difference between the predicted RMS value of each frequency band output by the regression model and the RMS value of the actual spectrum.

[0008] In some embodiments, the loss function is: in, This represents the true RMS value of the i-th frequency band. The predicted RMS value for the i-th frequency band output by the regression model.

[0009] In some embodiments, the objective function is constructed based on the weights of each frequency band and the corresponding predicted RMS values.

[0010] In some embodiments, the objective function is: Where n represents the number of frequency bands, This represents the weight of the i-th frequency band.

[0011] In some embodiments, the step of iteratively generating new filter parameters in the search space using a Bayesian optimization algorithm includes: Within a preset number of search iterations, the surrogate model is updated based on the prediction results of the regression model, and the new filter parameters are generated.

[0012] In some embodiments, an asymmetric search space design is used to search and generate new filter parameters.

[0013] In a second aspect, embodiments of the present invention provide a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, it implements the steps of the method described in the first aspect.

[0014] In some embodiments, the method includes: Obtain the original spectrum of the target device without the sound-absorbing structure installed, and the actual spectrum of the target device with the sound-absorbing structure installed; Multiple sets of initial filter parameters are constructed, and the original spectrum is convolved using each set of filters to obtain the corresponding equivalent spectrum; Calculate the true RMS difference between each equivalent spectrum and the true spectrum in multiple frequency bands, and use the initial filter parameters as input, the true RMS difference in multiple frequency bands as labels, and train a regression model using a loss function; The trained regression model is used as a surrogate model, and the Bayesian optimization algorithm is used to iteratively generate new filter parameters in the search space to obtain the optimal filter parameters that minimize the objective function.

[0015] Thirdly, according to an embodiment of the present invention, a computer-readable storage medium is provided thereon storing a computer program, which, when executed by a processor, implements the steps of the method described in the first aspect.

[0016] In some embodiments, the method includes: Obtain the original spectrum of the target device without the sound-absorbing structure installed, and the actual spectrum of the target device with the sound-absorbing structure installed; Multiple sets of initial filter parameters are constructed, and the original spectrum is convolved using each set of filters to obtain the corresponding equivalent spectrum; Calculate the true RMS difference between each equivalent spectrum and the true spectrum in multiple frequency bands, and use the initial filter parameters as input, the true RMS difference in multiple frequency bands as labels, and train a regression model using a loss function; The trained regression model is used as a surrogate model, and the Bayesian optimization algorithm is used to iteratively generate new filter parameters in the search space to obtain the optimal filter parameters that minimize the objective function.

[0017] Compared with existing technologies, the above-mentioned method, computer equipment, and storage medium have the following technical effects: They acquire the original spectrum of the target device without a sound-absorbing structure and the true spectrum of the target device with a sound-absorbing structure installed; they construct multiple sets of initial filter parameters and convolve the original spectrum using each set of filters to obtain the corresponding equivalent spectrum; they calculate the true RMS difference between each equivalent spectrum and the true spectrum in multiple frequency bands, using the initial filter parameters as input and the true RMS difference in multiple frequency bands as labels, and train a regression model using a loss function; they use the trained regression model as a surrogate model and iteratively generate new filter parameters in the search space using a Bayesian optimization algorithm to obtain the optimal filter parameters that minimize the objective function, thus achieving a quantitative description of the true noise reduction capability of the sound-absorbing structure. This method can be applied to any sound-absorbing structure and allows for horizontal comparison of different sound-absorbing structures at the same frequency under the same operating conditions, facilitating optimal solution selection. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of the structure of a computer device in one embodiment; Figure 2 This is a flowchart illustrating a sound absorption equivalence evaluation method based on regression model and Bayesian optimization in one embodiment. Figure 3 This is a schematic diagram of the original spectrum, the true spectrum, and the equivalent spectrum in one embodiment; Figure 4 This is a schematic diagram of the sound absorption coefficient curve in an example embodiment; Figure 5 This is a schematic diagram of the optimal filter in an example embodiment; Figure 6 This is a schematic diagram illustrating the fitting effect of the optimal filter in an example embodiment. Detailed Implementation

[0019] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are merely some examples or embodiments of the present invention. For those skilled in the art, the present invention can be applied to other similar scenarios based on these drawings without creative effort. Unless obvious from the context or otherwise specified, the same reference numerals in the drawings represent the same structures or operations.

[0020] As indicated in this invention and the claims, unless the context clearly indicates otherwise, the words "a," "an," "an," and / or "the" do not specifically refer to the singular and may also include the plural. Generally speaking, the terms "comprising" and "including" only indicate the inclusion of explicitly identified steps and elements, which do not constitute an exclusive list, and the method or apparatus may also include other steps or elements.

[0021] While this invention makes various references to certain modules in an apparatus according to embodiments of the invention, any number of different modules can be used and run on a computing device and / or processor. Modules are merely illustrative, and different aspects of the apparatus and methods may use different modules.

[0022] It should be understood that when a unit or module is described as "connected" or "coupled" to other units, modules, or blocks, it may refer to a direct connection or coupling, or communication with other units, modules, or blocks, or the presence of intermediate units, modules, or blocks, unless the context explicitly indicates otherwise. The term "and / or" as used herein may include any and all combinations of one or more of the related listed items.

[0023] The sound absorption equivalence assessment method based on regression models and Bayesian optimization provided in this application can be applied to, for example... Figure 1 In the computer device shown. For example... Figure 1 As shown, the computer device may include a processor 102 and a memory 104 for storing data. The processor 102 may be a processing device including, but not limited to, a microprocessor MCU or a programmable logic device FPGA.

[0024] Memory includes RAM and external storage, such as Flash and ROM.

[0025] The aforementioned computer equipment may further include a transmission device 106 for communication functions and an input / output device 108. Those skilled in the art will understand that... Figure 1 The structure shown is for illustrative purposes only and does not limit the structure of the computer device described above. For example, the computer device may also include components that are more... Figure 1The more or fewer components shown, or having the same Figure 1 The different configurations shown are illustrated.

[0026] like Figure 2 As shown, this embodiment of the invention provides a sound absorption equivalence evaluation method based on regression models and Bayesian optimization, which is applied to... Figure 2 Taking a computer device as an example, the explanation includes the following steps: S202: Obtain the original spectrum of the target device when no sound-absorbing structure is installed, and the actual spectrum of the target device when the sound-absorbing structure is installed; The target device is, for example, a range hood, and the sound-absorbing structure is, for example, a cotton-wrapped structure for the flue.

[0027] S204: Construct multiple sets of initial filter parameters, and use each set of filters to convolve the original spectrum to obtain the corresponding equivalent spectrum.

[0028] In some embodiments, the multiple sets of initial filter parameters are constructed based on the characteristics of the sound-absorbing structure and / or the characteristics of the true spectrum.

[0029] S206: Calculate the true RMS difference between each of the equivalent spectra and the true spectra in multiple frequency bands, and use the initial filter parameters as input, the true RMS difference of the multiple frequency bands as labels, and train the regression model using the loss function.

[0030] The regression model is used to learn the mapping relationship between the equivalent spectrum and the true RMS difference, which can evaluate the filter fitting ability and guide the Bayesian optimization algorithm to find the optimal filter.

[0031] S208: Using the trained regression model as a surrogate model, the Bayesian optimization algorithm is used to iteratively generate new filter parameters in the search space to obtain the optimal filter parameters that minimize the objective function.

[0032] To efficiently find the optimal filter in a high-dimensional feature space, a Bayesian optimization algorithm based on Gaussian processes was adopted.

[0033] Based on the above steps S202-S208, the original spectrum of the target device without the sound-absorbing structure and the real spectrum of the target device with the sound-absorbing structure are obtained; multiple sets of initial filter parameters are constructed, and the original spectrum is convolved using each set of filters to obtain the corresponding equivalent spectrum; the real RMS difference between each equivalent spectrum and the real spectrum in multiple frequency bands is calculated, and the initial filter parameters are used as input, the real RMS difference in multiple frequency bands is used as labels, and a regression model is trained using a loss function; the trained regression model is used as a surrogate model, and a Bayesian optimization algorithm is used to iteratively generate new filter parameters in the search space to obtain the optimal filter parameters that minimize the objective function, thereby realizing a quantitative description of the real noise reduction capability of the sound-absorbing structure. It can be applied to any sound-absorbing structure and can realize the horizontal comparison of different sound-absorbing structures at the same frequency under the same working conditions, which is convenient for scheme selection.

[0034] The specific solutions for steps S202-S208 will be described in detail below with reference to specific embodiments.

[0035] The dataset constructed in this embodiment contains several sets of initial filter parameters and their corresponding denoised equivalent spectra. The frequency range is 0-6400Hz, with intervals of 100Hz, totaling 65 frequency points. Therefore, each set of input features is a 65-dimensional vector. For ease of understanding, the following explanation uses a specific set of filter parameters as an example: like Figure 3 In the diagram, the red curve represents the original noise spectrum without cotton wrapping, the green curve represents the noise-reduced spectrum after cotton wrapping, and the blue curve represents the equivalent spectrum after filtering by one set of filters. It can be seen that this set of filters closely matches the original spectrum in the low-frequency range (below 100Hz) and mid-to-high frequency range (between 2000-4000Hz). However, the curve fitting error is significant in the 200-1500Hz and 4000-6400Hz ranges. Calculating the total RMS value within the 0-6400Hz range, the total noise of the actual spectrum after cotton wrapping is 64.10dB, while the fitted noise of the equivalent spectrum from the filters is 61.27dB, a difference of 2.83dB. Similarly, the difference between the actual noise of the actual spectrum after cotton wrapping and the RMS of the fitted noise from the filters in each frequency band is calculated. Clearly, the higher the feature dimension and the higher the label dimension, the higher the fitting accuracy. The filtering process can be implemented in Simcenter Testlab software or in Python using the firwin2 and lfiter libraries; details will not be elaborated here.

[0036] During the dataset construction phase, considering that sound-absorbing cotton material has significant sound absorption characteristics mainly in the high-frequency range, a filter with a "slide-type" characteristic was designed, such as... Figure 4As shown, the attenuation is smaller in the low-frequency band, gradually increases in the mid-frequency band, and larger in the high-frequency band. This data construction helps the regression model learn the mapping relationship between filter parameters and noise reduction effect more quickly, thereby improving the overall fitting accuracy and convergence speed.

[0037] In some embodiments, the loss function is constructed using the difference (mean squared error, MSE) between the predicted RMS values ​​of each frequency band output by the regression model and the true RMS values ​​of the true spectrum. The training objective of the regression model is the mean squared error, MSE.

[0038] The loss function is: in, This represents the true RMS value of the i-th frequency band. The predicted RMS value for the i-th frequency band output by the regression model.

[0039] The goal of the Bayesian optimization algorithm is to further search for an optimal set of 65-dimensional filter parameters based on the trained regression model, so that when applied to the original spectrum without cotton, the predicted RMS value of each frequency band is as close as possible to the actual RMS value with cotton, that is, the effect of fitting the real spectrum is the best.

[0040] Similar to the dataset construction, this embodiment preferably employs an asymmetric search space design to improve the efficiency of the Bayesian optimization algorithm when searching for the optimal filter. Specifically, a narrower search range (e.g., -3dB to 0dB) is set in the low-frequency band, while a larger search boundary (e.g., -20dB to -3dB) is set in the high-frequency band. This setting is based on the actual attenuation distribution of frequencies before and after cotton wrapping in actual tests, aiming to provide a reasonable exploration boundary for the Bayesian optimization algorithm, reduce meaningless search areas, improve the utilization of the parameter space, and thus improve the convergence rate of Bayesian optimization.

[0041] It should be noted that the Bayesian optimization search space and the "slide-type" filter construction in the regression model described in this invention are not restrictive solutions. They can be constructed based on the characteristics of the sound-absorbing structure and / or the characteristics of the real spectrum, and have strong versatility and scalability.

[0042] In the Bayesian optimization algorithm, a weighted objective function is defined, and the objective function is: Where n represents the number of frequency bands? This represents the weight of the i-th frequency band.

[0043] It's important to understand that the objective function does not contain label values; the goal of this function is to make the objective function... Close to 0.

[0044] In this embodiment, for example, the following weight configuration is used: The weight design principle is to reflect the strategy of "prioritizing total energy fitting, followed by mid-to-high frequency bands" to ensure that the optimization direction has practical effect and physical significance.

[0045] In this embodiment, within a preset number of search iterations, the surrogate model is updated based on the prediction results of the regression model, and new filter parameters are generated. After the Bayesian optimization algorithm completes the search, a set of optimal 65-dimensional filters will be obtained, such as... Figure 5 The example shown is the optimal filter obtained by the Bayesian optimization algorithm in this embodiment.

[0046] The optimal filter described above is applied to the original spectrum of the unwrapped material, and the filtered equivalent spectrum obtained by convolving with this filter is calculated. For example... Figure 6 As shown in the figure, the blue curve represents the equivalent spectrum generated after convolving the spectrum of the unwrapped cotton using the optimized filter, while the green curve represents the actual spectrum of the wrapped cotton. It can be seen that the overall curve fit is good, with a Total RMS difference of 0.01dB, and some deviation only exists in the 5000-5500Hz range. This filter meets the engineering accuracy requirements, and the fitting accuracy can be further improved by increasing the feature dimension, label dimension, and Bayesian iteration count.

[0047] It should be understood that although the steps in the flowchart above are shown sequentially as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowchart above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages in other steps.

[0048] In one embodiment, the present invention provides a computer device including a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, it implements the steps in any of the embodiments of the sound absorption equivalence evaluation method based on regression model and Bayesian optimization described above.

[0049] In one embodiment, the present invention provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps in any of the above embodiments of the sound absorption equivalence evaluation method based on regression model and Bayesian optimization.

[0050] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, or optical storage, etc. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.

[0051] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0052] The above embodiments merely illustrate several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A sound absorption equivalence evaluation method based on regression model and Bayesian optimization, characterized in that, The method includes: Obtain the original spectrum of the target device without the sound-absorbing structure installed, and the actual spectrum of the target device with the sound-absorbing structure installed; Multiple sets of initial filter parameters are constructed, and the original spectrum is convolved using each set of filters to obtain the corresponding equivalent spectrum; Calculate the true RMS difference between each equivalent spectrum and the true spectrum in multiple frequency bands, and use the initial filter parameters as input, the true RMS difference in multiple frequency bands as labels, and train a regression model using a loss function; The trained regression model is used as a surrogate model, and the Bayesian optimization algorithm is used to iteratively generate new filter parameters in the search space to obtain the optimal filter parameters that minimize the objective function.

2. The method according to claim 1, characterized in that, The multiple sets of initial filter parameters are constructed based on the characteristics of the sound-absorbing structure and / or the characteristics of the real spectrum.

3. The method according to claim 1, characterized in that, The loss function is constructed using the difference between the predicted RMS value of each frequency band output by the regression model and the RMS value of the actual spectrum.

4. The method according to claim 3, characterized in that, The loss function is: in, This represents the true RMS value of the i-th frequency band. The predicted RMS value for the i-th frequency band output by the regression model.

5. The method according to claim 1, characterized in that, The objective function is constructed based on the weights of each frequency band and the corresponding predicted RMS values.

6. The method according to claim 5, characterized in that, The objective function is: Where n represents the number of frequency bands, This represents the weight of the i-th frequency band.

7. The method according to claim 1, characterized in that, The step of iteratively generating new filter parameters in the search space using the Bayesian optimization algorithm includes: Within a preset number of search iterations, the surrogate model is updated based on the prediction results of the regression model, and the new filter parameters are generated.

8. The method according to claim 7, characterized in that, An asymmetric search space is used to design the search and generate new filter parameters.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 8.

Images