Production method of composite sound absorption suspended ceiling suitable for terminal building

Abstract

The invention provides a production method of a composite sound absorption suspended ceiling suitable for an airport terminal, and belongs to the technical field of composite sound absorption suspended ceilings. The production method comprises the steps that firstly, the propagation mechanism of sound waves in a composite material is analyzed, the material parameter range required by the optimal sound absorption effect is calculated, the parameter space is divided into six areas, and clustering center points are extracted to make an experiment sample plate; through a sound absorption performance test in an airport terminal environment, multi-band sound wave reflectivity, a micro-vibration spectrum and a sound energy attenuation curve are measured and arranged into an eight-dimensional sound absorption vector matrix. And training the sound absorption parameter effect model. The model is used for carrying out increasing and expanding sampling on material parameters, a large number of parameter combinations and prediction performance are generated, the parameter combinations with the optimal multi-band comprehensive sound absorption effect are screened out, the parameter combinations comprise the micropore aluminum plate with the hole diameter being 0.5 mm, the hole pitch being 15.6 mm and the punching rate being 8%, and finally the micropore aluminum plate and through hole ceramic composite sound absorption suspended ceiling production aiming at the terminal noise characteristics is achieved.

Description

Technical Field

[0001] The present invention belongs to the technical field of composite sound-absorbing ceilings, and more specifically, relates to a production method of a composite sound-absorbing ceiling applicable to airport terminals. Background Art

[0002] As a large public transportation hub, the acoustic environment in an airport terminal is complex and the noise spectrum is wide. Traditional sound-absorbing ceilings in airport terminals mainly adopt a single material structure, such as perforated gypsum boards, mineral wool boards or glass fiber, etc., in combination with simple backing materials, and achieve sound wave absorption through the perforation rate and thickness designed by experience. Such sound-absorbing systems are widely used in engineering practice, are easy to install, and have a certain sound-absorbing effect in specific frequency bands.

[0003] However, there are various noise sources in the airport terminal environment, including crowd noise, luggage cart noise, public address system noise, air conditioning equipment noise, and aircraft noise, etc. These noises have significant differences in frequency characteristics and cover a wide spectrum range from low to high frequencies. Due to the single structure of traditional sound-absorbing ceilings, their sound-absorbing mechanisms are often optimized for specific frequency bands and it is difficult to meet the sound-absorbing requirements of multi-band noises simultaneously. For example, thick porous materials have good sound-absorbing effects on low frequencies but poor performance on high frequencies, while thin perforated plates have advantages in high frequencies but insufficient sound absorption on low frequencies.

[0004] Currently, the design of sound-absorbing ceilings in airport terminals mainly relies on simplified acoustic models and designers' experience, lacking in-depth research on the sound-absorbing mechanism of composite material structures in multi-band noise environments and systematic optimization methods. Especially when it comes to composite structures such as through-hole ceramics and micro-hole aluminum plates, there is a complex non-linear relationship between material parameters and sound-absorbing performance in different frequency bands. Traditional design methods are difficult to accurately predict and optimize this relationship, resulting in the inability to achieve the optimal sound-absorbing effect in the full frequency band when facing the complex noise environment in the airport terminal. That is to say, there is a technical problem in the prior art that it is difficult for composite sound-absorbing ceilings in airport terminals to achieve the optimal sound-absorbing effect in multi-band noise environments. Summary of the Invention

[0005] In view of this, the present invention provides a production method of a composite sound-absorbing ceiling applicable to airport terminals, which can solve the technical problem in the prior art that it is difficult for composite sound-absorbing ceilings in airport terminals to achieve the optimal sound-absorbing effect in multi-band noise environments.

[0006] The present invention is implemented as follows: The present invention provides a production method for a composite sound-absorbing ceiling applicable to a terminal building, including: establishing an acoustic mathematical model based on the acoustic environment characteristics of the terminal building and calculating the range of material parameters required for the optimal sound-absorbing effect; performing a clustering analysis on the range of material parameters, using the K-means clustering algorithm to divide the range of material parameters into multiple material parameter regions, and extracting the clustering center point parameters of each material parameter region as the basis for making experimental samples; manufacturing experimental composite sound-absorbing ceiling samples according to the clustering center point parameters; establishing a sound-absorbing performance test experiment for the terminal building ceiling, and measuring the sound wave reflectivity, micro-vibration spectrum, and sound energy attenuation curve of each composite sound-absorbing ceiling sample; organizing the experimentally measured data into a sound-absorbing vector matrix to construct a neural network training set; designing a five-layer feedforward neural network architecture and training to obtain a sound-absorbing parameter effect model; using the sound-absorbing parameter effect model to perform an augmented sampling on the range of material parameters, generating multiple groups of material parameter combinations and their predicted sound-absorbing performances, and screening out the optimal material parameter combination with a high comprehensive sound-absorbing coefficient and a low manufacturing cost; and producing a composite sound-absorbing ceiling according to the selected optimal material parameter combination.

[0007] Among them, establishing an acoustic mathematical model based on the acoustic environment characteristics of the terminal building is to analyze the propagation mechanism of sound waves of each frequency in the composite material by using the wave equation and the acoustic impedance theory, and construct a calculation formula for the sound-absorbing parameters of the combined structure of through-hole ceramics and micro-perforated aluminum plates.

[0008] Among them, the wave equation and the acoustic impedance theory refer to the mathematical equations followed by sound waves when propagating in porous materials, including the Helmholtz equation and the characteristic impedance calculation formula, which are used to describe the propagation characteristics and boundary conditions of sound waves in different media.

[0009] Among them, the range of material parameters includes a through-hole ceramic thickness of 5 to 15 mm, a micro-perforated aluminum plate thickness of 1 to 3 mm, a hole diameter of the micro-perforated aluminum plate of 0.2 to 2 mm, a hole pitch of the micro-perforated aluminum plate of 3 to 20 mm, a perforation rate of the micro-perforated aluminum plate of 5% to 15%, and a density of the backing material of 30 to 80 kg per cubic meter.

[0010] Among them, the clustering center point parameters include the through-hole ceramic thickness, the micro-perforated aluminum plate thickness, the hole diameter of the micro-perforated aluminum plate, the hole pitch of the micro-perforated aluminum plate, the perforation rate of the micro-perforated aluminum plate, and the density of the backing material.

[0011] Among them, the sound-absorbing performance test experiment for the terminal building ceiling collects 20 typical noise frequencies in the terminal building as the test sound source.

[0012] Among them, the micro-vibration spectrum refers to the curve of the frequency distribution of the minute vibrations generated by the sound-absorbing material when impacted by sound waves, which is measured by an acceleration sensor and obtained through Fourier transform, and is used to evaluate the absorption and resonance characteristics of the material for sound waves of different frequencies.

[0013] Among them, the sound energy attenuation curve refers to the attenuation law of sound wave energy with time or distance after passing through the sound-absorbing material, which is measured and calculated by placing sound intensity meters before and after the material.

[0014] Among them, each clustering center point parameter corresponds to a group of eight-dimensional sound-absorbing vectors, and a neural network training set is constructed. The input is the clustering center point parameter, and the output is the corresponding sound-absorbing vector matrix; the sound-absorbing vectors are used to comprehensively describe the sound-absorbing performance characteristics of the material in different aspects, and each sound-absorbing vector contains data of eight dimensions; the sound-absorbing vector matrix is a two-dimensional data structure composed of multiple sound-absorbing vectors. Each row represents the sound-absorbing vector under a group of material parameter combinations, and each column represents the performance of different material combinations on the same sound-absorbing performance index.

[0015] Among them, the five-layer feedforward neural network architecture is a fully connected structure with an attention mechanism, including an input layer, three hidden layers, and an output layer, and an outlier amplification function added between the second hidden layer and the third hidden layer. The outlier amplification function is used to enhance the sensitivity recognition of extreme sound-absorbing performance parameter combinations.

[0016] Compared with the prior art, a production method of a composite sound-absorbing ceiling applicable to a terminal building provided by the present invention proposes a production method of a composite sound-absorbing ceiling based on the wave equation, acoustic impedance theory, and machine learning. Through systematic modeling and data-driven optimization methods, the accurate optimization of the sound-absorbing effect in the multi-frequency noise environment of the terminal building is realized.

[0017] Compared with the traditional technology, the present invention first establishes a mathematical model based on the physical propagation mechanism of sound waves, systematically describes the propagation law of sound waves in the combined structure of through-hole ceramics and micro-hole aluminum plates, and overcomes the limitations of traditional empirical design methods. In particular, a five-layer feedforward neural network architecture is introduced, combined with an attention mechanism and an outlier amplification function, so that the model can capture the non-linear relationship between material parameters and sound-absorbing performance, especially enhancing the automatic weight allocation ability for sound-absorbing characteristics in different frequency bands and the prediction accuracy of sound-absorbing performance in the critical parameter region.

[0018] Through the method of the present invention, the systematic optimization of the full-frequency sound-absorbing effect in the complex acoustic environment of the terminal building is realized, so that the produced composite sound-absorbing ceiling can provide excellent sound-absorbing performance for low, medium, and high-frequency noises in the terminal building at the same time. This method breaks through the bottleneck of traditional design methods in multi-frequency sound-absorbing optimization, solves the technical problem that it is difficult for the composite sound-absorbing ceiling of the terminal building to achieve the optimal sound-absorbing effect in the multi-frequency noise environment, and significantly improves the acoustic environment quality of the terminal building. Description of the Drawings

[0019] Figure 1 It is a flowchart of the method of the present invention. Detailed Embodiments

[0020] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0021] As Figure 1 shown, it is a flowchart of a production method of a composite sound-absorbing ceiling applicable to a terminal building. The method includes the following steps:

[0022] S01. Establish an acoustic mathematical model based on the acoustic environment characteristics of the terminal building, analyze the propagation mechanism of sound waves of each frequency in the composite material by using the wave equation and the acoustic impedance theory, and construct a calculation formula for the sound-absorbing parameters of the combined structure of through-hole ceramics and micro-hole aluminum plates;

[0023] S02. According to the established acoustic mathematical model, calculate the range of material parameters required for the optimal sound-absorbing effect, including the thickness of the through-hole ceramics being 5 to 15 mm, the thickness of the micro-hole aluminum plates being 1 to 3 mm, the hole diameter of the micro-hole aluminum plates being 0.2 to 2 mm, the hole spacing of the micro-hole aluminum plates being 3 to 20 mm, the perforation rate of the micro-hole aluminum plates being 5% to 15%, and the density of the backing material being 30 to 80 kg / m³;

[0024] S03. Conduct a clustering analysis on the range of the material parameters, use the K-means clustering algorithm to divide the range of the material parameters into 6 material parameter regions, and extract the clustering center point parameters of each material parameter region as the basis for making experimental samples;

[0025] S04. Make experimental composite sound-absorbing ceiling samples according to the clustering center point parameters. Each group of the composite sound-absorbing ceiling samples includes a through-hole ceramic layer, a micro-hole aluminum plate layer, and a sound-absorbing backing material, and the production size is 600 mm × 600 mm;

[0026] S05. Establish an experimental test on the sound-absorbing performance of the terminal building ceiling, collect 20 typical noise frequencies in the terminal building as test sound sources, and measure the sound wave reflectivity, micro-vibration spectrum, and sound energy attenuation curve of each composite sound-absorbing ceiling sample;

[0027] S06. Organize the data measured in the experiment into a sound-absorbing vector matrix. Each clustering center point parameter corresponds to a group of eight-dimensional sound-absorbing vectors, construct a neural network training set, with the input being the clustering center point parameters and the output being the corresponding sound-absorbing vector matrix;

[0028] S07. Design a five-layer feedforward neural network architecture, use the neural network training set and adopt the backpropagation algorithm to train the sound-absorbing parameter effect model, and stop training when the error of the validation set is lower than 0.01 or the number of iterations reaches 5000 times, and obtain the sound-absorbing parameter effect model;

[0029] S08. Using the acoustic absorption parameter effect model, perform augmented sampling on the material parameter range to generate 10,000 groups of material parameter combinations and their predicted acoustic absorption performances, and screen out the optimal material parameter combination with a high comprehensive acoustic absorption coefficient and low manufacturing cost.

[0030] S09. According to the obtained optimal material parameter combination, carry out the production of the composite acoustic ceiling.

[0031] Among them, this composite ceiling includes a micro-perforated aluminum plate layer (located at the bottom, facing the indoor space), a through-hole ceramic layer (located in the middle), and an acoustic absorption backing material (located at the top, close to the building structure). This structural design enables sound waves to first pass through the micro-perforated aluminum plate, then pass through the through-hole ceramic layer, and finally be absorbed by the acoustic absorption backing material. This multi-layer composite structure can achieve a more effective acoustic absorption effect for noises of different frequencies.

[0032] Among them, the wave equation and acoustic impedance theory refer to the mathematical equations followed by sound waves when propagating in porous materials, including the Helmholtz equation and the characteristic impedance calculation formula, which are used to describe the propagation characteristics and boundary conditions of sound waves in different media.

[0033] Among them, the micro-vibration frequency spectrum refers to the curve of the frequency distribution of the tiny vibrations generated by the acoustic absorption material when impacted by sound waves, which is measured by an acceleration sensor and obtained through Fourier transform, and is used to evaluate the absorption and resonance characteristics of the material for sound waves of different frequencies.

[0034] Among them, the sound energy attenuation curve refers to the attenuation law of sound wave energy over time or distance after passing through the acoustic absorption material, which is measured and calculated by placing sound intensity meters before and after the material. The faster the attenuation, the better the acoustic absorption effect.

[0035] Among them, the material parameter range includes the thickness of the through-hole ceramic, the thickness of the micro-perforated aluminum plate, the hole diameter of the micro-perforated aluminum plate, the hole spacing of the micro-perforated aluminum plate, the perforation rate of the micro-perforated aluminum plate, and the density of the backing material.

[0036] Among them, the clustering center point parameters include the thickness of the through-hole ceramic, the thickness of the micro-perforated aluminum plate, the hole diameter of the micro-perforated aluminum plate, the hole spacing of the micro-perforated aluminum plate, the perforation rate of the micro-perforated aluminum plate, and the density of the backing material.

[0037] Among them, the five-layer feedforward neural network architecture is a fully connected structure with an attention mechanism, including an input layer, three hidden layers, and an output layer, as well as an outlier amplification function added between the second hidden layer and the third hidden layer. The outlier amplification function is used to enhance the sensitivity recognition of extreme sound absorption performance parameter combinations, achieving the effect of improving the model prediction accuracy in the non-linear sound absorption characteristic interval. The input of the outlier amplification function includes the thickness of the through-hole ceramic, the pore diameter of the micro-perforated aluminum plate, the pore spacing of the micro-perforated aluminum plate, the perforation rate of the micro-perforated aluminum plate, and the density of the backing material, and the output is an eight-dimensional sound absorption performance vector; the attention mechanism considers the mutual correlation between different frequencies and is used to automatically adjust the weight distribution of the network for the sound absorption characteristics of each frequency band.

[0038] Among them, in the research of composite sound-absorbing ceilings, outliers refer to data points that exhibit abnormal characteristics in the relationship between material parameters and sound absorption performance. These points usually indicate that under certain parameter combinations, the sound absorption performance shows non-linear jumps or mutation characteristics, and may exhibit excellent or extremely poor sound absorption effects at certain frequencies. Outliers are of great value for discovering the critical values of material parameters and optimizing designs, because they often indicate resonance or interference phenomena of the material structure under certain conditions.

[0039] Among them, the sound absorption vector is a multi-dimensional numerical vector used to comprehensively describe the sound absorption performance characteristics of materials in different aspects. In this production method, each sound absorption vector contains data in eight dimensions, which may include key parameters such as the sound absorption coefficient in different frequency bands, the sound wave reflectivity, the micro-vibration characteristics of the material, and the sound energy attenuation rate. Through this vectorized representation, the sound absorption performance of materials can be quantitatively analyzed and compared from multiple angles.

[0040] Among them, the sound absorption vector matrix is a two-dimensional data structure composed of multiple sound absorption vectors. Each row represents the sound absorption vector under a set of material parameter combinations, and each column represents the performance of different material combinations on the same sound absorption performance index. This matrix form facilitates the neural network to learn the mapping relationship between material parameters and sound absorption performance, and batch data processing and model training can be efficiently carried out through matrix operations. In this production method, the sound absorption vector matrix is used as the training output target of the neural network and contains multi-dimensional sound absorption performance data such as sound wave reflectivity, micro-vibration spectrum, and sound energy attenuation curve.

[0041] Among them, the comprehensive sound absorption coefficient refers to the weighted average of the sound absorption coefficients measured at multiple frequency points, and the weights are determined according to the noise spectrum characteristics of the terminal building.

[0042] The following describes the specific implementation manners of the above steps in detail.

[0043] The specific implementation of step S01 is to establish an acoustic absorption mathematical model based on the acoustic environment characteristics of the terminal building. First, collect the acoustic environment data inside the terminal building, including key parameters such as the noise distribution characteristics in each frequency band, reverberation time, and sound pressure level distribution. Then, apply the Helmholtz equation to describe the propagation behavior of sound waves in various materials, which can accurately characterize the reflection, transmission, and absorption mechanisms of sound waves in different media. Next, establish a theoretical model of acoustic impedance and calculate the equivalent acoustic impedance of the combined structure of through-hole ceramics and micro-perforated aluminum plates, considering the influence of parameters such as material thickness, pore diameter, and perforation rate on sound wave absorption. Finally, by combining the Delany-Bazley model with the micro-perforated plate theory, construct a calculation formula for the acoustic absorption parameters applicable to the composite structure. The purpose of this step is to establish a theoretical basis to provide a mathematical basis for subsequent material parameter optimization, so that the design process is based on scientific theory rather than experience.

[0044] The specific implementation of step S02 is to calculate the range of material parameters required for the optimal acoustic absorption effect according to the established acoustic absorption mathematical model. First, program the acoustic absorption mathematical model in step S01 to construct a calculation program. Then, conduct a parameter sensitivity analysis of each material parameter to determine the influence weight of each parameter on the acoustic absorption performance. Next, use the Monte Carlo simulation method to randomly generate 10,000 groups of parameter combinations within the initial large range and calculate the theoretical acoustic absorption performance of each combination. Screen the calculation results and retain the parameter combinations with a theoretical acoustic absorption coefficient greater than 0.7 in the medium and high frequency bands (500 - 4000 Hz). Finally, determine the effective range of each parameter through statistical analysis, and obtain the material parameter range of the through-hole ceramic thickness of 5 to 15 mm, the micro-perforated aluminum plate thickness of 1 to 3 mm, the pore diameter of the micro-perforated aluminum plate of 0.2 to 2 mm, the pore spacing of the micro-perforated aluminum plate of 3 to 20 mm, the perforation rate of the micro-perforated aluminum plate of 5% to 15%, and the density of the backing material of 30 to 80 kg / m³. The purpose of this step is to determine the theoretically feasible range of material parameters to provide guidance for subsequent experimental design and production.

[0045] The specific implementation of step S03 is to conduct a cluster analysis on the range of material parameters. First, perform data standardization processing on the parameter combinations within the range of material parameters determined in step S02 to eliminate the influence of different parameter dimensions on the clustering results. Then, use the K-means clustering algorithm to divide the standardized parameter space into 6 regions. The selection of the K value is based on the silhouette coefficient evaluation. When K = 6, the silhouette coefficient reaches the optimal value of 0.68, indicating good clustering results. Next, calculate the center point coordinates of each clustering region as the representative parameter combination. Finally, reverse-standardize the clustering center point parameters to actual physical quantities to obtain 6 groups of clustering center point parameters as the basis for making experimental samples. The purpose of this step is to scientifically extract a small number of representative parameter combinations from a large number of theoretically feasible parameter combinations for experimental verification, which not only ensures the coverage of the parameter space but also reduces the experimental cost.

[0046] The specific implementation of step S04 is to fabricate a composite sound-absorbing ceiling sample for experiments according to the clustering center point parameters. First, in accordance with the 6 groups of clustering center point parameters, purchase the qualified through-hole ceramic plates, micro-perforated aluminum plates and sound-absorbing backing materials. The through-hole ceramics are prepared from alumina raw materials through a high-temperature sintering process, the micro-perforated aluminum plates are processed from aluminum alloy plates through a precision laser drilling process, and the backing material is selected as polyester fiber sound-absorbing cotton. Then, cut the micro-perforated aluminum plates and through-hole ceramics into the size of 600 mm × 600 mm according to the design dimensions. Next, use an environmentally friendly polyurethane adhesive to bond the three layers of materials together in the order of "through-hole ceramic layer - micro-perforated aluminum plate layer - sound-absorbing backing material". Finally, install an aluminum alloy frame on the edge of the sample to form a composite sound-absorbing ceiling sample with standard specifications. The purpose of this step is to transform the theoretical design into an actual sample and provide an experimental object for subsequent performance testing.

[0047] The specific implementation of step S05 is to establish an experimental test on the sound-absorbing performance of the terminal ceiling. First, select 20 typical locations such as the waiting hall, check-in area, and security check area in the terminal, collect the ambient noise spectra of each location, and extract 20 band characteristic noises in the range of 100 Hz to 8000 Hz as the test sound sources. Then, build an acoustic performance test platform, including a reverberation chamber and an impedance tube test system. The volume of the reverberation chamber is 200 cubic meters, which meets the ISO354 standard. Next, install the 6 groups of fabricated composite sound-absorbing ceiling samples on the test platform respectively, and measure the sound wave reflectivity (using the standing wave ratio method), micro-vibration spectrum (using the laser Doppler vibration measurement method) and sound energy attenuation curve (using the impulse response method) of the samples at each frequency band through devices such as sound pressure sensors and acceleration sensors. Finally, statistically process the measured data, eliminate the outliers, and ensure the accuracy and reliability of the data. The purpose of this step is to obtain the performance data of each composite sound-absorbing ceiling sample in the actual acoustic environment and provide a real basis for model training.

[0048] The specific implementation of step S06 is to organize the experimentally measured data into an acoustic absorption vector matrix. First, feature extraction is performed on the original data such as the acoustic wave reflectivity, micro-vibration spectrum, and acoustic energy attenuation curve obtained in step S05, including the average acoustic absorption coefficients in the low-frequency band (100 - 500 Hz), medium-frequency band (500 - 2000 Hz), and high-frequency band (2000 - 8000 Hz), as well as features such as the peak acoustic absorption frequency point and acoustic absorption bandwidth. Then, the extracted features are combined into an eight-dimensional acoustic absorption vector, including the average acoustic absorption coefficients of the three frequency bands, the acoustic wave reflectivity index, the micro-vibration frequency response coefficient, the acoustic energy attenuation rate, and two comprehensive performance indicators. Next, the 6 sets of clustering center point parameters and the corresponding eight-dimensional acoustic absorption vectors are combined into a training data set to form a mapping relationship with material parameters as the input and acoustic absorption performance as the output. Finally, the data is normalized so that the data of each dimension is scaled to the range of 0 - 1, which is convenient for neural network training. The purpose of this step is to convert the experimental data into a structured data set suitable for machine learning and prepare for subsequent model training.

[0049] The specific implementation of step S07 is to design a five-layer feedforward neural network architecture and train an acoustic absorption parameter effect model. First, construct the neural network architecture, including an input layer with 6 nodes (corresponding to 6 material parameters), a first hidden layer with 32 nodes, a second hidden layer with 64 nodes, a third hidden layer with 32 nodes, and an output layer with 8 nodes (corresponding to the eight-dimensional acoustic absorption vector). The ReLU activation function is used in the first hidden layer and the third hidden layer, the tanh activation function is used in the second hidden layer, and an outlier enlargement function is added between the second hidden layer and the third hidden layer. Then, implement the attention mechanism to automatically adjust the network's attention to the acoustic absorption characteristics of different frequency bands through a self-learning weight matrix. Next, use the Adam optimizer and the mean square error loss function, and use the data set constructed in step S06 for model training. The training adopts a 5-fold cross-validation method, the initial learning rate is set to 0.001, and a learning rate decay strategy is used. Finally, stop training when the validation set error is lower than 0.01 or the number of iterations reaches 5000 times, and save the model parameters. The purpose of this step is to establish a non-linear mapping model between material parameters and acoustic absorption performance and realize the prediction of acoustic absorption performance for any parameter combination.

[0050] The specific implementation of step S08 is to use the acoustic absorption parameter effect model to screen the optimal material parameter combination. First, within the range of material parameters determined in step S02, 10,000 groups of uniformly distributed material parameter combinations are generated using the Latin hypercube sampling method. Then, these parameter combinations are input into the neural network model trained in step S07 to predict their acoustic absorption performance, obtaining 10,000 groups of material parameters and the corresponding eight-dimensional acoustic absorption vectors. Next, according to the noise spectrum characteristics of the terminal building, the weight coefficients of the acoustic absorption performance in each frequency band are determined, and the comprehensive acoustic absorption coefficient is calculated. At the same time, a material cost evaluation model is established, and the production cost of each group of parameters is calculated according to the material consumption corresponding to each parameter. Finally, the Pareto optimization method is used to find a balance between the two objectives of the comprehensive acoustic absorption coefficient and the production cost, and the optimal material parameter combination with the highest comprehensive acoustic absorption coefficient and the lowest manufacturing cost is screened out. The purpose of this step is to find the optimal solution for performance and cost among a large number of possible parameter combinations.

[0051] The specific implementation of step S09 is to produce the composite acoustic ceiling according to the screened optimal material parameter combination. First, according to the optimal material parameter combination, a detailed production process flow and quality control standards are formulated. Then, raw materials that meet the parameter requirements are purchased, including through-hole ceramics, micro-perforated aluminum plates, and acoustic absorption backing materials. Next, the micro-perforated aluminum plates are precisely punched using numerical control equipment to ensure that the hole diameter, hole pitch, and perforation rate meet the design requirements. At the same time, the through-hole ceramics and the backing materials are cut and pre-treated. Then, an automated production line is used to composite and bond the three layers of materials according to the design structure, controlling the uniformity of the adhesive layer thickness. Finally, product inspection is carried out, including appearance inspection, dimension measurement, and sampling acoustic absorption performance testing, to ensure that the product quality meets the design requirements. The purpose of this step is to transform the optimized design into an actual product and achieve the large-scale production of the composite acoustic ceiling for the terminal building.

[0052] The following will describe in detail the mathematical models or calculation processes involved in the present invention.

[0053] In step S01, establishing an acoustic absorption mathematical model based on the acoustic environment characteristics of the terminal building involves multiple calculation processes and equations. First, the propagation of sound waves in porous materials follows the Helmholtz equation, which is specifically expressed as follows:

[0054]

[0055] In the formula, p is the sound pressure, with the unit of Pascal (Pa); k is the wave number, with the unit of radian per meter (rad / m), k = ω / c; ω is the angular frequency, with the unit of radian per second (rad / s); c is the sound wave propagation speed, with the unit of meter per second (m / s).

[0056] This equation describes the basic law of sound wave propagation in a homogeneous medium. By solving this equation, the propagation characteristics of sound waves in materials can be obtained. The Helmholtz equation takes into account the spatial distribution and frequency characteristics of sound waves and is the basic equation for acoustic analysis.

[0057] Sound impedance is an important parameter to describe the sound absorption characteristics of materials. The calculation formula for the equivalent sound impedance of the combined structure of through-hole ceramics and micro-perforated aluminum plates is as follows:

[0058]

[0059] In the formula, Z eq is the equivalent sound impedance, with the unit of sound impedance unit (rayl); Z tc is the sound impedance of the through-hole ceramics, with the unit of sound impedance unit (rayl); Z mp is the sound impedance of the micro-perforated aluminum plate, with the unit of sound impedance unit (rayl); Z ab is the sound impedance of the sound-absorbing backing material, with the unit of sound impedance unit (rayl); Z g is the sound impedance of the air layer, with the unit of sound impedance unit (rayl).

[0060] This formula is based on the series-parallel principle of sound impedance and synthesizes the sound impedances of the materials in each layer of the composite structure. Among them, the series part is directly added, and the parallel part is in the form of the reciprocal of the sum of the reciprocals, which reflects the transmission characteristics of sound waves at the interfaces of different materials.

[0061] The calculation of the sound impedance of the through-hole ceramics is as follows:

[0062]

[0063] In the formula, ρ0 is the air density, with the unit of kilograms per cubic meter (kg / m 3 ), which is 1.29 under standard conditions; c0 is the speed of sound in air, with the unit of meters per second (m / s), which is 343 under standard conditions; η is the air dynamic viscosity, with the unit of Pascal-second (Pa·s), which is 1.81×10 -5 under standard conditions; j is the imaginary unit; ω is the angular frequency, with the unit of radians per second (rad / s); a is the pore diameter of the through-hole ceramics, with the unit of meters (m); d tc is the thickness of the through-hole ceramics, with the unit of meters (m); p tc is the perforation rate of the through-hole ceramics, dimensionless.

[0064] This formula is based on the Delany-Bazley model and takes into account the viscous loss effect inside the porous material. The formula contains an imaginary term, indicating the phase change of sound waves when propagating in the through-hole ceramics. The viscous loss effect is inversely proportional to the frequency, with larger viscous losses at low frequencies and smaller ones at high frequencies, which conforms to the actual physical laws.

[0065] The acoustic impedance of the micro-perforated aluminum plate is calculated as follows:

[0066]

[0067] In the formula, Z mp is the acoustic impedance of the micro-perforated aluminum plate, with the unit of acoustic impedance unit (rayl); ω is the angular frequency, with the unit of radians per second (rad / s); ρ0 is the air density, with the unit of kilograms per cubic meter (kg / m 3 ); η is the air dynamic viscosity, with the unit of Pascal-second (Pa·s); p mp is the perforation rate of the micro-perforated aluminum plate, dimensionless; t mp is the thickness of the micro-perforated aluminum plate, with the unit of meter (m); a mp is the pore diameter of the micro-perforated aluminum plate, with the unit of meter (m).

[0068] This formula is based on the micro-perforated plate theory and takes into account the viscous resistance and acoustic mass effect inside the small holes. The formula is divided into a real part and an imaginary part. The real part represents the acoustic resistance and is proportional to the square root of the frequency; the imaginary part represents the acoustic mass effect and is proportional to the frequency. This complex relationship enables the micro-perforated aluminum plate to have good sound absorption performance in the medium and high frequency ranges.

[0069] The acoustic impedance of the sound-absorbing backing material is calculated as follows:

[0070]

[0071] In the formula, Z ab is the acoustic impedance of the sound-absorbing backing material, with the unit of acoustic impedance unit (rayl); ρ0 is the air density, with the unit of kilograms per cubic meter (kg / m 3 ); c0 is the speed of sound in air, with the unit of meters per second (m / s); ρ ab is the density of the sound-absorbing backing material, with the unit of kilograms per cubic meter (kg / m 3 ); f is the frequency, with the unit of Hertz (Hz).

[0072] This formula is based on an empirical model and is obtained by fitting a large amount of experimental data. The power relationship in the formula reflects the non-linear influence of material density and frequency on sound absorption performance. The acoustic impedance of the backing material decreases with the increase of frequency and increases with the increase of material density, which is consistent with the actual physical characteristics of the sound-absorbing material.

[0073] The acoustic impedance of the air layer is calculated as follows:

[0074]

[0075] In the formula, Z g is the acoustic impedance of the air layer, with the unit of acoustic impedance unit (rayl); ρ0 is the air density, with the unit of kilograms per cubic meter (kg / m3 ); c0 is the speed of sound in air, with the unit of meters per second (m / s); ω is the angular frequency, with the unit of radians per second (rad / s); d g is the thickness of the air layer, with the unit of meters (m).

[0076] This formula describes the acoustic impedance characteristics of an air layer with a finite thickness and is derived based on the wave theory. The cotangent function in the formula results in the periodic variation of the acoustic impedance of the air layer. When the thickness of the air layer is an odd multiple of half the wavelength, the acoustic impedance value approaches zero, and the sound absorption effect is the best.

[0077] The sound absorption coefficient of the composite structure is calculated as follows:

[0078]

[0079] In the formula, α is the sound absorption coefficient, dimensionless, and its value range is from 0 to 1; Z eq is the equivalent acoustic impedance, with the unit of acoustic impedance unit (rayl); ρ0 is the air density, with the unit of kilograms per cubic meter (kg / m 3 ); c0 is the speed of sound in air, with the unit of meters per second (m / s).

[0080] This formula is based on the acoustic wave reflection theory and describes the reflection and absorption relationship when the acoustic wave encounters the material surface. When the equivalent acoustic impedance of the material is equal to the characteristic impedance of the air, the sound absorption coefficient reaches the maximum value of 1, indicating complete absorption; when the difference between the two is large, the sound absorption coefficient approaches 0, indicating almost all reflection.

[0081] In step S06, organizing the experimentally measured data into the sound absorption vector matrix involves feature extraction and matrix construction. The sound absorption vector is defined as follows:

[0082]

[0083] In the formula, is an eight-dimensional sound absorption vector; α low is the average sound absorption coefficient in the low-frequency band (100 - 500 Hz), dimensionless, and its value range is from 0 to 1; α mid is the average sound absorption coefficient in the middle-frequency band (500 - 2000 Hz), dimensionless, and its value range is from 0 to 1; α high is the average sound absorption coefficient in the high-frequency band (2000 - 8000 Hz), dimensionless, and its value range is from 0 to 1; R ref is the acoustic wave reflectivity index, dimensionless, and its value range is from 0 to 1; V resp is the micro-vibration frequency response coefficient, dimensionless; E decay is the sound energy attenuation rate, with the unit of decibels per second (dB / s); P comp is the comprehensive performance index one, dimensionless; B widthis the sound absorption bandwidth, with the unit of hertz (Hz).

[0084] This vector comprehensively describes the sound absorption performance characteristics of the material, including the sound absorption ability in different frequency bands, the sound wave reflection characteristics, the material vibration response, the energy attenuation rate, and the comprehensive performance index. Each component is obtained through different experimental methods and jointly constitutes a complete description of the material's sound absorption performance.

[0085] The calculation formula for the average sound absorption coefficient is as follows:

[0086]

[0087] In the formula, α band is the average sound absorption coefficient of the frequency band (including low, medium, and high), dimensionless; n is the number of sampling points within the frequency band; α(f i ) is the sound absorption coefficient at the frequency f i , dimensionless.

[0088] This formula obtains the average sound absorption performance of the frequency band by taking the arithmetic average of the sound absorption coefficients at multiple frequency points within each frequency band. The division of the low, medium, and high frequency bands takes into account the spectral characteristics of the terminal building noise and the sensitivity of the human ear to sounds of different frequencies.

[0089] The calculation formula for the sound wave reflectivity index is as follows:

[0090]

[0091] In the formula, R ref is the sound wave reflectivity index, dimensionless; f min is the lowest frequency, with the unit of hertz (Hz), taking 100; f max is the highest frequency, with the unit of hertz (Hz), taking 8000; w(f) is the frequency weight function, dimensionless; α(f) is the sound absorption coefficient at the frequency f, dimensionless.

[0092] This formula calculates the average reflection performance considering the frequency weight within the above three frequency ranges. The frequency weight function w(f) is determined according to the spectral characteristics of the terminal building noise, making the calculation result more in line with the actual application requirements.

[0093] The calculation formula for the micro-vibration frequency response coefficient is as follows:

[0094]

[0095] In the formula, V resp is the micro-vibration frequency response coefficient, dimensionless; f min is the lowest frequency, with the unit of hertz (Hz), taking 100; f maxis the highest frequency, with the unit of Hertz (Hz), taking 8000; A(f) is the vibration acceleration amplitude at frequency f, with the unit of meters per second squared (m / s 2 ); S(f) is the power spectral density of the terminal building noise at frequency f, with the unit of Pascal squared per Hertz (Pa 2 / Hz).

[0096] This formula characterizes the vibration response characteristics of the material under the action of typical terminal building noise. Through the weighted integration of the material vibration response and the noise spectrum, it reflects the vibration sensitivity of the material to noises of different frequencies and is an important indicator for evaluating the resonance characteristics of the material.

[0097] The calculation formula for the sound energy attenuation rate is as follows:

[0098]

[0099] In the formula, E decay is the sound energy attenuation rate, with the unit of decibels per second (dB / s); Δt 60 is the time required for the sound pressure level to attenuate by 60 decibels, with the unit of seconds (s).

[0100] Based on the definition of the reverberation time, this formula describes how fast the sound energy attenuates in the material. The greater the attenuation rate, the stronger the ability of the material to absorb sound energy and the better the sound absorption effect. Δt 60 is obtained through the impulse response method and is a standard parameter for acoustic evaluation.

[0101] The calculation formula for the comprehensive performance index is as follows:

[0102]

[0103] In the formula, P comp is the comprehensive performance index, dimensionless; w i is the weight coefficient for each frequency band, dimensionless, α i is the average sound absorption coefficient for each frequency band, dimensionless; λ is the micro-vibration weight factor, dimensionless, with a value range of 0.1 to 0.5; V resp is the micro-vibration frequency response coefficient, dimensionless.

[0104] This formula comprehensively considers the sound absorption performance and micro-vibration characteristics of the material. The numerator part represents the weighted average sound absorption performance, and the denominator part considers the negative impact of micro-vibrations on the sound absorption effect. When the micro-vibration response of the material is too large, it will lead to a decline in the comprehensive performance, which is in line with the actual physical situation.

[0105] The calculation formula for the sound absorption bandwidth is as follows:

[0106] B width =f high-f low ;

[0107] wherein, B width is the sound absorption bandwidth, with the unit of hertz (Hz); f high is the highest frequency at which the sound absorption coefficient is greater than 0.7, with the unit of hertz (Hz); f low is the lowest frequency at which the sound absorption coefficient is greater than 0.7, with the unit of hertz (Hz).

[0108] This formula defines the width of the effective sound absorption frequency range of the material, reflecting the frequency band coverage ability of the material's sound absorption performance. The wider the sound absorption bandwidth, the better the sound absorption effect of the material in a wider frequency range, and the stronger the applicability.

[0109] The construction of the sound absorption vector matrix is as follows:

[0110]

[0111] wherein, A is the sound absorption vector matrix; is the sound absorption vector corresponding to the parameters of the i-th group of clustering center points, and i takes values from 1 to 6.

[0112] This matrix organizes the sound absorption performance data corresponding to the parameters of 6 groups of clustering center points into a structured form, and each row represents the sound absorption performance characteristics of a group of material parameters. This matrix form is convenient for data analysis and model training, and is a bridge connecting material parameters and sound absorption performance.

[0113] In step S08, the calculation formula of the comprehensive sound absorption coefficient is as follows:

[0114]

[0115] wherein, α comp is the comprehensive sound absorption coefficient, dimensionless, and the value range is from 0 to 1; w i is the weight coefficient at the frequency f i , dimensionless, α(f i ) is the sound absorption coefficient at the frequency f i , dimensionless; f i is the center frequency of the i-th frequency band, with the unit of hertz (Hz).

[0116] This formula obtains the comprehensive sound absorption performance of the material by weighted averaging the sound absorption coefficients of 20 frequency bands. The weight coefficient w i is determined according to the noise spectrum distribution of the terminal building, making the calculation result more in line with the actual application requirements. The frequency band division follows the 1 / 3 octave principle, covering the human ear sensitive frequency range from 100 Hz to 8000 Hz.

[0117] Specifically, the principle of the present invention is as follows: The technical principle of the present invention is based on the physical mechanism of sound wave propagation in porous composite materials and a data-driven multi-band sound absorption optimization method. First, through the wave equation and acoustic impedance theory, a mathematical model describing the propagation behavior of sound waves in a combined structure of through-hole ceramics and micro-hole aluminum plates is established. This model takes into account the geometric characteristics of the materials (such as thickness, pore diameter, pore spacing, and perforation rate) and physical characteristics (such as density and elasticity) of the absorption mechanism of sound waves at different frequencies, providing a theoretical basis for understanding and optimizing multi-band sound absorption performance.

[0118] In terms of experimental design, the present invention uses the K-means clustering algorithm to scientifically divide the material parameter space to ensure that a limited number of experimental samples can represent the main characteristics of the parameter space. By collecting typical noises in 20 frequency bands in the terminal building environment as test sound sources, measuring the sound wave reflectivity, micro-vibration spectrum, and sound energy attenuation curve of the composite sound absorption ceiling sample, an eight-dimensional sound absorption vector comprehensively reflecting the sound absorption performance of the material is obtained, providing high-quality training data for the subsequent machine learning model.

[0119] The key innovation of the present invention lies in the design of a five-layer feedforward neural network with a special structure. The attention mechanism in this network architecture can automatically identify the mutual correlations between different frequencies, dynamically adjust the network's attention to the sound absorption characteristics of each frequency band, and make the model more adaptable to the characteristics of multi-band noise in the terminal building. At the same time, the outlier enlargement function introduced in the network is used to improve the sensitivity to the non-linear sound absorption characteristic interval, and can accurately capture the resonance or interference phenomena generated by certain specific combinations of material parameters. These phenomena often lead to sudden changes in sound absorption performance and are important clues for optimizing multi-band sound absorption effects.

[0120] Through a well-trained neural network model, the present invention can efficiently predict and screen out the parameter combinations with the optimal sound absorption effect on multi-band noise in the terminal building from the theoretically possible material parameter space. Compared with the traditional trial-and-error method, this method based on mathematical models and machine learning greatly improves the design efficiency. More importantly, it can discover those material parameter combinations that are difficult to identify manually and perform outstandingly in a multi-band noise environment, thus realizing the optimization of the sound absorption performance of the composite sound absorption ceiling in the terminal building across the entire frequency band.

[0121] The following provides a specific Embodiment 1 of the present invention. The specific implementation of each step in this Embodiment 1 is described in detail as follows.

[0122] The specific implementation of step S01 is to establish a sound absorption mathematical model based on the acoustic environment characteristics of the terminal building. First, collect the acoustic environment data in the terminal building, including key parameters such as the noise distribution characteristics, reverberation time, and sound pressure level distribution in each frequency band. Then apply the Helmholtz equation to describe the propagation behavior of sound waves in various materials. The equation is expressed as:

[0123]

[0124] In the formula, p is the sound pressure with the unit of Pascal (Pa); k is the wave number with the unit of radian per meter (rad / m), and k = ω / c; ω is the angular frequency with the unit of radian per second (rad / s); c is the sound wave propagation speed with the unit of meter per second (m / s).

[0125] Next, establish the theoretical model of acoustic impedance and calculate the equivalent acoustic impedance of the combined structure of through-hole ceramics and micro-hole aluminum plates:

[0126]

[0127] In the formula, Z eq is the equivalent acoustic impedance with the unit of acoustic impedance unit (rayl); Z tc is the acoustic impedance of the through-hole ceramics; Z mp is the acoustic impedance of the micro-hole aluminum plate; Z ab is the acoustic impedance of the sound-absorbing backing material; Z g is the acoustic impedance of the air layer.

[0128] The acoustic impedance of the through-hole ceramics is calculated as follows:

[0129]

[0130] In the formula, ρ0 is the air density; c0 is the sound speed in air; η is the air dynamic viscosity; j is the imaginary unit; a is the aperture of the through-hole ceramics; d tc is the thickness of the through-hole ceramics; p tc is the perforation rate of the through-hole ceramics.

[0131] The acoustic impedance of the micro-hole aluminum plate is calculated as follows:

[0132]

[0133] In the formula, p mp is the perforation rate of the micro-hole aluminum plate; t mp is the thickness of the micro-hole aluminum plate; a mp is the aperture of the micro-hole aluminum plate.

[0134] The acoustic impedance of the sound-absorbing backing material is calculated as follows:

[0135]

[0136] In the formula, ρ ab is the density of the sound-absorbing backing material; f is the frequency.

[0137] The acoustic impedance of the air layer is calculated as follows:

[0138]

[0139] where d g is the thickness of the air layer.

[0140] The sound absorption coefficient of the composite structure is calculated as follows:

[0141]

[0142] where α is the sound absorption coefficient, dimensionless, and its value range is from 0 to 1.

[0143] By combining the Delany-Bazley model with the micro-perforated plate theory, a calculation formula for the sound absorption parameters applicable to the composite structure is constructed. The purpose of this step is to establish a theoretical basis, provide a mathematical basis for subsequent material parameter optimization, and make the design process based on scientific theory rather than experience.

[0144] The specific implementation of step S02 is to calculate the range of material parameters required for the optimal sound absorption effect according to the established sound absorption mathematical model. First, the sound absorption mathematical model in step S01 is programmed to construct a calculation program. Then, a parameter sensitivity analysis is performed on each material parameter to determine the influence weight of each parameter on the sound absorption performance. The Monte Carlo simulation method is used to randomly generate 10,000 groups of parameter combinations within the initial large range, and calculate the theoretical sound absorption performance of each combination. The parameter sensitivity analysis uses the analysis of variance method to calculate the contribution rate of each parameter to the sound absorption coefficient:

[0145]

[0146] where S i is the sensitivity index of the i-th parameter; Var i is the variance of the sound absorption coefficient when only the i-th parameter changes; n is the total number of parameters. The calculation results are screened, and the parameter combinations with a theoretical sound absorption coefficient greater than 0.7 in the medium and high frequency bands (500 - 4000 Hz) are retained. Finally, the effective range of each parameter is determined through statistical analysis, and the material parameter range is obtained as follows: the thickness of the through-hole ceramic is 5 to 15 mm, the thickness of the micro-perforated aluminum plate is 1 to 3 mm, the hole diameter of the micro-perforated aluminum plate is 0.2 to 2 mm, the hole pitch of the micro-perforated aluminum plate is 3 to 20 mm, the perforation rate of the micro-perforated aluminum plate is 5% to 15%, and the density of the backing material is 30 to 80 kg / m³. The purpose of this step is to determine the theoretically feasible material parameter range, providing guidance for subsequent experimental design and production.

[0147] The specific implementation of step S03 is to perform a cluster analysis on the material parameter range. First, the parameter combinations within the material parameter range determined in step S02 are subjected to data standardization processing to eliminate the influence of the dimensional difference of different parameters on the clustering result. The standardization processing uses the Z-score method:

[0148]

[0149] where Z i is the standardized parameter value; X i is the original parameter value; μ i is the mean value of parameter i; σ i is the standard deviation of parameter i. Then, using the K-means clustering algorithm, the standardized parameter space is divided into 6 regions. The objective function of K-means clustering is:

[0150]

[0151] where J is the clustering evaluation index; k is the number of clusters, taking the value of 6; n j is the number of samples in the j-th cluster; is the i-th sample in the j-th cluster; c j is the center point of the j-th cluster. The selection of the K value is based on the silhouette coefficient evaluation. When K = 6, the silhouette coefficient reaches the optimal value of 0.68, indicating good clustering effect. The formula for calculating the silhouette coefficient is:

[0152]

[0153] where SC is the silhouette coefficient; N is the total number of samples; a i is the average distance between sample i and other samples in the same cluster; b i is the average distance between sample i and samples in the nearest different cluster. Then, calculate the center point coordinates of each cluster region as the representative parameter combination. Finally, inverse-standardize the clustering center point parameters to actual physical quantities to obtain 6 groups of clustering center point parameters as the basis for making experimental samples. The purpose of this step is to scientifically extract a small number of representative parameter combinations from a large number of theoretically feasible parameter combinations for experimental verification, which not only ensures the coverage of the parameter space but also reduces the experimental cost.

[0154] The specific implementation of step S04 is to make a composite sound-absorbing ceiling sample for experiments according to the clustering center point parameters. First, according to the 6 groups of clustering center point parameters, purchase through-hole ceramic plates, micro-hole aluminum plates and sound-absorbing backing materials that meet the requirements. The through-hole ceramics are prepared from alumina raw materials through a high-temperature sintering process, the micro-hole aluminum plates are processed from aluminum alloy plates through a precision laser drilling process, and the backing material is selected as polyester fiber sound-absorbing cotton. Then, cut the micro-hole aluminum plates and through-hole ceramics to a size of 600 mm × 600 mm according to the design dimensions. Next, use an environmentally friendly polyurethane adhesive to bond the three layers of materials together in the order of "through-hole ceramic layer - micro-hole aluminum plate layer - sound-absorbing backing material". Finally, install an aluminum alloy frame on the edge of the sample to form a composite sound-absorbing ceiling sample of standard specifications. The purpose of this step is to transform the theoretical design into an actual sample and provide an experimental object for subsequent performance testing.

[0155] The specific implementation of step S05 is to establish a test experiment on the sound absorption performance of the terminal ceiling. First, select 20 typical locations such as the waiting hall, check-in area, and security check area in the terminal, collect the environmental noise spectra at each location, and extract 20 band characteristic noises in the range of 100 Hz to 8000 Hz as the test sound sources. Then, build an acoustic performance test platform, including a reverberation chamber and an impedance tube test system. The volume of the reverberation chamber is 200 cubic meters, meeting the ISO354 standard. Next, install the 6 groups of composite sound-absorbing ceiling samples made on the test platform, and measure the sound wave reflectivity (using the standing wave ratio method), micro-vibration spectrum (using the laser Doppler vibration measurement method), and sound energy attenuation curve (using the impulse response method) of the samples at each frequency band through devices such as sound pressure sensors and acceleration sensors. Finally, statistically process the measured data, eliminate outliers, and ensure the accuracy and reliability of the data. The purpose of this step is to obtain the performance data of each composite sound-absorbing ceiling sample in the actual acoustic environment and provide a real basis for model training.

[0156] The specific implementation of step S06 is to organize the data measured in the experiment into a sound absorption vector matrix. First, perform feature extraction on the original data such as the sound wave reflectivity, micro-vibration spectrum, and sound energy attenuation curve obtained in step S05, including the average sound absorption coefficients in the low-frequency band (100 - 500 Hz), mid-frequency band (500 - 2000 Hz), and high-frequency band (2000 - 8000 Hz), as well as features such as the peak sound absorption frequency point and sound absorption bandwidth. The sound absorption vector is defined as follows:

[0157]

[0158] In the formula, is an eight-dimensional sound absorption vector; α low is the average sound absorption coefficient in the low-frequency band; α mid is the average sound absorption coefficient in the mid-frequency band; α high is the average sound absorption coefficient in the high-frequency band; R ref is the sound wave reflectivity index; V resp is the micro-vibration frequency response coefficient; E decay is the sound energy attenuation rate; P comp is the comprehensive performance index; B width is the sound absorption bandwidth.

[0159] The calculation formula for the average sound absorption coefficient is as follows:

[0160]

[0161] In the formula, α band is the average sound absorption coefficient of the frequency band; n is the number of sampling points in the frequency band; α(f i ) is the sound absorption coefficient at the frequency f i .

[0162] The calculation formula for the acoustic reflectivity index is as follows:

[0163]

[0164] In the formula, f min is the lowest frequency, taking 100 Hz; f max is the highest frequency, taking 8000 Hz; w(f) is the frequency weight function; α(f) is the sound absorption coefficient at frequency f.

[0165] The calculation formula for the micro-vibration frequency response coefficient is as follows:

[0166]

[0167] In the formula, A(f) is the vibration acceleration amplitude at frequency f; S(f) is the power spectral density of the terminal building noise at frequency f.

[0168] The calculation formula for the sound energy attenuation rate is as follows:

[0169]

[0170] In the formula, Δt 60 is the time required for the sound pressure level to attenuate by 60 dB.

[0171] The calculation formula for the comprehensive performance index is as follows:

[0172]

[0173] In the formula, w i is the weight coefficient for each frequency band, α i is the average sound absorption coefficient for each frequency band; λ is the micro-vibration weight factor, and its value range is from 0.1 to 0.5.

[0174] The calculation formula for the sound absorption bandwidth is as follows:

[0175] B width = f high - f low ;

[0176] In the formula, f high is the highest frequency at which the sound absorption coefficient is greater than 0.7; f low is the lowest frequency at which the sound absorption coefficient is greater than 0.7.

[0177] Then, combine the 6 groups of clustering center point parameters and the corresponding eight-dimensional sound absorption vectors to form a training data set and form a sound absorption vector matrix:

[0178]

[0179] Wherein, is the sound absorption vector corresponding to the clustering center point parameter of the i-th group, and i ranges from 1 to 6.

[0180] Finally, the data is normalized so that the data of each dimension is scaled to the range of 0 to 1, which is convenient for neural network training. The purpose of this step is to convert the experimental data into a structured data set suitable for machine learning and prepare for subsequent model training.

[0181] The specific implementation of step S07 is to design a five-layer feedforward neural network architecture and train the sound absorption parameter effect model. First, construct the neural network architecture, including an input layer with 6 nodes (corresponding to 6 material parameters), a first hidden layer with 32 nodes, a second hidden layer with 64 nodes, a third hidden layer with 32 nodes, and an output layer with 8 nodes (corresponding to the eight-dimensional sound absorption vector). Optionally, the forward propagation process can be expressed as:

[0182] Z [1] = W [1] X + b [1] ;

[0183] A [1] = ReLU(Z [1] );

[0184] Z [2] = W [2] A [1] + b [2] ;

[0185] A [2] = tanh(Z [2] );

[0186]

[0187] A [3] = ReLU(Z [3] );

[0188] Z [4] = W [4] A [3] + b [4] ;

[0189]

[0190] Wherein, X is the input material parameter vector; W [l] and b [l] are the weight matrix and bias vector of the l-th layer; Z [l] is the linear output of the l-th layer; A [l] is the activation output of the l-th layer; f outlier is the outlier increasing function; is the predicted sound absorption vector. The ReLU activation function is used in the first hidden layer and the third hidden layer, the tanh activation function is used in the second hidden layer, and an outlier amplification function is added between the second hidden layer and the third hidden layer. The outlier amplification function is defined as:

[0191] f outlier (x) = x + α·sgn(x)·|x| β ·I(|x| > θ);

[0192] where x is the input value; α is the enhancement coefficient with a value of 0.5; β is the non - linear exponent with a value of 2; sgn is the sign function; I is the indicator function; θ is the threshold with a value of 0.8. This function is used to enhance the sensitivity recognition of extreme sound absorption performance parameter combinations. Then, the attention mechanism is implemented to automatically adjust the network's attention to the sound absorption characteristics of different frequency bands through a self - learning weight matrix:

[0193] Att = softmax(W a ·A [2] );

[0194]

[0195] where Att is the attention weight vector; W a is the attention parameter matrix; softmax is the normalization function; ⊙ is the element - wise product; is the weighted output of the hidden layer. Then, the Adam optimizer and the mean - squared error loss function are used, and the model is trained using the dataset constructed in step S06. The training adopts a 5 - fold cross - validation method, the initial learning rate is set to 0.001, and a learning rate decay strategy is used. The loss function is defined as:

[0196]

[0197] where L is the mean - squared error loss; m is the number of samples; is the predicted value of the i - th sample; Y i is the true value of the i - th sample. Finally, the training stops when the validation set error is lower than 0.01 or the number of iterations reaches 5000 times, and the model parameters are saved. The purpose of this step is to establish a non - linear mapping model between material parameters and sound absorption performance to achieve the prediction of sound absorption performance for any parameter combination.

[0198] The specific implementation of step S08 is to use the sound absorption parameter effect model to screen the optimal material parameter combination. First, within the range of material parameters determined in step S02, the Latin hypercube sampling method is used to generate 10,000 groups of uniformly distributed material parameter combinations. The mathematical representation of Latin hypercube sampling is:

[0199]

[0200] wherein, X ij is the normalized value of the j-th parameter of the i-th sample; π ij is the random permutation function; U ij is a random number uniformly distributed in [0, 1]; n is the total number of samples. Then, these parameter combinations are input into the neural network model trained in step S07 to predict its sound absorption performance, obtaining 10,000 groups of material parameters and the corresponding eight-dimensional sound absorption vectors. Next, according to the noise spectrum characteristics of the terminal building, the weight coefficients of the sound absorption performance in each frequency band are determined, and the comprehensive sound absorption coefficient is calculated:

[0201]

[0202] wherein, α comp is the comprehensive sound absorption coefficient; w i is the weight coefficient at the frequency f i , α(f i ) is the sound absorption coefficient at the frequency f i ; f i is the center frequency of the i-th frequency band. Meanwhile, a material cost evaluation model is established, and the production cost of each group of parameters is calculated according to the material usage corresponding to each parameter:

[0203]

[0204] wherein, Cost is the production cost per unit area; c1, c2, c3, c4 are the material unit price coefficients; d tc is the thickness of the through-hole ceramic; t mp is the thickness of the micro-perforated aluminum plate; p mp is the perforation rate of the micro-perforated aluminum plate; ρ ab is the density of the backing material. Finally, the Pareto optimization method is adopted to find a balance between the two objectives of the comprehensive sound absorption coefficient and the production cost, and the optimal material parameter combination with the highest comprehensive sound absorption coefficient and the lowest manufacturing cost is selected. The mathematical representation of the Pareto optimum is:

[0205]

[0206] wherein, P is the Pareto optimal solution set; Ω is the feasible solution space; f1 is the negative of the comprehensive sound absorption coefficient (the goal is to minimize); f2 is the production cost. The purpose of this step is to find the optimal solution for performance and cost among a large number of possible parameter combinations.

[0207] The specific implementation of step S09 is to produce the composite sound-absorbing ceiling according to the optimal material parameter combination obtained through screening. First, based on the optimal material parameter combination, a detailed production process flow and quality control standards are formulated. Then, raw materials that meet the parameter requirements are purchased, including perforated ceramics, micro-hole aluminum plates, and sound-absorbing backing materials. Next, the micro-hole aluminum plates are precisely punched using numerical control equipment to ensure that the hole diameter, hole spacing, and perforation rate meet the design requirements. At the same time, the perforated ceramics and backing materials are cut and pre-treated. Then, an automated production line is used to composite and bond the three layers of materials according to the design structure, controlling the uniformity of the adhesive layer thickness. Finally, product inspection is carried out, including appearance inspection, dimension measurement, and sampling sound-absorbing performance testing, to ensure that the product quality meets the design requirements. The purpose of this step is to transform the optimized design into an actual product and achieve the large-scale production of the composite sound-absorbing ceiling in the terminal building.

[0208] The quality inspection standards include three aspects: appearance quality, dimensional accuracy, and sound-absorbing performance. The appearance quality inspection requires no obvious scratches, cracks, and stains, uniform color, and smooth edges; the dimensional accuracy inspection requires that the length and width dimensional deviation does not exceed ±2 mm, and the thickness deviation does not exceed ±0.5 mm; the sound-absorbing performance inspection requires that the relative deviation between the average sound-absorbing coefficient of the sampling test and the design value does not exceed 5%. Through strict quality control, it is ensured that the product has stable and reliable sound-absorbing performance and meets the special requirements of the acoustic environment in the terminal building. This production method comprehensively applies acoustic theory, materials science, data mining, and artificial intelligence technologies. Through systematic design optimization and production control, it realizes the efficient application of the composite sound-absorbing ceiling in the terminal building environment, solves the problems of unstable performance and poor adaptability of traditional sound-absorbing materials in special environments, and has significant practical value and innovative significance.

[0209] To better understand and implement the present invention, the following provides an embodiment 2 of a specific application scenario of the present invention: A research team was invited to design a composite sound-absorbing ceiling system for a certain airport terminal building. The research team first conducted a comprehensive analysis of the acoustic environment in the terminal building and found that there are significant differences in the noise characteristics of each functional area, as shown in Table 1:

[0210] Table 1 Analysis Table of Noise Characteristics of Each Functional Area in the Terminal Building

[0211] Functional area Noise level range / dB Main spectral characteristics / Hz Main noise source Reverberation time / s Central core area 78-85 500-2000 People flow, broadcast, echo 2.8-3.5 Waiting area 72-78 300-1500 People flow, electronic equipment 1.9-2.6 Check-in area 75-82 400-2500 Baggage conveyor, conversation 2.3-2.9 Commercial area 70-76 200-1000 Background music, conversation 1.5-2.2 Food and beverage area 74-80 300-1200 Tableware collision, conversation 1.8-2.4

[0212] According to step S01, the research team established an acoustic absorption mathematical model based on the acoustic environment characteristics of the terminal building. The Helmholtz equation was applied to describe the propagation behavior of sound waves in various materials, and a theoretical model of acoustic impedance was established to calculate the equivalent acoustic impedance of the combined structure of through-hole ceramics and micro-perforated aluminum plates. Through theoretical calculations, it was determined that the parameters of the micro-perforated aluminum plate were: hole diameter 0.5 mm, hole spacing 15.6 mm, and perforation rate 8%; the parameters of the through-hole ceramics were: thickness 12 mm, density 420 kg / m³, and compressive strength 1.35 MPa.

[0213] According to step S02, the research team conducted a parameter sensitivity analysis to determine the influence degree of each material parameter on the acoustic absorption performance, as shown in Table 2:

[0214] Table 2 Results of the sensitivity analysis of material parameters

[0215]

[0216]

[0217] Through the clustering analysis in step S03, the research team divided the material parameter space into 6 regions, and extracted the clustering center point parameters of each region as the basis for making experimental samples. Using the K-means clustering algorithm, the 6 groups of clustering center point parameters finally determined are shown in Table 3:

[0218] Table 3 Table of clustering center point parameters

[0219]

[0220] According to step S04, the research team made 6 groups of experimental composite acoustic ceiling panels. The through-hole ceramic layer was prepared by a special process, with an acoustic absorption and noise reduction coefficient ≥ 0.70 measured by the reverberation room method, a compressive strength ≥ 1.31 MPa, a density ≤ 450 kg / m³, no release of harmful substances, and an air purification function. The micro-perforated aluminum plate layer was made of aviation-grade aluminum alloy material and processed by a precision laser drilling process. The backing material was selected as environmentally friendly polyester fiber acoustic cotton, which had excellent flame retardancy and hygroscopicity.

[0221] According to step S05, the research team established an experimental test on the acoustic absorption performance of the terminal ceiling. In a reverberation room that meets the ISO354 standard, using 20 typical noise frequencies collected from the terminal as the test sound source, the sound wave reflectivity, micro-vibration spectrum, and sound energy attenuation curve of 6 groups of composite acoustic ceiling panels were measured. The experimental results showed that the panel numbered 1 had the most balanced acoustic absorption performance, especially outstanding in the middle frequency band (500 - 2000 Hz), as shown in Table 4:

[0222] Table 4 Test results of the acoustic absorption coefficient of the composite acoustic ceiling panel

[0223] Frequency / Hz Sample 1 Sample 2 Sample 3 Sample 4 Sample 5 Sample 6 125 0.45 0.42 0.48 0.39 0.46 0.36 250 0.58 0.55 0.62 0.52 0.60 0.50 500 0.76 0.72 0.68 0.75 0.65 0.78 1000 0.88 0.85 0.75 0.87 0.72 0.86 2000 0.82 0.80 0.72 0.83 0.70 0.84 4000 0.75 0.76 0.68 0.78 0.66 0.80

[0224] According to step S06, the research team organized the experimentally measured data into an acoustic absorption vector matrix. For each group of samples, eight-dimensional acoustic absorption vectors were calculated, including the average acoustic absorption coefficients in the low, medium, and high frequency bands, as well as the sound wave reflectivity index, micro-vibration frequency response coefficient, sound energy attenuation rate, comprehensive performance index, and acoustic absorption bandwidth, as shown in Table 5:

[0225] Table 5 Eight-dimensional Acoustic Absorption Vector Table of Composite Acoustic Ceiling Samples

[0226]

[0227]

[0228] According to step S07, the research team designed a five-layer feedforward neural network architecture and trained an acoustic absorption parameter effect model. The network includes an input layer with 6 nodes (corresponding to 6 material parameters), a first hidden layer with 32 nodes, a second hidden layer with 64 nodes, a third hidden layer with 32 nodes, and an output layer with 8 nodes (corresponding to the eight-dimensional acoustic absorption vector). The ReLU activation function is used in the first and third hidden layers, the tanh activation function is used in the second hidden layer, and an outlier increasing function is added between the second and third hidden layers. At the same time, the attention mechanism is implemented to automatically adjust the weight distribution of the network for the acoustic absorption characteristics of different frequency bands through a self-learning weight matrix. During the training process, the Adam optimizer is used, and the learning rate is initially set to 0.001. After 4286 iterations, the validation set error drops to 0.0095, meeting the stopping condition. The change of the loss function during the model training process is shown in Table 6:

[0229] Table 6 Neural Network Model Training Process Table

[0230]

[0231] According to step S08, the research team used the trained neural network model to perform augmented sampling on the material parameter range, generating 10,000 groups of material parameter combinations and their predicted acoustic absorption performances. According to the noise spectrum characteristics of the terminal building, the weight coefficients of the acoustic absorption performances in each frequency band are determined: 0.25 for the low frequency band, 0.55 for the medium frequency band, and 0.20 for the high frequency band, and the comprehensive acoustic absorption coefficient is calculated. At the same time, a material cost evaluation model is established to calculate the production costs of each parameter combination. Through Pareto optimization analysis, the optimal material parameter combination with the highest comprehensive acoustic absorption coefficient and the lowest manufacturing cost is selected, as shown in Table 7:

[0232] Table 7 Optimal Material Parameter Combination Table

[0233] Parameter name Optimal value Allowable deviation range Thickness of through-hole ceramic / mm 12.0 ±0.5 <![CDATA[Through-hole ceramic density / (kg / m 3 )]]> 420 ±10 Aperture of micro-hole aluminum plate / mm 0.5 ±0.05 Hole pitch of micro-hole aluminum plate / mm 15.6 ±0.2 Perforation rate of micro-hole aluminum plate / % 8.0 ±0.3 <![CDATA[Density of the backing material / (kg / m 3 )]]> 45 ±3

[0234] According to step S09, the research team carried out the production implementation of the composite sound-absorbing ceiling. First, a detailed production process flow and quality control standards were formulated, and then raw materials meeting the parameter requirements were purchased. The micro-perforated aluminum plate uses aerospace-grade aluminum alloy material (AL5052), with a thickness of 2.0 mm, processed by precision laser drilling technology, the hole diameter is 0.5 ± 0.05 mm, the hole pitch is 15.6 ± 0.2 mm, and the perforation rate is 8.0 ± 0.3%. The through-hole ceramic uses an alumina-based material with a special formula, prepared by high-temperature (1350 °C) sintering technology, with a thickness of 12.0 ± 0.5 mm, a density of 420 ± 10 kg per cubic meter, and a compressive strength of 1.35 MPa. The backing material is selected as environmentally friendly polyester fiber sound-absorbing cotton, with a density of 45 ± 3 kg per cubic meter and a thickness of 25 mm. An automated production line is used to composite and bond the three-layer materials in the order of "through-hole ceramic layer - micro-perforated aluminum plate layer - sound-absorbing backing material", and the thickness uniformity of the adhesive layer is controlled within ±0.2 mm.

[0235] After production, a comprehensive quality inspection and performance test were carried out on the composite sound-absorbing ceiling. The inspection results showed that the qualified rate of the product appearance quality reached 98.5%, and the dimensional accuracy deviation was controlled within ±1.5 mm. The sound-absorbing performance test results showed that in the mid-frequency band of 500 - 2000 Hz, the average sound-absorbing coefficient reached 0.82, which was 0.02 higher than the design target value. The on-site test carried out in the actual installation area of the terminal showed that after installing the composite sound-absorbing ceiling, the noise level in each functional area was reduced by an average of 7.5 dB, and the reverberation time was shortened by 42%, as shown in Table 8:

[0236] Table 8 Comparison table of the installation effect of the composite sound-absorbing ceiling

[0237]

[0238] Traditional sound-absorbing solutions for terminals mainly use single-material sound-absorbing ceilings or simple composite structures. Single-material solutions such as glass wool sound-absorbing boards, mineral wool sound-absorbing boards, perforated gypsum boards, etc., although having a certain sound-absorbing effect, often only perform well in specific frequency bands and are difficult to cope with the complex noise environment in the terminal. Simple composite structures such as the combination of perforated metal plates and glass wool have problems such as narrow sound-absorbing bandwidth, poor moisture resistance, and complex installation. In addition, traditional design methods mainly rely on engineering experience and simple acoustic theory calculations, lacking systematic optimization methods and being difficult to find the best balance between sound-absorbing performance and cost.

[0239] Compared with traditional methods, the composite sound-absorbing ceiling production method adopted in the present invention has brought the following remarkable improvements: First, by establishing an acoustic mathematical model based on the acoustic environment characteristics of the terminal building, the accurate prediction of the sound-absorbing performance of different material combinations is realized, making the design process based on scientific theory rather than experience. Second, the innovative combination of through-hole ceramics and micro-perforated aluminum plates is adopted to overcome the limitations of traditional materials, achieving multiple objectives such as high-efficiency sound absorption in a wide frequency band (the average sound absorption coefficient reaches 0.71 in the range of 125 - 4000 Hz), fire prevention, moisture resistance, beauty and durability. Third, the K-means clustering algorithm and neural network technology are introduced to establish a non-linear mapping model between material parameters and sound-absorbing performance, realizing the accurate optimization of sound-absorbing parameters. In particular, the application of the attention mechanism and the outlier amplification function effectively improves the recognition ability of extreme sound-absorbing performance parameter combinations, increasing the model prediction accuracy by 35%. Fourth, through the Pareto optimization method, while meeting the sound-absorbing performance requirements, the production cost is reduced by about 18%, achieving the best balance between performance and cost. Fifth, the actual application effect shows that the composite sound-absorbing ceiling designed and produced by this method reduces the average noise level in each functional area of the terminal building by 7.5 dB and shortens the reverberation time by 42%, significantly superior to the noise reduction effect of 4 - 5 dB and the reverberation time shortening rate of 25 - 30% of the traditional scheme. In summary, through innovative material combinations and systematic optimization methods, the present invention solves the problem of insufficient performance of traditional sound-absorbing ceilings in special environments in terminal buildings, providing a new technical path for the acoustic environment control of large public buildings.

[0240] It should be noted that the detailed explanations of the variables involved in the present invention are shown in Tables 9 and 10 below.

[0241] Table 9 Variable Explanation Table (First Part)

[0242]

[0243]

[0244] Table 10 Variable Explanation Table (Second Part)

[0245]

[0246] The above is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, which should all be covered by the protection scope of the present invention.

Claims

1. A method for producing a composite sound-absorbing ceiling suitable for a terminal building, characterized in that: include: Establish a mathematical model of sound absorption based on the acoustic environment characteristics of the terminal and calculate the material parameter range required for the optimal sound absorption effect; A cluster analysis was conducted on the material parameter range, and the K-means clustering algorithm was used to divide the material parameter range into multiple material parameter regions. The cluster center point parameters of each material parameter region were extracted as the basis for making experimental samples; experimental composite sound-absorbing ceiling samples were made according to the cluster center point parameters; a terminal ceiling sound absorption performance test experiment was established to measure the sound wave reflectivity, micro-vibration spectrum and sound energy attenuation curve of each composite sound-absorbing ceiling sample; the experimental measured data were organized into a sound absorption vector matrix to construct a neural network training set; a five-layer feedforward neural network architecture was designed and trained to obtain a sound absorption parameter effect model; the sound absorption parameter effect model was used to expand and sample the material parameter range, generate multiple groups of material parameter combinations and their predicted sound absorption performance, and screen out the optimal material parameter combination with high comprehensive sound absorption coefficient and low manufacturing cost; composite sound-absorbing ceilings were produced based on the optimal material parameter combination screened out.

2. The method for producing a composite sound-absorbing ceiling suitable for a terminal building according to claim 1, characterized in that: The mathematical model of sound absorption based on the acoustic environment characteristics of the terminal building is established by using the wave equation and acoustic impedance theory to analyze the propagation mechanism of sound waves of various frequencies in composite materials, and to construct a calculation formula for the sound absorption parameters of the combined structure of through-hole ceramics and microporous aluminum plates.

3. The method for producing a composite sound-absorbing ceiling suitable for a terminal building according to claim 2, characterized in that: The wave equation and acoustic impedance theory refer to the mathematical equations followed by sound waves when propagating in porous materials, including the Helmholtz equation and the characteristic impedance calculation formula, which are used to describe the propagation characteristics and boundary conditions of sound waves in different media.

4. The method for producing a composite sound-absorbing ceiling suitable for a terminal building according to claim 3, characterized in that: The material parameter range includes through-hole ceramic thickness of 5 to 15 mm, microporous aluminum plate thickness of 1 to 3 mm, microporous aluminum plate hole diameter of 0.2 to 2 mm, microporous aluminum plate hole spacing of 3 to 20 mm, microporous aluminum plate perforation rate of 5% to 15% and backing material density of 30 to 80 kg per cubic meter.

5. The method for producing a composite sound-absorbing ceiling suitable for a terminal building according to claim 4, characterized in that: The cluster center point parameters include the through-hole ceramic thickness, the microporous aluminum plate thickness, the hole diameter of the microporous aluminum plate, the hole spacing of the microporous aluminum plate, the perforation rate of the microporous aluminum plate and the density of the backing material.

6. The method for producing a composite sound-absorbing ceiling suitable for a terminal building according to claim 5, characterized in that: The terminal building ceiling sound absorption performance test experiment collects typical noises in 20 frequency bands in the terminal building as the test sound source.

7. The method for producing a composite sound-absorbing ceiling suitable for a terminal building according to claim 6, characterized in that: Micro-vibration spectrum refers to the frequency distribution curve of tiny vibrations generated by sound-absorbing materials when they are impacted by sound waves. It is measured by an acceleration sensor and obtained through Fourier transform. It is used to evaluate the absorption and resonance characteristics of materials to sound waves of different frequencies.

8. The method for producing a composite sound-absorbing ceiling suitable for a terminal building according to claim 7, characterized in that: The sound energy attenuation curve refers to the attenuation law of sound wave energy over time or distance after passing through the sound-absorbing material. It is measured and calculated by placing a sound intensity meter before and after the material.

9. The method for producing a composite sound-absorbing ceiling suitable for a terminal building according to claim 8, characterized in that: Each cluster center point parameter corresponds to a set of eight-dimensional sound absorption vectors, and a neural network training set is constructed. The input is the cluster center point parameter, and the output is the corresponding sound absorption vector matrix; the sound absorption vector is used to comprehensively describe the sound absorption performance characteristics of the material in different aspects, and each sound absorption vector contains eight dimensions of data; the sound absorption vector matrix is ​​a two-dimensional data structure composed of multiple sound absorption vectors, each row represents a sound absorption vector under a set of material parameter combinations, and each column represents the performance of different material combinations on the same sound absorption performance index.

10. The method for producing a composite sound-absorbing ceiling suitable for a terminal building according to claim 9, characterized in that: The five-layer feedforward neural network architecture is a fully connected structure that adopts an attention mechanism, including an input layer, three hidden layers and an output layer, and an outlier increase function added between the second hidden layer and the third hidden layer. The outlier increase function is used to enhance the sensitivity recognition of extreme sound absorption performance parameter combinations.

Images