Microperforated panel sound absorption coefficient prediction method based on optimization algorithm deep learning

Abstract

The invention discloses a micro-perforated panel sound absorption coefficient prediction method based on optimization algorithm deep learning. The method comprises the following steps: generating structure parameters of a plurality of micro-perforated plates in a random generation and logic arrangement mode; constructing a theoretical sound absorption coefficient curve of the corresponding relation between the structure parameters and the theoretical sound absorption coefficients of the micro-perforated plates; establishing a sound absorption coefficient prediction model, and training the sound absorption coefficient prediction model through the theoretical sound absorption coefficient curve to obtain a trained sound absorption coefficient prediction model; and inputting the structure parameters of the to-be-predicted micro-perforated plate into the trained sound absorption coefficient prediction model, and outputting the sound absorption coefficient of the to-be-predicted micro-perforated plate based on the processed sound absorption coefficient so as to realize prediction. The method is suitable for different types of micro-perforated plates and changes thereof, does not need a complex numerical simulation process, is higher in calculation speed, is suitable for calculation in engineering practice, has higher generalization ability, and can be applied to other acoustic metamaterials.

Description

Technical Field

[0001] The invention relates to a method for predicting a sound absorption coefficient and relates to the field of artificial intelligence, and in particular to a method for predicting a sound absorption coefficient of a micro-perforated plate based on deep learning of an optimization algorithm. Background Art

[0002] Microperforated Panel (MPP) is an acoustic metamaterial with extraordinary acoustic properties. By optimizing the design of its structural parameters, it can meet the specific requirements of sound absorption performance in engineering applications.

[0003] Since the sound absorption performance of micro-perforated panels is affected by multiple factors such as pore size, perforation rate, panel thickness, back cavity depth, etc., the prediction of its sound absorption coefficient is crucial for acoustic design.

[0004] Traditional methods for calculating sound absorption coefficients are generally based on empirical formulas or numerical simulations. Although these methods can provide certain prediction results, their efficiency is relatively poor.

[0005] In recent years, deep learning has been widely used in multiple interdisciplinary fields, such as computer science, materials, chemistry, physics, etc., and deep learning is involved in various monitoring and prediction applications. However, there is no efficient and systematic prediction method in the field of micro-perforated plate prediction. Summary of the invention

[0006] In order to solve the problems existing in the background technology, the present invention provides a method for predicting the sound absorption coefficient of a micro-perforated plate based on deep learning of an optimization algorithm. The method of the present invention constructs a deep neural network DNN (Deep Neural Networks) model, which can input the physical parameters of the micro-perforated plate, such as pore size, perforation rate, plate thickness, back cavity depth and other structural parameters, learn the complex relationship between the structural parameters of the micro-perforated plate and its sound absorption performance, and quickly and accurately predict its sound absorption coefficient, thereby providing an efficient prediction method.

[0007] The technical solution adopted by the present invention is:

[0008] The method for predicting the sound absorption coefficient of a micro-perforated plate based on deep learning of an optimization algorithm of the present invention comprises:

[0009] In the first step, the structural parameters of several micro-perforated plates are generated by random generation and logical arrangement.

[0010] The second step is to construct a theoretical sound absorption coefficient curve of the corresponding relationship between the structural parameters of each micro-perforated plate and the theoretical sound absorption coefficient.

[0011] The third step is to establish a sound absorption coefficient prediction model, train the sound absorption coefficient prediction model through the theoretical sound absorption coefficient curve, and obtain a trained sound absorption coefficient prediction model.

[0012] The fourth step is to input the structural parameters of the micro-perforated plate to be predicted into the trained sound absorption coefficient prediction model, and output the sound absorption coefficient of the micro-perforated plate to be predicted based on the processed sound absorption coefficient to achieve prediction.

[0013] In the first step, the structural parameters of the micro-perforated plate include pore size, perforation rate, plate thickness and back cavity depth.

[0014] In the third step, the sound absorption coefficient prediction model specifically adopts a multi-layer fully connected neural network DNN, and the hidden layer uses the activation function ReLU.

[0015] In the third step, the sound absorption coefficient prediction model is trained by the root mean square optimization algorithm RMSProp (Root Mean Square Propagation) and the learning rate decay mechanism, uses the mean square error as the loss function, and is optimized by the back propagation algorithm.

[0016] In the third step, the sound absorption coefficient predicted by the sound absorption coefficient prediction model is compared with the theoretical sound absorption coefficient curve to verify the prediction effect of the model.

[0017] The beneficial effects of the present invention are:

[0018] The method of the present invention adopts a deep learning model of the RMSProp optimization algorithm to provide more efficient and accurate prediction results, especially in the case of complex structures and multi-factor influences; by introducing the RMSProp optimization algorithm, a more efficient and accurate method for predicting the sound absorption coefficient of micro-perforated plates is provided than the traditional method. The method can handle complex multi-parameter relationships, is applicable to different types of micro-perforated plates and their changes, and does not require complex numerical simulation processes, has a faster calculation speed, and is suitable for calculations in engineering practice. The method not only has high prediction efficiency, but also has strong generalization ability, and can be applied to other acoustic metamaterials. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 It is a schematic diagram of the process of the present invention;

[0020] Figure 2 This is a schematic diagram of the neural network architecture of the present invention;

[0021] Figure 3 This is a flow chart of model training of the present invention;

[0022] Figure 4 A graph showing changes in loss values ​​during the training process of the present invention;

[0023] Figure 5 Four sets of comparison diagrams of structural parameter model prediction results and theoretical values ​​are selected for the present invention, among which: Figure 5 (a) is a comparison chart of the prediction results of the first set of structural parameter models of the present invention and the theoretical values, Figure 5 (b) is a comparison chart of the predicted results of the second set of structural parameter models of the present invention and the theoretical values, Figure 5 (c) is a comparison chart of the prediction results of the third group of structural parameter models of the present invention and the theoretical values, Figure 5 (d) is a comparison chart of the predicted results of the fourth group of structural parameter models of the present invention and the theoretical values. DETAILED DESCRIPTION

[0024] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0025] like Figure 1 As shown, the method for predicting the sound absorption coefficient of a micro-perforated plate based on deep learning of an optimization algorithm of the present invention is as follows:

[0026] In the first step, several structural parameters of micro-perforated plates are generated by random generation and logical arrangement. The structural parameters of the micro-perforated plates include pore size, perforation rate, plate thickness and back cavity depth.

[0027] The second step is to construct a theoretical sound absorption coefficient curve of the corresponding relationship between the structural parameters of each micro-perforated plate and the theoretical sound absorption coefficient.

[0028] The third step is to establish a sound absorption coefficient prediction model, train the sound absorption coefficient prediction model through the theoretical sound absorption coefficient curve, and obtain the trained sound absorption coefficient prediction model. The sound absorption coefficient prediction model specifically adopts a multi-layer fully connected neural network DNN, and the hidden layer uses the activation function ReLU. The sound absorption coefficient prediction model has an input layer with 4 structural parameters, three hidden layers (the number of nodes is 64-128-512 respectively), and an output layer containing 256 sound absorption coefficient values ​​of different frequencies. The sound absorption coefficient prediction model is trained by the root mean square optimization algorithm RMSProp and the learning rate decay mechanism, using the mean square error as the loss function, and optimized by the back propagation algorithm. The sound absorption coefficient predicted by the sound absorption coefficient prediction model is compared with the theoretical sound absorption coefficient curve to verify the prediction effect of the model.

[0029] The fourth step is to input the structural parameters of the micro-perforated plate to be predicted into the trained sound absorption coefficient prediction model, and output the sound absorption coefficient of the micro-perforated plate to be predicted based on the processed sound absorption coefficient to achieve prediction.

[0030] The sample data set of the present invention outputs 500,000 data groups by randomly generating and logically arranging them within a specified range. Each data group extracts 256 characteristic frequencies in the frequency range of 0-3200 Hz to calculate the sound absorption coefficient value. The 256 sound absorption coefficient values ​​are output, and finally the sound absorption coefficient curve is output. Figure 2 As shown in the figure, the neural network includes an input layer with 4 structural parameters, three hidden layers (the number of nodes is 64-128-512 respectively) and an output layer containing 256 sound absorption coefficient values. In the data preprocessing stage, the normalization method is used to standardize the input data. The deep learning model uses a multi-layer fully connected neural network. The hidden layer uses the ReLU activation function. The RMSProp (Root Mean Square Propagation) algorithm is selected as the optimization algorithm. The loss function uses the mean square error MSE (Mean-Square Error). During the training process, the learning rate attenuation mechanism and cross-validation methods are used to accelerate convergence and prevent overfitting. Figure 3 and Figure 4 As shown in the figure, during the model training process, both the training loss and the validation loss showed a downward trend and stabilized after about 100 iterations. The mean square error (MSE) of the training set and the validation set was less than 0.001, and the model training was completed. In practical applications, users only need to provide the four structural parameters of the micro-perforated plate, and then the trained deep learning model can quickly predict the sound absorption coefficient of the micro-perforated plate.

[0031] The selection range of micro-perforated plate structural parameters is shown in Table 1 below:

[0032] Table 1

[0033]

[0034] The acoustic parameters of the micro-perforated plate include relative acoustic resistance r, relative acoustic mass m and sound absorption coefficient α. Four sets of structural parameters are now input to compare the model prediction values ​​with the theoretical calculation values. The four sets of structural parameters are shown in Table 2 below.

[0035] In the embodiment of the present invention, after the micro-perforated plate sound absorption coefficient prediction model is trained, the known structural parameters are input into the model, and the output structure is the sound absorption curve of the micro-perforated plate. After comparison with the theoretical sound absorption curve, it is proved that the model can more accurately predict the sound absorption coefficient of the micro-perforated plate. Figure 5 (a) Figure 5 (b) Figure 5 (c) and Figure 5 as shown in (d).

[0036] Table 2

[0037]

[0038] The above is only a preferred specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by any technician familiar with the technical field within the technical scope disclosed in the present invention should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims. The information disclosed in the background technology section of this article is only intended to deepen the understanding of the overall background technology of the present invention, and should not be regarded as an admission or in any form that the information constitutes prior art known to those skilled in the art.

Claims

1. A method for predicting the sound absorption coefficient of a micro-perforated plate based on deep learning of an optimization algorithm, characterized in that: include: In the first step, the structural parameters of several micro-perforated plates are generated by random generation and logical arrangement; The second step is to construct a theoretical sound absorption coefficient curve of the corresponding relationship between the structural parameters of each micro-perforated plate and the theoretical sound absorption coefficient; The third step is to establish a sound absorption coefficient prediction model, train the sound absorption coefficient prediction model through a theoretical sound absorption coefficient curve, and obtain a trained sound absorption coefficient prediction model; The fourth step is to input the structural parameters of the micro-perforated plate to be predicted into the trained sound absorption coefficient prediction model, and output the sound absorption coefficient of the micro-perforated plate to be predicted based on the processed sound absorption coefficient to achieve prediction.

2. The method for predicting the sound absorption coefficient of a micro-perforated plate based on deep learning of an optimization algorithm according to claim 1 is characterized in that: In the first step, the structural parameters of the micro-perforated plate include pore size, perforation rate, plate thickness and back cavity depth.

3. The method for predicting the sound absorption coefficient of a micro-perforated plate based on deep learning of an optimization algorithm according to claim 1 is characterized in that: In the third step, the sound absorption coefficient prediction model specifically adopts a multi-layer fully connected neural network DNN, and the hidden layer uses the activation function ReLU.

4. The method for predicting the sound absorption coefficient of a micro-perforated plate based on deep learning of an optimization algorithm according to claim 1, characterized in that: In the third step, the sound absorption coefficient prediction model is trained by the root mean square optimization algorithm RMSProp and the learning rate decay mechanism, the mean square error is used as the loss function, and it is optimized by the back propagation algorithm.

5. The method for predicting the sound absorption coefficient of a micro-perforated plate based on deep learning of an optimization algorithm according to claim 1 is characterized in that: In the third step, the sound absorption coefficient predicted by the sound absorption coefficient prediction model is compared with the theoretical sound absorption coefficient curve to verify the prediction effect of the model.

Images