Sound insulation wall high-frequency sound transmission loss prediction method and device based on Bayesian neural network, equipment and medium

Abstract

The invention discloses a Bayesian neural network-based sound insulation wall high-frequency sound transmission loss prediction method and device, equipment and a medium, and relates to the technical field of computers. Comprising the following steps: constructing an original Bayesian neural network, and training the original Bayesian neural network based on historical measurement data to obtain a target Bayesian neural network; obtaining structure attribute data and low-frequency sound reduction index data of the to-be-predicted sound insulation wall; the structure attribute data comprises parameter data of a constituent material of the to-be-predicted sound insulation wall; the parameter data comprises material thickness data and material surface density data; the low-frequency sound reduction index data is sound reduction index data of a preset target low-frequency band; and inputting the structure attribute data of the to-be-predicted sound insulation wall and the low-frequency sound insulation index data into the target Bayesian neural network so as to generate high-frequency sound insulation index data corresponding to the to-be-predicted sound insulation wall. Therefore, high-precision prediction of the high-frequency sound transmission loss of the sound insulation wall of the transformer substation can be realized under the condition that the sound insulation wall is not damaged.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a method, device, equipment and medium for predicting high-frequency sound transmission loss of a sound insulation wall based on a Bayesian neural network. Background Art

[0002] In modern urban environments, some open-air substations are built near residential areas. The broadband noise generated by substation operations can significantly impact the surrounding environment. In engineering projects, noise reduction is often achieved through the construction of soundproof walls. Sound transmission loss (STL) is a key technical indicator of the sound insulation performance of soundproof walls, and its full-band evaluation is crucial for the acceptance and retrofitting of substation noise reduction projects.

[0003] Currently, there are four main technologies for estimating and measuring the sound insulation performance of sound insulation walls: theoretical calculation, empirical formulas, laboratory measurements, and field measurements. However, these methods each have limitations in determining the full-band sound transmission loss of substation sound insulation walls, making efficient and accurate measurements difficult. Theoretical calculation methods predict sound insulation performance by building mathematical models (such as finite element analysis and boundary element methods). This eliminates the need for actual samples, saving cost and time, and allows for the simulation of various material combinations for design optimization. However, this method is highly dependent on model accuracy and cannot incorporate field measurement data for dynamic correction, resulting in significant deviations between predicted and actual performance. Empirical formulas, based on empirical data, provide simple formulas for rapid sound insulation prediction, making them easy to apply in engineering projects. However, existing formulas lack universality and are only applicable to specific materials and structures, making them unsuitable for the complex structures of substation sound insulation walls. Laboratory measurements, which calculate sound transmission loss using the sound pressure level difference between the source and receiving rooms, offer high accuracy. However, this method requires destructive sampling of the sound insulation wall, making it impractical for on-site assessment of existing substation sound insulation walls. Furthermore, the limited size of the specimens makes it difficult to characterize the acoustic properties of large-scale structures. On-site measurement methods, which do not require structural damage to the sound insulation wall, can be performed directly in engineering environments such as substations. However, this method is affected by diffraction and air absorption in the high-frequency band (>1000Hz), significantly reducing the accuracy of high-frequency sound transmission loss measurements and failing to meet full-band assessment requirements.

[0004] As can be seen from the above, how to achieve high-precision prediction of high-frequency sound transmission loss of substation sound insulation walls without destroying the sound insulation walls is an urgent problem to be solved. Summary of the Invention

[0005] In view of this, the present invention aims to provide a Bayesian neural network-based method, device, equipment, and medium for predicting high-frequency sound transmission loss in sound insulation walls. This method can achieve high-precision prediction of high-frequency sound transmission loss in substation sound insulation walls without damaging the sound insulation walls. The specific solution is as follows:

[0006] In a first aspect, the present application provides a method for predicting high-frequency sound transmission loss of a sound insulation wall based on a Bayesian neural network, comprising:

[0007] Constructing an original Bayesian neural network and training the original Bayesian neural network based on historical measurement data to obtain a target Bayesian neural network; the target Bayesian neural network is a fully connected neural network; the historical measurement data includes parameter data of the target material and corresponding full-band sound insulation data;

[0008] Acquiring structural property data and low-frequency sound insulation data of the sound insulation wall to be predicted; the structural property data includes parameter data of the constituent material of the sound insulation wall to be predicted; the parameter data includes material thickness data and material surface density data; the low-frequency sound insulation data is sound insulation data of a preset target low-frequency band;

[0009] The structural property data and low-frequency sound insulation data of the sound insulation wall to be predicted are input into the target Bayesian neural network to generate high-frequency sound insulation data corresponding to the sound insulation wall to be predicted; the high-frequency sound insulation data is the sound insulation data of a preset target high frequency band.

[0010] Optionally, the target Bayesian neural network includes an input layer, two hidden layers, and an output layer; the activation function of the target Bayesian neural network is a ReLU function; and the loss function of the target Bayesian neural network includes a data fitting term and a KL regularization term;

[0011] The input layer consists of a material thickness data node, a material surface density data node, and four low-frequency sound insulation data nodes; the output layer consists of two high-frequency sound insulation data nodes of different frequencies; and the hidden layer consists of several data nodes.

[0012] Optionally, before constructing the original Bayesian neural network and training the original Bayesian neural network based on historical measurement data to obtain the target Bayesian neural network, the method further includes:

[0013] The historical measurement data are preprocessed to obtain target historical measurement data, and a training data set and a validation data set are constructed based on the target historical measurement data according to a preset ratio; the historical measurement data include parameter data of target materials and corresponding full-band sound insulation data; the target materials include metal materials, building materials, and sound-absorbing materials.

[0014] Optionally, constructing an original Bayesian neural network and training the original Bayesian neural network based on historical measurement data to obtain a target Bayesian neural network includes:

[0015] Constructing an original Bayesian neural network, and training the original Bayesian neural network using the training data set to obtain a trained Bayesian neural network;

[0016] The trained Bayesian neural network is validated and optimized using the validation data set to obtain the target Bayesian neural network.

[0017] Optionally, the using the training data set to train the original Bayesian neural network to obtain a trained Bayesian neural network includes:

[0018] The original Bayesian neural network is trained using the training data set by means of variational inference to determine parameter data of the Bayesian neural network, and a trained Bayesian neural network is generated based on the parameter data.

[0019] Optionally, the using the validation data set to validate and optimize the trained Bayesian neural network to obtain the target Bayesian neural network includes:

[0020] Inputting the parameter data of the target material and the corresponding low-frequency sound insulation data in the verification data set into the trained Bayesian neural network to generate corresponding predicted high-frequency sound insulation data;

[0021] Based on the predicted high-frequency sound insulation data, a preset Monte Carlo sampling algorithm is used to calculate the mean and standard deviation of the high-frequency sound transmission loss, and the obtained data is compared and analyzed with the historical measurement data to obtain the target Bayesian neural network.

[0022] Optionally, obtaining the structural property data and low-frequency sound insulation data of the sound insulation wall to be predicted includes:

[0023] The structural property data and low-frequency sound insulation data of the sound insulation wall to be predicted are obtained based on the on-site measurement method.

[0024] In a second aspect, the present application provides a device for predicting high-frequency sound transmission loss of a sound insulation wall based on a Bayesian neural network, comprising:

[0025] A target Bayesian neural network construction module is used to construct an original Bayesian neural network and train the original Bayesian neural network based on historical measurement data to obtain a target Bayesian neural network; the target Bayesian neural network is a fully connected neural network; the historical measurement data includes parameter data of the target material and corresponding full-band sound insulation data;

[0026] An input data acquisition module is configured to acquire structural property data and low-frequency sound insulation data of the sound insulation wall to be predicted; the structural property data includes parameter data of the material constituting the sound insulation wall to be predicted; the parameter data includes material thickness data and material surface density data; and the low-frequency sound insulation data is sound volume data of a preset target low-frequency band;

[0027] A high-frequency sound insulation data output module is configured to input the structural attribute data of the sound insulation wall to be predicted and the low-frequency sound insulation data into the target Bayesian neural network to generate high-frequency sound insulation data corresponding to the sound insulation wall to be predicted; the high-frequency sound insulation data is sound insulation data in a preset target high-frequency band.

[0028] In a third aspect, the present application provides an electronic device, comprising:

[0029] Memory, used to store computer programs;

[0030] A processor is used to execute the computer program to implement the aforementioned method for predicting high-frequency sound transmission loss of sound insulation walls based on Bayesian neural network.

[0031] In a fourth aspect, the present application provides a computer-readable storage medium for storing a computer program, wherein when the computer program is executed by a processor, the aforementioned method for predicting high-frequency sound transmission loss of a sound insulation wall based on a Bayesian neural network is implemented.

[0032] The present application provides a method for predicting high-frequency sound transmission loss of a sound insulation wall based on a Bayesian neural network. The method first constructs an original Bayesian neural network and trains the original Bayesian neural network based on historical measurement data to obtain a target Bayesian neural network; the target Bayesian neural network is a fully connected neural network. Structural attribute data and low-frequency sound insulation data of the sound insulation wall to be predicted are then obtained; the structural attribute data include parameter data of the material constituting the sound insulation wall to be predicted; the parameter data include material thickness data and material surface density data. Finally, the structural attribute data and low-frequency sound insulation data of the sound insulation wall to be predicted are input into the target Bayesian neural network to generate high-frequency sound insulation data corresponding to the sound insulation wall to be predicted.

[0033] As can be seen above, this application uses a targeted Bayesian neural network to predict the high-frequency sound insulation data of the predicted sound insulation wall using the acquired structural property data and low-frequency sound insulation data. This method integrates engineering parameters such as surface density and thickness with measured low-frequency data, breaking through the linear assumption limitations of traditional empirical formulas and improving the prediction accuracy of complex sound insulation walls. By extrapolating high-frequency performance from low-frequency data and combining it with low-frequency data measured on-site, a complete assessment of the sound transmission loss of the sound insulation wall across the entire frequency range is achieved. This allows for highly accurate prediction of the high-frequency sound transmission loss of substation sound insulation walls without damaging the sound insulation wall. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are merely embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0035] Figure 1 This is a flow chart of a method for predicting high-frequency sound transmission loss of a sound insulation wall based on a Bayesian neural network disclosed in this application;

[0036] Figure 2 This is a schematic diagram of the high-frequency sound transmission loss prediction results disclosed in this application;

[0037] Figure 3 This is a schematic diagram of a device for predicting high-frequency sound transmission loss of a sound insulation wall based on a Bayesian neural network disclosed in this application;

[0038] Figure 4 This is a structural diagram of an electronic device disclosed in this application. DETAILED DESCRIPTION

[0039] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0040] Currently, there are four main technologies for estimating and measuring the sound insulation performance of sound insulation walls: theoretical calculation, empirical formulas, laboratory measurements, and field measurements. However, these methods each have limitations in determining the full-band sound transmission loss of substation sound insulation walls, making efficient and accurate measurements difficult. Theoretical calculation methods predict sound insulation performance by building mathematical models (such as finite element analysis and boundary element methods). This eliminates the need for actual samples, saving cost and time, and allows for the simulation of various material combinations for design optimization. However, this method is highly dependent on model accuracy and cannot incorporate field measurement data for dynamic correction, resulting in significant deviations between predicted and actual performance. Empirical formulas, based on empirical data, provide simple formulas for rapid sound insulation prediction, making them easy to apply in engineering projects. However, existing formulas lack universality and are only applicable to specific materials and structures, making them unsuitable for the complex structures of substation sound insulation walls. Laboratory measurements, which calculate sound transmission loss using the sound pressure level difference between the source and receiving rooms, offer high accuracy. However, this method requires destructive sampling of the sound insulation wall, making it impossible to conduct on-site evaluations of already constructed substation sound insulation walls. Furthermore, the limited size of the specimens makes it difficult to reflect the acoustic properties of large-scale structures. The on-site measurement method does not require destruction of the sound insulation wall structure and can be implemented directly in engineering environments such as substations. However, this method is affected by diffraction and air absorption effects in the high-frequency band (>1000Hz), resulting in a significant reduction in the measurement accuracy of high-frequency sound transmission loss, making it unable to meet the needs of full-band evaluation. Therefore, this application provides a Bayesian neural network-based prediction scheme for high-frequency sound transmission loss of sound insulation walls, which can achieve high-precision prediction of high-frequency sound transmission loss of substation sound insulation walls without destroying the sound insulation walls.

[0041] See also Figure 1 As shown, the embodiment of the present application discloses a method for predicting high-frequency sound transmission loss of a sound insulation wall based on a Bayesian neural network, comprising:

[0042] Step S11: construct an original Bayesian neural network, and train the original Bayesian neural network based on historical measurement data to obtain a target Bayesian neural network.

[0043] In this embodiment, the historical measurement data includes parameter data for the target material and the corresponding full-band sound insulation data. This historical measurement data covers the surface density, thickness, and corresponding full-band sound insulation of various materials. The target materials include, but are not limited to, common metal plates (such as steel and aluminum), gypsum boards, brick walls commonly used in construction, and various types of sound-absorbing materials such as ultrafine wool, glass wool, and mineral wool.

[0044] Furthermore, the collected data is strictly screened and processed to remove abnormal data that does not meet the standards to ensure the reliability and consistency of the data. To ensure the effectiveness and accuracy of the network, the data set is divided into a training set and a validation set in a certain ratio, and it is ensured that the data types of the two remain uniform in distribution. Specifically, for example, in a specific embodiment, the data set is divided into a training set and a validation set in a ratio of 70% and 30%. Specifically, before constructing the original Bayesian neural network and training the original Bayesian neural network based on historical measurement data to obtain a target Bayesian neural network, it may also include: preprocessing the historical measurement data to obtain target historical measurement data, and constructing a training data set and a validation data set based on the target historical measurement data in a ratio of 70% and 30%.

[0045] In this embodiment, the training set is used to train the network, while the validation set is used to evaluate the accuracy of the network's predictions. Specifically, constructing an original Bayesian neural network and training the original Bayesian neural network based on historical measurement data to obtain a target Bayesian neural network may include: constructing an original Bayesian neural network and training the original Bayesian neural network using the training data set to obtain a trained Bayesian neural network; and validating and optimizing the trained Bayesian neural network using the validation data set to obtain the target Bayesian neural network.

[0046] In this embodiment, the target Bayesian neural network is a fully connected neural network. The input layer of this fully connected neural network has six nodes, representing the thickness and surface density of the sound barrier, as well as four low-frequency sound transmission losses (125-1000 Hz, octave). The output layer has two nodes, representing the sound barrier's sound transmission losses at 2 kHz and 4 kHz. There are two hidden layers, each with 50 nodes, and the activation function is the Reluctant Unified Unit (ReLU). The loss function consists of two parts: a data fitting term and a KL regularization term. The data fitting term is consistent with that of a general fully connected network, while the KL regularization term is calculated by calculating the KL divergence between the variational posterior and the prior distribution. Specifically, the Adam optimizer was used during training, with an initial learning rate of 0.005, a training cycle of 10,000 epochs, and no early stopping strategy. The original Bayesian neural network is trained using the training dataset by variational inference. To support gradient backpropagation, the weights and biases are sampled using a reparameterization technique in each forward propagation. For a Gaussian distribution, the KL divergence of each weight and bias is as follows:

[0047] ;

[0048] in, and are the weights and biases of the prior distribution; and are the weights and biases of the posterior distribution.

[0049] Furthermore, after the Bayesian neural network training is completed, its prediction effect needs to be verified to evaluate the network's prediction accuracy and reliability. Specifically, the use of the validation dataset to verify and optimize the trained Bayesian neural network to obtain the target Bayesian neural network may include: inputting the parameter data of the target material and the corresponding low-frequency sound insulation data in the validation dataset into the trained Bayesian neural network to generate corresponding predicted high-frequency sound insulation data; calculating the mean and standard deviation of the high-frequency sound transmission loss based on the predicted high-frequency sound insulation data using a preset Monte Carlo sampling algorithm, and comparing and analyzing the obtained data with the historical measurement data to obtain the target Bayesian neural network. That is, using the validation dataset, the structural properties (areal density, thickness) and low-frequency sound insulation of the material are input into the network to calculate its high-frequency sound insulation. In order to evaluate the network error, the mean and standard deviation of the high-frequency sound transmission loss are calculated through Monte Carlo sampling (repeated forward propagation 5000 times). Then, the predicted mean is compared and analyzed with the actual measurement data to evaluate the network's prediction error and applicability. See Figure 2 As shown in the figure, the error between the model prediction value and the actual sound insulation value is within an acceptable range, indicating that the model has good prediction performance.

[0050] Step S12: Acquire structural property data and low-frequency sound insulation data of the sound insulation wall to be predicted.

[0051] In this embodiment, field measurement is used to obtain the structural property data and low-frequency sound insulation data of the sound insulation wall to be predicted. The parameter data includes material thickness data and material surface density data. The low-frequency sound insulation data is the sound insulation data for a preset target low-frequency band. Field measurement is a technical method for directly evaluating the sound transmission loss of a sound insulation wall in an actual engineering environment. Unlike laboratory measurement, field measurement does not require damage to the sound insulation wall structure and can be performed directly at construction sites such as substations. It is suitable for the acceptance, renovation, and long-term monitoring of existing sound insulation facilities. Field measurement uses high-precision microphones and sound level meters to synchronously collect sound pressure signals from the source and receiving sides. The received signals are processed and the sound pressure level difference between the two sides and the sound transmission loss of the sound insulation wall are calculated.

[0052] Step S13: inputting the structural attribute data and low-frequency sound insulation data of the sound insulation wall to be predicted into the target Bayesian neural network to generate high-frequency sound insulation data corresponding to the sound insulation wall to be predicted.

[0053] In this embodiment, high-frequency sound waves are significantly affected by diffraction and air absorption, resulting in low accuracy in high-frequency sound transmission loss measurements using on-site measurement methods. Therefore, after obtaining high-precision low-frequency sound insulation data through on-site measurement, this low-frequency sound insulation data is input into the target Bayesian neural network to generate high-frequency sound insulation data corresponding to the sound insulation wall to be predicted; this high-frequency sound insulation data is the sound insulation data for the preset target high-frequency range. In other words, the target Bayesian neural network predicts the high-frequency sound insulation data for the sound insulation wall to be predicted based on the structural property data and low-frequency sound insulation data of the sound insulation wall to be predicted. This compensates for the significant reduction in accuracy of high-frequency sound transmission loss measurements in the high-frequency range (>1000Hz) of the on-site measurement method due to diffraction and air absorption.

[0054] As can be seen from the above, the embodiments of the present application address the limitations of traditional methods in obtaining the full-band sound transmission loss of substation sound insulation walls. Based on field measurement methods, a Bayesian neural network-based method for predicting the high-frequency sound transmission loss of substation sound insulation walls is proposed, thereby obtaining the full-band sound transmission loss of the sound insulation wall. This application integrates engineering parameters such as surface density and thickness with low-frequency measured data, breaking through the linear assumption limitations of traditional empirical formulas and improving the prediction accuracy of complex sound insulation walls. Through the probabilistic output of the Bayesian neural network, confidence intervals for the prediction results are provided, measurement noise and model errors are quantified, and the reliability of engineering decisions is improved. By extrapolating high-frequency performance from low-frequency data and combining it with low-frequency data measured in-situ, a complete assessment of the full-band sound transmission loss of the sound insulation wall is achieved.

[0055] See also Figure 3 As shown, the embodiment of the present application discloses a device for predicting high-frequency sound transmission loss of a sound insulation wall based on a Bayesian neural network, comprising:

[0056] A target Bayesian neural network construction module 11 is used to construct an original Bayesian neural network and train the original Bayesian neural network based on historical measurement data to obtain a target Bayesian neural network; the target Bayesian neural network is a fully connected neural network; the historical measurement data includes parameter data of the target material and corresponding full-band sound insulation data;

[0057] The input data acquisition module 12 is used to acquire structural attribute data and low-frequency sound insulation data of the sound insulation wall to be predicted; the structural attribute data includes parameter data of the constituent material of the sound insulation wall to be predicted; the parameter data includes material thickness data and material surface density data; the low-frequency sound insulation data is the sound volume data of the preset target low-frequency band;

[0058] The high-frequency sound insulation data output module 13 is configured to input the structural attribute data of the sound insulation wall to be predicted and the low-frequency sound insulation data into the target Bayesian neural network to generate high-frequency sound insulation data corresponding to the sound insulation wall to be predicted; the high-frequency sound insulation data is sound insulation data in a preset target high-frequency band.

[0059] The target Bayesian neural network includes an input layer, two hidden layers, and an output layer; the activation function of the target Bayesian neural network is a ReLU function; the loss function of the target Bayesian neural network includes a data fitting term and a KL regularization term; the input layer is composed of a material thickness data node, a material surface density data node, and four low-frequency sound insulation data nodes; the output layer is composed of two high-frequency sound insulation data nodes of different frequencies; and the hidden layer is composed of a plurality of data nodes.

[0060] As can be seen from the above, the present embodiment uses a target Bayesian neural network to predict the high-frequency sound insulation data of the predicted sound insulation wall using the acquired structural property data and low-frequency sound insulation data. This method integrates engineering parameters such as surface density and thickness with measured low-frequency data, breaking through the linear assumption limitations of traditional empirical formulas and improving the prediction accuracy of complex sound insulation walls. By extrapolating high-frequency performance from low-frequency data and combining it with low-frequency data measured on-site, a complete assessment of the sound transmission loss of the sound insulation wall across all frequency bands is achieved. This allows for highly accurate prediction of the high-frequency sound transmission loss of substation sound insulation walls without damaging the sound insulation wall.

[0061] In some specific implementations, the input data acquisition module 12 may specifically include:

[0062] The input data acquisition unit is used to acquire the structural property data and low-frequency sound insulation data of the sound insulation wall to be predicted based on the on-site measurement method.

[0063] In some specific embodiments, the device for predicting high-frequency sound transmission loss of a sound insulation wall based on a Bayesian neural network may further include:

[0064] a historical measurement data processing unit, configured to pre-process the historical measurement data to obtain target historical measurement data, and construct a training data set and a validation data set based on the target historical measurement data according to a preset ratio; the historical measurement data includes parameter data of target materials and corresponding full-band sound insulation data; the target materials include metal materials, building materials, and sound-absorbing materials;

[0065] Accordingly, the target Bayesian neural network construction module 11 may specifically include:

[0066] An original Bayesian neural network construction submodule is used to construct an original Bayesian neural network and train the original Bayesian neural network using the training data set to obtain a trained Bayesian neural network;

[0067] The original Bayesian neural network training submodule is used to verify and optimize the trained Bayesian neural network using the verification data set to obtain the target Bayesian neural network.

[0068] Furthermore, in some specific implementations, the original Bayesian neural network construction submodule may specifically include:

[0069] a Bayesian neural network parameter determination unit, configured to train the original Bayesian neural network using the training data set by means of variational inference to determine parameter data of the Bayesian neural network, and generate a trained Bayesian neural network based on the parameter data;

[0070] The original Bayesian neural network training submodule may specifically include:

[0071] a predicted high-frequency sound insulation data generating unit, configured to input the parameter data of the target material and the corresponding low-frequency sound insulation data in the verification data set into the trained Bayesian neural network to generate corresponding predicted high-frequency sound insulation data;

[0072] A target Bayesian neural network generation unit is used to calculate the mean and standard deviation of high-frequency sound transmission loss based on the predicted high-frequency sound insulation data using a preset Monte Carlo sampling algorithm, and compare and analyze the obtained data with the historical measurement data to obtain the target Bayesian neural network.

[0073] Furthermore, the embodiment of the present application also discloses an electronic device, Figure 4 This is a structural diagram of an electronic device 20 according to an exemplary embodiment. The content in this diagram should not be construed as limiting the scope of application of this application. The electronic device 20 may specifically include: at least one processor 21, at least one memory 22, a power supply 23, a communication interface 24, an input / output interface 25, and a communication bus 26. The memory 22 is used to store a computer program, which is loaded and executed by the processor 21 to implement the relevant steps of the Bayesian neural network-based method for predicting high-frequency sound transmission loss in a sound insulation wall as disclosed in any of the aforementioned embodiments. Furthermore, the electronic device 20 in this embodiment may specifically be an electronic computer.

[0074] In this embodiment, the power supply 23 is used to provide operating voltage for each hardware device on the electronic device 20; the communication interface 24 can create a data transmission channel between the electronic device 20 and the external device. The communication protocol it follows is any communication protocol that can be applied to the technical solution of this application and is not specifically limited here; the input and output interface 25 is used to obtain external input data or output data to the outside world. Its specific interface type can be selected according to specific application needs and is not specifically limited here.

[0075] In addition, the memory 22, as a carrier for resource storage, can be a read-only memory, random access memory, disk or CD, etc. The resources stored thereon can include an operating system 221, a computer program 222, etc., and the storage method can be temporary storage or permanent storage.

[0076] The operating system 221 is used to manage and control the hardware devices and computer program 222 on the electronic device 20. It can be Windows Server, NetWare, Unix, Linux, etc. In addition to including a computer program capable of implementing the Bayesian neural network-based method for predicting high-frequency sound transmission loss of a sound insulation wall, as disclosed in any of the aforementioned embodiments and executed by the electronic device 20, the computer program 222 may further include computer programs capable of performing other specific tasks.

[0077] Furthermore, this application discloses a computer-readable storage medium for storing a computer program. When executed by a processor, the computer program implements the aforementioned Bayesian neural network-based method for predicting high-frequency sound transmission loss in sound insulation walls. The specific steps of this method can be found in the corresponding sections disclosed in the aforementioned embodiments and will not be further elaborated here.

[0078] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from the other embodiments. Reference can be made to the descriptions of the identical or similar parts between the various embodiments. For the devices disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple, and the relevant parts can be referred to the descriptions of the methods.

[0079] Professionals may further appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the above description has generally described the components and steps of each example according to their functions. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians may use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0080] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein may be implemented directly using hardware, a software module executed by a processor, or a combination of the two. The software module may be placed in random access memory (RAM), internal memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

[0081] Finally, it should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or device. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of additional identical elements in the process, method, article, or device comprising the element.

[0082] The above is a detailed introduction to the technical solution provided by the present application. Specific examples are used herein to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method of the present application and its core idea. At the same time, for those skilled in the art, according to the ideas of the present application, there may be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.

Claims

1. A method for predicting high-frequency sound transmission loss of sound insulation walls based on Bayesian neural networks, characterized in that: include: Constructing an original Bayesian neural network, and training the original Bayesian neural network based on historical measurement data to obtain a target Bayesian neural network; The target Bayesian neural network is a fully connected neural network; the historical measurement data includes parameter data of the target material and corresponding full-band sound insulation data; Acquiring structural property data and low-frequency sound insulation data of the sound insulation wall to be predicted; the structural property data includes parameter data of the constituent material of the sound insulation wall to be predicted; the parameter data includes material thickness data and material surface density data; the low-frequency sound insulation data is sound insulation data of a preset target low-frequency band; The structural property data and low-frequency sound insulation data of the sound insulation wall to be predicted are input into the target Bayesian neural network to generate high-frequency sound insulation data corresponding to the sound insulation wall to be predicted; the high-frequency sound insulation data is the sound insulation data of a preset target high frequency band.

2. The method for predicting high-frequency sound transmission loss of a sound insulation wall based on a Bayesian neural network according to claim 1, characterized in that: The target Bayesian neural network includes an input layer, two hidden layers and an output layer; the activation function of the target Bayesian neural network is a ReLU function; the loss function of the target Bayesian neural network includes a data fitting term and a KL regularization term; The input layer consists of a material thickness data node, a material surface density data node, and four low-frequency sound insulation data nodes; the output layer consists of two high-frequency sound insulation data nodes of different frequencies; and the hidden layer consists of several data nodes.

3. The method for predicting high-frequency sound transmission loss of a sound insulation wall based on a Bayesian neural network according to claim 1, characterized in that: Before constructing the original Bayesian neural network and training the original Bayesian neural network based on historical measurement data to obtain the target Bayesian neural network, the method further includes: The historical measurement data are preprocessed to obtain target historical measurement data, and a training data set and a validation data set are constructed based on the target historical measurement data according to a preset ratio; the historical measurement data include parameter data of target materials and corresponding full-band sound insulation data; the target materials include metal materials, building materials, and sound-absorbing materials.

4. The method for predicting high-frequency sound transmission loss of a sound insulation wall based on a Bayesian neural network according to claim 3, characterized in that: The constructing of the original Bayesian neural network and training the original Bayesian neural network based on the historical measurement data to obtain the target Bayesian neural network includes: Constructing an original Bayesian neural network, and training the original Bayesian neural network using the training data set to obtain a trained Bayesian neural network; The trained Bayesian neural network is validated and optimized using the validation data set to obtain the target Bayesian neural network.

5. The method for predicting high-frequency sound transmission loss of a sound insulation wall based on a Bayesian neural network according to claim 4, characterized in that: The method of training the original Bayesian neural network using the training data set to obtain a trained Bayesian neural network includes: The original Bayesian neural network is trained using the training data set by means of variational inference to determine parameter data of the Bayesian neural network, and a trained Bayesian neural network is generated based on the parameter data.

6. The method for predicting high-frequency sound transmission loss of a sound insulation wall based on a Bayesian neural network according to claim 4, characterized in that: The method of verifying and optimizing the trained Bayesian neural network using the verification data set to obtain the target Bayesian neural network includes: Inputting the parameter data of the target material and the corresponding low-frequency sound insulation data in the verification data set into the trained Bayesian neural network to generate corresponding predicted high-frequency sound insulation data; Based on the predicted high-frequency sound insulation data, a preset Monte Carlo sampling algorithm is used to calculate the mean and standard deviation of the high-frequency sound transmission loss, and the obtained data is compared and analyzed with the historical measurement data to obtain the target Bayesian neural network.

7. The method for predicting high-frequency sound transmission loss of a sound insulation wall based on a Bayesian neural network according to any one of claims 1 to 6, characterized in that: The step of obtaining the structural property data and low-frequency sound insulation data of the sound insulation wall to be predicted includes: The structural property data and low-frequency sound insulation data of the sound insulation wall to be predicted are obtained based on the on-site measurement method.

8. A device for predicting high-frequency sound transmission loss of sound insulation walls based on Bayesian neural networks, characterized in that: include: A target Bayesian neural network construction module is used to construct an original Bayesian neural network and train the original Bayesian neural network based on historical measurement data to obtain a target Bayesian neural network; The target Bayesian neural network is a fully connected neural network; the historical measurement data includes parameter data of the target material and corresponding full-band sound insulation data; An input data acquisition module is configured to acquire structural property data and low-frequency sound insulation data of the sound insulation wall to be predicted; the structural property data includes parameter data of the material constituting the sound insulation wall to be predicted; the parameter data includes material thickness data and material surface density data; and the low-frequency sound insulation data is sound volume data of a preset target low-frequency band; A high-frequency sound insulation data output module is configured to input the structural attribute data of the sound insulation wall to be predicted and the low-frequency sound insulation data into the target Bayesian neural network to generate high-frequency sound insulation data corresponding to the sound insulation wall to be predicted; the high-frequency sound insulation data is sound insulation data in a preset target high-frequency band.

9. An electronic device, characterized in that: include: Memory, used to store computer programs; A processor, configured to execute the computer program to implement the method for predicting high-frequency sound transmission loss of a sound insulation wall based on a Bayesian neural network according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that Used to store a computer program, wherein when the computer program is executed by a processor, it implements the method for predicting high-frequency sound transmission loss of a sound insulation wall based on a Bayesian neural network according to any one of claims 1 to 7.

Images