A method, apparatus, device, and medium for sound field prediction based on RBF neural network

Abstract

This invention relates to a sound field prediction method, apparatus, device, and medium based on an RBF neural network, belonging to the field of vibration and noise control technology. The sound field prediction method based on the RBF neural network includes establishing a sound field structure model, simulating sound field vibration on the sound field structure model to obtain sound field vibration data under different working conditions; training the constructed RBF neural network based on the sound field vibration data and adjusting the parameters of the RBF neural network, wherein the RBF neural network includes an input layer, a hidden layer, and an output layer; and predicting the sound field vibration data to be predicted based on the fully trained RBF neural network to obtain the predicted total sound pressure level, thereby improving the accuracy and efficiency of sound field prediction.

Description

Technical Field

[0001] This invention relates to the field of vibration and noise control technology, and in particular to a sound field prediction method, device, equipment and medium based on RBF neural network. Background Technology

[0002] Sound field prediction is an important area of ​​research in vibration and noise control. It involves the accurate prediction of sound propagation and distribution characteristics. However, the complexity of the sound field often exceeds the scope of traditional prediction methods, especially when dealing with the correlation between near-field and far-field sound fields. These differences and dynamic changes make it extremely challenging to build accurate prediction models.

[0003] Near-field sound fields typically contain rich details of sound waves, such as interference, diffraction, and reflection, which make the distribution of sound waves extremely complex. Far-field sound fields, on the other hand, are relatively simple, with sound waves mainly propagating in the form of spherical waves. However, the correlation between the two is crucial, because even small changes in the near-field sound field can have a significant impact on the far-field sound field.

[0004] Traditional sound field prediction methods, such as physical models and numerical calculations, can simulate the propagation characteristics of sound fields to some extent. However, these methods are not only computationally intensive and time-consuming, but their prediction accuracy is also limited when dealing with complex sound field environments. Summary of the Invention

[0005] In view of this, it is necessary to provide a sound field prediction method, device, equipment and medium based on RBF neural network to solve the technical problems of low sound field prediction accuracy and low efficiency.

[0006] To address the above problems, this invention provides a sound field prediction method based on an RBF neural network, comprising:

[0007] A sound field structure model is established, and sound field vibration simulation is performed on the sound field structure model to obtain sound field vibration data under different working conditions.

[0008] The constructed RBF neural network is trained based on the sound field vibration data, and the parameters of the RBF neural network are adjusted. The RBF neural network includes an input layer, a hidden layer, and an output layer.

[0009] The RBF neural network, after being fully trained, is used to predict the sound field vibration data to be predicted, and the predicted total sound pressure level is obtained.

[0010] In one possible implementation, establishing the sound field structure model includes:

[0011] A sound field structure model was established using aluminum sheet as the sound field structure and air as the sound field domain.

[0012] In one possible implementation, the step of performing sound field vibration simulation on the sound field structure model to obtain sound field vibration data under different working conditions includes:

[0013] A frequency range and frequency step size are determined, and a point load is applied to the aluminum sheet based on the frequency range and frequency step size to obtain sound field vibration data under different working conditions. The sound field vibration data includes vibration data and noise data.

[0014] In one possible implementation, the frequency range is 5Hz to 8000Hz, and the frequency step size is 5Hz.

[0015] In one possible implementation, training the constructed RBF neural network based on the acoustic field vibration data and adjusting the parameters of the RBF neural network includes:

[0016] Construct an RBF neural network and determine the parameters of the RBF neural network;

[0017] The sound field vibration data is divided to obtain training set data and test set data;

[0018] The training set data is input into the input layer, and the input layer transforms the training set data to obtain a numerical vector.

[0019] The numerical vector is input into the hidden layer, and the hidden layer performs nonlinear transformation and feature extraction on the numerical vector to obtain a feature vector, wherein the neurons use the radial basis function as the activation function of the hidden layer;

[0020] The feature vector is input into the output layer, and the feature vector is linearly combined through the output layer to obtain the prediction result;

[0021] The prediction results are denormalized to obtain prediction data. Error is calculated on the prediction data to obtain the error. The parameters of the RBF neural network are adjusted based on the error to obtain a fully trained RBF neural network.

[0022] The test set data is input into the fully trained RBF neural network to obtain the predicted data of the test set, and the fully trained RBF neural network is evaluated based on the predicted data of the test set.

[0023] In one possible implementation, the parameters of the RBF neural network include the beta value, the center point, and the dimension of the output layer.

[0024] In one possible implementation, the prediction of the sound field vibration data to be predicted based on the fully trained RBF neural network to obtain the predicted total sound pressure level includes:

[0025] The vibration data in the sound field vibration data to be predicted is input into the fully trained RBF neural network. Based on the fully trained RBF neural network, the vibration data is predicted to obtain noise data at the same frequency as the vibration data.

[0026] The predicted total sound pressure level is obtained based on the noise data.

[0027] On the other hand, the present invention also provides a sound field prediction device based on an RBF neural network, comprising:

[0028] The data acquisition module is used to establish a sound field structure model, perform sound field vibration simulation on the sound field structure model, and obtain sound field vibration data under different working conditions.

[0029] The training module is used to train the constructed RBF neural network based on the sound field vibration data and adjust the parameters of the RBF neural network, wherein the RBF neural network includes an input layer, a hidden layer, and an output layer.

[0030] The prediction module is used to predict the sound field vibration data to be predicted based on the fully trained RBF neural network, and obtain the predicted value of the total sound pressure level.

[0031] On the other hand, the present invention also provides an electronic device, including: a processor and a memory;

[0032] The memory stores a computer-readable program that can be executed by the processor;

[0033] When the processor executes the computer-readable program, it implements the steps in the sound field prediction method based on the RBF neural network as described above.

[0034] On the other hand, the present invention also provides a computer-readable storage medium storing one or more programs that can be executed by one or more processors to implement the steps in the sound field prediction method based on the RBF neural network as described above.

[0035] The beneficial effects of this invention are: training the constructed RBF neural network based on sound field vibration data and adjusting the parameters of the RBF neural network; predicting the sound field vibration data to be predicted based on the fully trained RBF neural network and obtaining the total sound pressure level prediction value, thereby improving the sound field prediction accuracy and prediction efficiency. Attached Figure Description

[0036] Figure 1 A flowchart illustrating an embodiment of the sound field prediction method based on RBF neural network provided by the present invention;

[0037] Figure 2 A schematic diagram of the RBF neural network model for the sound field prediction method based on the RBF neural network provided by the present invention;

[0038] Figure 3 A schematic diagram showing the comparison between a predicted value and an actual value for the sound field prediction method based on RBF neural network provided by this invention;

[0039] Figure 4 A schematic diagram showing the comparison between another predicted value and the actual value of the sound field prediction method based on RBF neural network provided by the present invention;

[0040] Figure 5 A schematic diagram of an embodiment of the sound field prediction device based on RBF neural network provided by the present invention;

[0041] Figure 6 A schematic diagram of an embodiment of the electronic device provided by the present invention. Detailed Implementation

[0042] Preferred embodiments of the present invention will now be described in detail with reference to the accompanying drawings, which form part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, but are not intended to limit the scope of the present invention.

[0043] This invention discloses a sound field prediction method, apparatus, device, and medium based on an RBF neural network, which can be used in a computer. The method, apparatus, or computer-readable storage medium involved in this invention can be integrated with the aforementioned apparatus or be relatively independent.

[0044] One specific embodiment of the present invention discloses a sound field prediction method based on an RBF neural network, which can be executed by a computer, specifically by one or more processors of the computer. For example... Figure 1 As shown, the sound field prediction method based on RBF neural network includes:

[0045] S101. Establish a sound field structure model, perform sound field vibration simulation on the sound field structure model, and obtain sound field vibration data under different working conditions.

[0046] S102. The constructed RBF neural network is trained based on the sound field vibration data, and the parameters of the RBF neural network are adjusted. The RBF neural network includes an input layer, a hidden layer, and an output layer.

[0047] S103. Based on the fully trained RBF neural network, predict the sound field vibration data to be predicted and obtain the predicted value of the total sound pressure level.

[0048] Among them, the mapping relationship between vibration data and noise data is established by using RBF (radial basis) neural network; the propagation and distribution characteristics of sound waves in various complex sound field environments can be predicted quickly and accurately by using RBF neural network, without the need for tedious physical models and numerical calculations, saving a lot of manpower costs and reducing the possibility of errors caused by repetitive work.

[0049] Compared with existing technologies, the sound field prediction method based on RBF neural network provided in this embodiment establishes a sound field structure model, simulates sound field vibration on the sound field structure model, and obtains sound field vibration data under different working conditions; trains the constructed RBF neural network based on the sound field vibration data, and adjusts the parameters of the RBF neural network, wherein the RBF neural network includes an input layer, a hidden layer, and an output layer; and predicts the sound field vibration data to be predicted based on the fully trained RBF neural network to obtain the predicted total sound pressure level, thereby improving the accuracy and efficiency of sound field prediction.

[0050] In some embodiments, in step S101, a sound field structure model is established, using an aluminum sheet as the sound field structure and the air domain as the sound field domain. This model is then implemented in COMSOL. In Multiphysics finite element software, an aluminum sheet was selected as the acoustic field structure, and the air domain was used as the acoustic field domain to establish an acoustic field structure model. Acoustic vibration simulation was performed on the acoustic field structure model to obtain acoustic vibration data under different working conditions. The frequency range and frequency step size were determined: 5Hz~8000Hz, with a frequency step size of 5Hz. Based on the frequency range and frequency step size, a point load was applied to the aluminum sheet to obtain acoustic vibration data under different working conditions. This acoustic vibration data included vibration data and noise data. A point load was applied to the aluminum sheet, with the study frequency being (5, 5, 8000), i.e., a frequency range of 5Hz~8000Hz, and a frequency step size of 5Hz. Acoustic vibration simulations were performed on the structure under different working conditions, obtaining multiple sets of acoustic vibration data under different working conditions. For each working condition, one-third of the octave band data within the same frequency band was selected. A large portion of the working condition data was used as the training set, and the remaining small portion was used as the test set.

[0051] In some embodiments, in step S102, the constructed RBF neural network is trained based on acoustic field vibration data, and the parameters of the RBF neural network are adjusted. The RBF neural network includes an input layer, a hidden layer, and an output layer. First, the RBF neural network is constructed; for a schematic diagram of the RBF neural network model, please refer to [link to schematic diagram]. Figure 2 ,like Figure 2 As shown, This serves as the input to the neural network, corresponding to the sound field vibration data. The output of the neural network, An activation function maps the inputs of neurons in a neural network to their outputs. The weight matrix represents the influencing factors of the neural network. The parameters of the RBF neural network are determined, including the beta value (self.beta), the center points (num_cebters), and the output layer dimension (out_dim). The sound field vibration data is divided into training and testing sets. The training set data is then input into the input layer, which transforms the data to obtain numerical vectors. These numerical vectors are then input into the hidden layer, where nonlinear transformations and feature extraction are performed to obtain feature vectors, where the radial basis function (RBF) is used. The function is the activation function of the hidden layer neuron; the feature vector is input into the output layer, and the feature vector is linearly combined through the output layer to obtain the prediction result; finally, the prediction result is denormalized to obtain the prediction data, the error of the prediction data is calculated, and the parameters of the RBF neural network are adjusted based on the error to obtain a fully trained RBF neural network; the test set data is input into the fully trained RBF neural network to obtain the prediction data of the test set, and the fully trained RBF neural network is evaluated based on the prediction data of the test set to evaluate the generalization ability and performance in practical applications of the RBF neural network.

[0052] The RBF neural network has only three layers: the first is the input layer, the second is the hidden layer, and the third is the output layer. All weights between neurons in the input and hidden layers are equal to 1. The hidden layers use radial basis functions (RBFs) as activation functions. The connections between the hidden and output layers are the same as in a regular neural network. Their weights can be changed through training. The beta value, center points, and expansion range of the RBF neural network are defined. Different beta values ​​will produce different effects; a smaller beta value will result in a smoother function that covers a wider input space. The pre-prepared training data is imported and standardized to ensure the neural network can process it. The predicted results are then denormalized to obtain the predicted data, and error calculations are performed. Finally, a curve comparing the predicted and actual values ​​is plotted, and the trained neurophysical model `rbf_model.joblib` is exported. By comparing the curves, the neural network model parameters are continuously adjusted. Finally, a better-fitting model is obtained when `self.beta=0.35`, `num_centers=85`, and `out_dim=1`, indicating a fully trained RBF neural network.

[0053] In some embodiments, in step S103, the sound field vibration data to be predicted is predicted based on the fully trained RBF neural network to obtain the total sound pressure level prediction value. The vibration data in the sound field vibration data to be predicted is input into the fully trained RBF neural network. The vibration data is predicted based on the fully trained RBF neural network to obtain noise data at the same frequency as the vibration data. The total sound pressure level prediction value is obtained based on the noise data.

[0054] Comparing the predicted results with the actual values, the predicted total sound pressure level for test set 1 was 144.226 dB, while the actual total sound pressure level was 145.167 dB, a difference of 0.941 dB. For a comparison diagram, please refer to [link to diagram]. Figure 3 The predicted total sound pressure level (SPL) for test set 2 was 135.047 dB, while the actual SPL was 136.229 dB, a difference of 1.182 dB. Please refer to the comparison diagram. Figure 4 The comparison shows that it can predict the sound field more efficiently and ensure a certain level of prediction accuracy in various complex and variable sound field environments, such as different working conditions, saving a lot of time, manpower and resources.

[0055] To better implement the sound field prediction method based on RBF neural network in the embodiments of the present invention, based on the sound field prediction method based on RBF neural network, correspondingly, as follows: Figure 5 As shown, this embodiment of the invention also provides a sound field prediction device based on an RBF neural network. The sound field prediction device 500 based on an RBF neural network includes:

[0056] The data acquisition module 501 is used to establish a sound field structure model, perform sound field vibration simulation on the sound field structure model, and obtain sound field vibration data under different working conditions.

[0057] Training module 502 is used to train the constructed RBF neural network based on sound field vibration data and adjust the parameters of the RBF neural network. The RBF neural network includes an input layer, a hidden layer, and an output layer.

[0058] The prediction module 503 is used to predict the sound field vibration data to be predicted based on the fully trained RBF neural network, and obtain the predicted value of the total sound pressure level.

[0059] like Figure 6 As shown, the present invention also provides an electronic device 600, which can be a mobile terminal, desktop computer, laptop, handheld computer, server, or other computing device. The electronic device 600 includes a processor 601, a memory 602, and a display 603. Figure 6Only some components of the electronic device 600 are shown, but it should be understood that it is not required to implement all the components shown, and more or fewer components may be implemented instead.

[0060] In some embodiments, memory 602 may be an internal storage unit of the electronic device 600, such as a hard disk or memory of the electronic device 600. In other embodiments, memory 602 may be an external storage device of the electronic device 600, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc., equipped on the electronic device 600. Furthermore, memory 602 may include both internal and external storage units of the electronic device 600. Memory 602 is used to store application software and various types of data installed on the electronic device 600, such as program code installed on the electronic device 600. Memory 602 may also be used to temporarily store data that has been output or will be output. In one embodiment, memory 602 stores a sound field prediction program based on an RBF neural network, which can be executed by processor 601 to implement the sound field prediction method based on an RBF neural network according to various embodiments of the present invention.

[0061] In some embodiments, processor 601 may be a central processing unit (CPU), microprocessor, or other data processing chip, used to run program code stored in memory 602 or process data, such as a sound field prediction method based on RBF neural network.

[0062] In some embodiments, display 603 may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, or an OLED (Organic Light-Emitting Diode) touchscreen. Display 603 is used to display recognition information from an RBF neural network-based sound field prediction program and to display a visual user interface. Components 601-603 of electronic device 600 communicate with each other via a system bus.

[0063] In some embodiments, when the processor 601 executes the sound field prediction program based on the RBF neural network in the memory 602, it implements each step of the sound field prediction method based on the RBF neural network as described in the above embodiments. Since the sound field prediction method based on the RBF neural network has been described in detail above, it will not be repeated here.

[0064] Accordingly, the present invention also provides a computer-readable storage medium for storing a computer-readable program or instruction, which, when executed by a processor, can implement the steps or functions of the sound field prediction method based on the RBF neural network provided in the above-described method embodiments.

[0065] In summary, the sound field prediction method, apparatus, device, and medium based on RBF neural network provided by this invention establish a sound field structure model, simulate sound field vibration on the sound field structure model, and obtain sound field vibration data under different working conditions; train the constructed RBF neural network based on the sound field vibration data, and adjust the parameters of the RBF neural network, wherein the RBF neural network includes an input layer, a hidden layer, and an output layer; predict the sound field vibration data to be predicted based on the fully trained RBF neural network, and obtain the predicted value of the total sound pressure level, thereby improving the accuracy and efficiency of sound field prediction.

[0066] Those skilled in the art will understand that all or part of the processes of the methods described in the above embodiments can be implemented by a computer program instructing related hardware, and the program can be stored in a computer-readable storage medium. The computer-readable storage medium may be a disk, optical disk, read-only memory, or random access memory, etc.

[0067] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A sound field prediction method based on RBF neural network, characterized in that, include: A sound field structure model is established, and sound field vibration simulation is performed on the sound field structure model to obtain sound field vibration data under different working conditions. The sound field vibration data includes vibration data and noise data. The constructed RBF neural network is trained based on the sound field vibration data, and the parameters of the RBF neural network are adjusted. The RBF neural network includes an input layer, a hidden layer, and an output layer. The training of the constructed RBF neural network based on the sound field vibration data and the adjustment of the parameters of the RBF neural network include: Construct an RBF neural network and determine the parameters of the RBF neural network; The sound field vibration data is divided to obtain training set data and test set data; The training set data is input into the input layer, and the input layer transforms the training set data to obtain a numerical vector. The numerical vector is input into the hidden layer, and the hidden layer performs nonlinear transformation and feature extraction on the numerical vector to obtain a feature vector, wherein the neurons use the radial basis function as the activation function of the hidden layer; The feature vector is input into the output layer, and the feature vector is linearly combined through the output layer to obtain the prediction result; The prediction results are denormalized to obtain prediction data. Error is calculated on the prediction data to obtain the error. The parameters of the RBF neural network are adjusted based on the error to obtain a fully trained RBF neural network. The test set data is input into the fully trained RBF neural network to obtain the predicted data of the test set, and the fully trained RBF neural network is evaluated based on the predicted data of the test set. Based on the fully trained RBF neural network, a prediction of the sound field vibration data to be predicted is performed to obtain a predicted total sound pressure level. This prediction of the sound field vibration data to be predicted using the fully trained RBF neural network includes: The vibration data in the sound field vibration data to be predicted is input into the fully trained RBF neural network. Based on the fully trained RBF neural network, the vibration data is predicted to obtain noise data at the same frequency as the vibration data. The predicted total sound pressure level is obtained based on the noise data.

2. The sound field prediction method based on RBF neural network according to claim 1, characterized in that, The establishment of the sound field structure model includes: A sound field structure model was established using aluminum sheet as the sound field structure and air as the sound field domain.

3. The sound field prediction method based on RBF neural network according to claim 2, characterized in that, The process of simulating the acoustic field vibration of the acoustic field structure model to obtain acoustic field vibration data under different working conditions includes: The frequency range and frequency step size are determined, and a point load is applied to the aluminum sheet based on the frequency range and frequency step size to obtain sound field vibration data under different working conditions.

4. The sound field prediction method based on RBF neural network according to claim 3, characterized in that, The frequency range is 5Hz to 8000Hz, and the frequency step size is 5Hz.

5. The sound field prediction method based on RBF neural network according to claim 1, characterized in that, The parameters of the RBF neural network include the beta value, the center point, and the dimension of the output layer.

6. A sound field prediction device based on an RBF neural network, characterized in that, include: The data acquisition module is used to establish a sound field structure model, perform sound field vibration simulation on the sound field structure model, and obtain sound field vibration data under different working conditions, wherein the sound field vibration data includes vibration data and noise data. A training module is used to train a constructed RBF neural network based on the sound field vibration data and to adjust the parameters of the RBF neural network. The RBF neural network includes an input layer, a hidden layer, and an output layer. The training of the constructed RBF neural network based on the sound field vibration data and the adjustment of the parameters of the RBF neural network include: Construct an RBF neural network and determine the parameters of the RBF neural network; The sound field vibration data is divided to obtain training set data and test set data; The training set data is input into the input layer, and the input layer transforms the training set data to obtain a numerical vector. The numerical vector is input into the hidden layer, and the hidden layer performs nonlinear transformation and feature extraction on the numerical vector to obtain a feature vector, wherein the neurons use the radial basis function as the activation function of the hidden layer; The feature vector is input into the output layer, and the feature vector is linearly combined through the output layer to obtain the prediction result; The prediction results are denormalized to obtain prediction data. Error is calculated on the prediction data to obtain the error. The parameters of the RBF neural network are adjusted based on the error to obtain a fully trained RBF neural network. The test set data is input into the fully trained RBF neural network to obtain the predicted data of the test set, and the fully trained RBF neural network is evaluated based on the predicted data of the test set. The prediction module is used to predict the sound field vibration data to be predicted based on the fully trained RBF neural network, and obtain a predicted total sound pressure level value. The process of predicting the sound field vibration data to be predicted based on the fully trained RBF neural network and obtaining the predicted total sound pressure level value includes: The vibration data in the sound field vibration data to be predicted is input into the fully trained RBF neural network. Based on the fully trained RBF neural network, the vibration data is predicted to obtain noise data at the same frequency as the vibration data. The predicted total sound pressure level is obtained based on the noise data.

7. An electronic device, characterized in that, Including memory and processor; The memory stores a computer-readable program that can be executed by the processor; When the processor executes the computer-readable program, it implements the steps in the sound field prediction method based on the RBF neural network as described in any one of claims 1-5.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores one or more programs, which can be executed by one or more processors to implement the steps in the sound field prediction method based on the RBF neural network as described in any one of claims 1-5.

Images