Production method of road sound barrier with light weight, high strength, corrosion resistance and strong sound absorption

Abstract

The invention provides a production method of a road sound barrier with light weight, high strength, corrosion resistance and strong sound absorption, and belongs to the technical field of road sound barriers, the production method comprises the following steps: collecting a traffic noise sample by using a high-precision microphone array, extracting noise spectrum characteristics by using a Noi seSpecNet model, and determining that the main noise frequency band is 500-4000 Hz; optimal perforation parameters are calculated through the model; accurately perforating the UHPC board according to the optimized parameters; the perforated UHPC board, basalt fiber cloth, a through-hole ceramic board, a sound insulation felt and other materials are bonded according to the sequence and technology to form a multi-layer composite structure; curing and curing in a constant-temperature and constant-humidity environment; carrying out acoustic performance testing; and finally, the qualified sound barrier unit is installed on a foundation structure and subjected to protection treatment. According to the method, the technical problem that a road sound barrier in the prior art is difficult to realize efficient sound absorption for traffic noise of a specific frequency band is solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of road sound barriers, and specifically relates to a production method of a road sound barrier with light weight, high strength, strong corrosion resistance and strong sound absorption. Background Art

[0002] Road traffic noise is one of the main sources of urban environmental pollution. As a key facility for isolating traffic noise, sound barriers are widely used in areas such as highways, urban expressways, and along railways. Traditional sound barriers are mainly made of materials such as concrete slabs, metal plates, wooden boards, or plexiglass, and reduce the propagation of noise through physical mechanisms such as sound wave reflection, absorption, or diffraction. These sound barrier designs mainly rely on empirical methods and simple physical models, and usually adopt unified structural parameters to deal with various road environments.

[0003] Traditional sound barriers have obvious defects in sound absorption performance: Most sound barriers adopt uniform perforation or simple porous structures, lacking optimized designs for the main frequency band of traffic noise from 500 Hz to 4000 Hz, resulting in low absorption efficiency for specific frequency noise in actual applications; At the same time, the existing sound barrier designs lack accurate analysis and matching of the spectral characteristics of actual road noise, and cannot perform adaptive optimization for the noise characteristics under different road types and traffic flow conditions. In addition, traditional design methods are difficult to consider the influence of environmental temperature and humidity changes on acoustic performance.

[0004] Currently, there is a lack of a systematic method in the industry that can accurately analyze the spectral distribution and propagation characteristics of traffic noise based on actual road noise characteristic data, and optimize the structural parameters of sound barriers accordingly to achieve efficient sound absorption of traffic noise in specific frequency bands. Especially in complex and changeable road environments and traffic flow conditions, the existing technologies are difficult to provide sound barrier design solutions with strong pertinence and good adaptability. Summary of the Invention

[0005] In view of this, the present invention provides a production method of a road sound barrier with light weight, high strength, strong corrosion resistance and strong sound absorption, which can solve the technical problem that it is difficult for existing road sound barriers to achieve efficient sound absorption for traffic noise in specific frequency bands.

[0006] The present invention is implemented as follows: The present invention provides a production method for a lightweight, high-strength, corrosion-resistant, and highly sound-absorbing road sound barrier, which includes the following steps: Set up a microphone array at the road edge to collect traffic noise samples, and extract the traffic noise spectrum features through the NoiseSpecNet noise spectrum feature extraction model; Obtain the optimal parameter combination of the perforated plate according to the extracted traffic noise spectrum features; Perform precise perforation on the ordinary perforated UHPC plate according to the optimized parameters; Lay basalt fiber cloth on the back of the perforated UHPC plate; Bond the through-hole ceramic plate to the basalt fiber cloth; Paste a first non-perforated UHPC plate, a sound insulation felt, and a second non-perforated UHPC plate on the back of the through-hole ceramic plate in sequence to form a composite structure; Place the composite structure in a constant temperature and humidity curing chamber for curing; Conduct acoustic performance tests and mechanical performance tests on the cured sound barrier units; Install the qualified sound barrier units on the prefabricated foundation structure and perform waterproof, dustproof treatment, and anti-ultraviolet coating spraying.

[0007] Among them, the step of setting up a microphone array to collect traffic noise samples is specifically as follows: Set up multiple high-precision microphone arrays at the road edge, collect traffic noise samples at different times, input the collected traffic noise samples into the NoiseSpecNet noise spectrum feature extraction model, extract the traffic noise spectrum features, determine that the main noise frequency band is 500 Hz to 4000 Hz, and calculate the influence degree of the sound wave superposition effect at different frequencies on the human auditory sense.

[0008] Among them, the step of obtaining the optimal parameter combination of the perforated plate is specifically as follows: According to the processing results of the NoiseSpecNet noise spectrum feature extraction model, extract the traffic noise spectrum distribution data, the incident angle range, the expected sound absorption coefficient target value, the sound barrier thickness limit value, and the environmental temperature and humidity change range, and output the optimal aperture size, hole spacing, perforation rate, and plate layer thickness combination parameters of the perforated plate through the fully connected layer of the NoiseSpecNet noise spectrum feature extraction model.

[0009] Among them, the step of performing precise perforation on the ordinary perforated UHPC plate is specifically as follows: According to the optimized parameters, place the ordinary perforated UHPC plate in a mold, and use a special drill bit array to perform precise perforation on the UHPC plate according to the set hole spacing, so that the error of the optimal aperture size is controlled within ±0.1 mm, and the perforation rate reaches the calculated optimal perforation rate.

[0010] Among them, the step of laying basalt fiber cloth on the back of the perforated UHPC plate is specifically as follows: Lay basalt fiber cloth on the back of the perforated UHPC plate, and use epoxy resin glue to bond and fix the basalt fiber cloth to the perforated UHPC plate to ensure that the two are tightly combined without gaps.

[0011] Among them, the steps of bonding the through-hole ceramic plate and the basalt fiber cloth are specifically as follows: Place the through-hole ceramic plate formed by high-temperature sintering and the basalt fiber cloth correspondingly, and bond them with waterproof polyurethane glue. The thickness of the glue layer is controlled between 1 mm and 1.5 mm to ensure that the bonding strength reaches more than 1.5 MPa.

[0012] Among them, the steps of pasting the first non-porous UHPC plate, the sound insulation felt, and the second non-porous UHPC plate on the back of the through-hole ceramic plate in sequence to form a composite structure are specifically as follows: Paste the first non-porous UHPC plate, the sound insulation felt, and the second non-porous UHPC plate on the back of the through-hole ceramic plate in sequence to form a composite structure. A staggered joint design is adopted between each layer, and the thickness of the plate layer is set according to the calculated optimal plate layer thickness combination parameters to enhance the overall structural strength.

[0013] Among them, the steps of curing the composite structure in a constant temperature and humidity curing chamber are specifically as follows: Place the composite structure in a constant temperature and humidity curing chamber, control the temperature at 20 °C to 25 °C, keep the relative humidity at 85% to 95%, and the curing time is not less than 14 days to ensure that the bonding material is fully cured.

[0014] Among them, the NoiseSpecNet noise spectrum feature extraction model specifically refers to a hybrid architecture model based on a convolutional neural network and a long short-term memory network, which includes five convolutional layers for extracting local features of the noise spectrum, two bidirectional long short-term memory network layers for capturing temporal features, and three fully connected layers for parameter prediction.

[0015] Among them, the steps of establishing the training data set of the NoiseSpecNet noise spectrum feature extraction model include collecting noise samples under different road types, different traffic flows, and different meteorological conditions, performing time-frequency domain conversion on the noise samples to obtain spectrograms, and annotating the corresponding optimal sound barrier structure parameters and measured sound absorption coefficient data for each spectrogram.

[0016] Compared with the prior art, the present invention provides a production method of a lightweight, high-strength, highly corrosion-resistant, and highly sound-absorbing road sound barrier. The present invention proposes a production method of a road sound barrier based on the NoiseSpecNet noise spectrum feature extraction model. By collecting and analyzing traffic noise data in real time, accurately extracting noise spectrum features, and automatically optimizing the sound barrier structure parameters using deep learning algorithms, an efficient sound absorption design for traffic noise in a specific frequency band is realized. This method combines materials such as perforated UHPC plates, basalt fiber cloth, and through-hole ceramic plates to form a targeted multi-layer composite sound absorption structure.

[0017] The present invention solves the problem of insufficient sound absorption performance of traditional sound barriers. By precisely controlling the aperture diameter, hole spacing, and perforation rate of the perforated panel and combining with the design of a multi-layer composite structure, efficient sound absorption is achieved for the main traffic noise frequency band from 500 Hz to 4000 Hz; the application of the NoiseSpecNet model enables the sound barrier design to shift from experience-driven to data-driven, and it can automatically adjust the structural parameters according to the noise characteristics of a specific road environment, improving the sound absorption ability and adaptability of the sound barrier to different frequency noises. Especially in the optimization of perforation parameters and the collaborative design of multi-layer structures, the present invention realizes the maximum dissipation of sound energy during the propagation process.

[0018] The present invention successfully solves the technical problem that it is difficult for road sound barriers to achieve efficient sound absorption for traffic noise in a specific frequency band, and provides a precise and intelligent solution for road noise control. Through the precise analysis of traffic noise characteristics and the adaptive optimization of the sound barrier structure, the best match between the sound barrier and the noise characteristics of a specific road environment is achieved, significantly improving the noise reduction effect of the sound barrier in practical applications, especially its adaptability under different traffic flows and environmental conditions. 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 with reference to 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 lightweight, high-strength, corrosion-resistant, and highly sound-absorbing road sound barrier provided by the present invention. This method includes the following steps:

[0022] S01. Set up multiple high-precision microphone arrays at the edge of the road, collect traffic noise samples at different times, input the collected traffic noise samples into the NoiseSpecNet noise spectrum feature extraction model, extract the traffic noise spectrum features, determine that the main noise frequency band is from 500 Hz to 4000 Hz, and calculate the influence degree of the sound wave superposition effect on human hearing at different frequencies;

[0023] S02. According to the processing results of the NoiseSpecNet noise spectrum feature extraction model, extract the traffic noise spectrum distribution data, incident angle range, expected sound absorption coefficient target value, sound barrier thickness limit value, and environmental temperature and humidity change range, and output the optimal aperture size, hole spacing, perforation rate, and plate layer thickness combination parameters of the perforated panel through the fully connected layer of the NoiseSpecNet noise spectrum feature extraction model;

[0024] S03. Place the ordinary perforated UHPC board in the mold according to the optimized parameters, and use a special drill bit array to precisely perforate the UHPC board at the set hole spacing, so that the error of the optimal hole diameter size is controlled within ±0.1 mm, and the perforation rate reaches the calculated optimal perforation rate;

[0025] S04. Lay basalt fiber cloth on the back of the perforated UHPC board, and use epoxy resin glue to bond and fix the basalt fiber cloth to the perforated UHPC board to ensure that the two are tightly combined without gaps;

[0026] S05. Place the through-hole ceramic board formed by high-temperature sintering corresponding to the basalt fiber cloth, and use waterproof polyurethane glue for bonding. The thickness of the glue layer is controlled between 1 mm and 1.5 mm to ensure that the bonding strength reaches more than 1.5 MPa;

[0027] S06. Paste the first non-perforated UHPC board, sound insulation felt, and the second non-perforated UHPC board on the back of the through-hole ceramic board in sequence to form a composite structure. A staggered joint design is adopted between layers, and the thickness of the board layer is set according to the calculated optimal board layer thickness combination parameters to enhance the overall structural strength;

[0028] S07. Put the composite structure into a constant temperature and humidity curing chamber, control the temperature at 20 °C to 25 °C, keep the relative humidity at 85% to 95%, and the curing time is not less than 14 days to ensure that the bonding material is fully cured;

[0029] S08. Conduct acoustic performance tests on the cured sound barrier unit, including tests of sound absorption coefficient, sound insulation amount, sound attenuation value in each frequency band, and mechanical performance tests, including tests of flexural strength, compressive strength, and shear strength;

[0030] S09. Install the qualified sound barrier units on the precast foundation structure, connect the units with rubber sealing strips, and conduct waterproof and dustproof treatments. Finally, spray the overall anti-ultraviolet coating to enhance the weather resistance and service life.

[0031] Among them, UHPC specifically refers to a high-strength and high-durability concrete material based on cement, incorporated with silica fume, quartz sand, mineral admixtures, and high-performance water reducers, as well as steel fibers or other reinforcing fibers. Its compressive strength usually exceeds 150 MPa, which is 3 to 5 times that of ordinary concrete.

[0032] Among them, basalt fiber cloth specifically refers to an inorganic fiber fabric made by melting and spinning basalt rock, which has excellent high-temperature resistance, acid and alkali resistance, heat insulation, and sound absorption properties. Its tensile strength is not less than 2000 MPa, and its melting point is above 1100 °C.

[0033] Among them, the through-hole ceramic plate is specifically a porous ceramic material with a porosity between 40% and 60%, having good sound absorption characteristics and fire resistance, with a fire resistance rating of A1, and a service temperature range of -40 degrees Celsius to 800 degrees Celsius.

[0034] Among them, the sound insulation felt is specifically a flexible sound insulation material made of polyester fiber or mineral fiber, with a density between 80 kg / m³ and 120 kg / m³, having good sound insulation performance and vibration damping characteristics, and can effectively block the propagation of medium and low-frequency sound waves.

[0035] Among them, the acoustic parameter optimization function is specifically a mathematical function used to calculate the structural parameters of the sound barrier. Based on the Helmholtz resonance principle and the acoustic impedance model of the multi-layer composite structure, it locates the optimal combination in the parameter space through the adaptive grid search algorithm. The input includes the traffic noise spectrum distribution data, the incident angle range, the expected sound absorption coefficient target value, the sound barrier thickness limit value, and the environmental temperature and humidity change range. The output is the optimal aperture size, hole spacing, perforation rate, and plate layer thickness combination parameters of the perforated plate.

[0036] Among them, the NoiseSpecNet noise spectrum feature extraction model is specifically a hybrid architecture model based on the convolutional neural network and the long short-term memory network. It contains five convolutional layers for extracting local features of the noise spectrum, two bidirectional long short-term memory network layers for capturing temporal features, and three fully connected layers for parameter prediction. The NoiseSpecNet noise spectrum feature extraction model uses a weighted combination of the cross-entropy loss function and the mean squared error loss function as the optimization objective, and uses the Adam optimizer for parameter update. The NoiseSpecNet noise spectrum feature extraction model is pre-trained with a large number of noise samples under different road types, different traffic flows, and different meteorological conditions, and uses synthetic data generated by acoustic simulation software to enhance the generalization ability of the model.

[0037] The steps for establishing the training dataset of the NoiseSpecNet noise spectrum feature extraction model include collecting noise samples under different road types, different traffic flows, and different meteorological conditions, performing time-frequency domain conversion on the noise samples to obtain spectrograms, annotating the corresponding optimal sound barrier structure parameters and measured sound absorption coefficient data for each spectrogram, dividing the dataset into a training set, a validation set, and a test set in a ratio of 7:2:1, using acoustic simulation software to generate synthetic data, simulating the noise propagation characteristics under various complex road environments and vehicle combination scenarios, and mixing the synthetic data with the real data to enhance the generalization ability and robustness of the NoiseSpecNet noise spectrum feature extraction model.

[0038] The steps for training the NoiseSpecNet noise spectrum feature extraction model include using a weighted combination of the cross-entropy loss function and the mean squared error loss function as the optimization objective, using the Adam optimizer for parameter update, setting the initial learning rate to 0.001 and dynamically adjusting it using the cosine annealing strategy, reducing the learning rate to 80% of the original every 50 training epochs, introducing noise augmentation technology during training, adding Gaussian white noise and random frequency masking to the input data to improve the ability of the NoiseSpecNet noise spectrum feature extraction model to cope with environmental changes, monitoring the performance metrics on the validation set during training, and terminating the training early when there is no performance improvement for 10 consecutive training epochs. Finally, the prediction error of the sound absorption coefficient of the NoiseSpecNet noise spectrum feature extraction model on different types of road noise test sets is less than 5%.

[0039] The specific implementation manners of the above steps are described in detail below. The specific implementation manner of step S01 is to select representative positions at the road edge, set up a high-precision microphone array according to the requirements specified in the acoustic survey specification GB / T4959. The microphone array uses 8 omnidirectional capacitive measurement microphones evenly distributed in a ring. The microphone sensitivity is not less than 50 mV / Pa, the frequency response range is from 20 Hz to 20000 Hz, and the signal-to-noise ratio is not less than 70 dB. The acquisition time covers different periods on weekdays and weekends, including the morning rush hour (7:00 - 9:00), the flat peak period (10:00 - 16:00), the evening rush hour (17:00 - 19:00), and the night (22:00 - 6:00). Each period is continuously acquired for not less than 60 minutes, the sampling rate is set to 48 kHz, and the quantization precision is 24 bit. The collected traffic noise samples are converted into time-frequency domain representation through fast Fourier transform (FFT). The frame length is set to 1024 points, the frame shift is 512 points, and the window function is selected as the Hanning window. Through the analysis and processing of the NoiseSpecNet noise spectrum feature extraction model, the traffic noise spectrum features are extracted, and the main noise frequency band is determined to be from 500 Hz to 4000 Hz, and the sound pressure level is concentrated between 70 dB and 85 dB. The A-weighting network is used to calculate the weighted values of sound waves at different frequencies to evaluate the degree of influence on human hearing. The results show that the noise has the most significant influence on human hearing in the frequency band from 1000 Hz to 2500 Hz. The purpose of this step is to accurately locate the traffic noise frequency bands that need to be key treated through scientific sampling and professional analysis, and provide accurate parameter basis for the subsequent sound barrier structure design.

[0040] The specific implementation of step S02 is to input the traffic noise spectrum distribution data processed by the NoiseSpecNet noise spectrum feature extraction model into the algorithm processing system to extract key acoustic parameters, including the spectral energy distribution characteristics, the direction of the main noise source and its incident angle range (usually ±60° on the horizontal plane and 0° to 45° on the vertical plane), the environmental temperature change range (-30°C to 50°C), and the humidity change range (30% to 100%). The target value of the expected sound absorption coefficient is set to be not less than 0.85 on average in the frequency band of 500 Hz to 4000 Hz and not less than 0.9 on average in the frequency band of 1000 Hz to 2500 Hz, considering the sound barrier thickness limit of 180 mm to 250 mm. Through the fully connected layer of the NoiseSpecNet noise spectrum feature extraction model, based on the Helmholtz resonator theory and the sound absorption mechanism of porous materials, the optimal perforation parameters are calculated. The calculation process uses an adaptive grid search algorithm to find the best combination in the parameter space, treating the acoustic performance and structural requirements as a multi-objective optimization problem. Finally, the optimal parameter combination of the perforated plate is output: the aperture size is 8 mm to 12 mm (the best value is 10 mm when the main frequency band is 1000 Hz to 2000 Hz), the hole pitch is 25 mm to 35 mm, the perforation rate is 25% to 30%, and the plate layer thickness configuration is: perforated UHPC plate 15 mm to 20 mm, basalt fiber cloth layer 3 mm to 5 mm, through-hole ceramic plate 30 mm to 40 mm, first non-perforated UHPC plate 20 mm to 25 mm, sound insulation felt 30 mm to 40 mm, second non-perforated UHPC plate 20 mm to 25 mm. The purpose of this step is to determine the optimal structural parameters of each component layer of the sound barrier through scientific calculation according to the actual noise characteristics and environmental conditions, providing an accurate basis for subsequent production.

[0041] The specific implementation of step S03 is to select UHPC plates that meet the requirements of the GB / T31387 standard, with a compressive strength of not less than 150 MPa, a flexural strength of not less than 30 MPa, and an elastic modulus of not less than 45 GPa. Design and customize a drill bit array according to the optimized parameters. The drill bits are made of diamond-coated cemented carbide, and the diameter accuracy of the drill bits is controlled within ±0.05 mm. Fix the UHPC plate on the fixture to ensure that the flatness error of the plate surface does not exceed 0.5 mm / m. Use a CNC drill press for precise perforation. During the drilling process, keep the drill bit speed constant at 1200 rpm to 1500 rpm, and control the feed rate at 15 mm / min to 20 mm / min. Use a water cooling system for continuous cooling to prevent overheating. The layered drilling strategy is adopted during the perforation process, and the feed depth each time does not exceed 1 / 3 of the plate thickness to reduce stress concentration and avoid edge cracking. Use precision measuring instruments to conduct quality inspection on the completed perforated plate. The aperture size error is controlled within ±0.1 mm, the hole position error is controlled within ±0.2 mm, and the perforation rate reaches the calculated optimal perforation rate of 25% to 30%. After perforation, perform edge treatment to remove burrs and microcracks to ensure the smoothness of the holes. The purpose of this step is to accurately achieve the perforation structure calculated theoretically, ensure the actual implementation of the acoustic performance design parameters, and lay a foundation for the assembly of the subsequent composite structure.

[0042] The specific implementation of step S04 is to select high-performance basalt fiber cloth with a tensile strength of not less than 2000 MPa and a density of 2.7 g / cm 3 to 2.9 g / cm 3 The fiber diameter is 9 μm to 12 μm, and the areal density is 300 g / m 2 to 400 g / m 2Cut the basalt fiber cloth into the same size as the perforated UHPC board and pre-dry it in an 80°C environment for 4 hours to reduce the moisture content to below 0.5%. Select a two-component epoxy resin adhesive with a shear strength of not less than 15 MPa, a tensile strength of not less than 25 MPa, and a temperature resistance range of -40°C to 120°C. Mix the epoxy resin main agent and the curing agent in a weight ratio of 2:1, stir evenly, and then let it stand for 10 minutes to remove air bubbles. Place the perforated UHPC board on a horizontal workbench, and use a gluing device to evenly coat the epoxy resin adhesive on the back of the board, with the gluing thickness controlled between 0.8 mm and 1.0 mm. Carefully lay the basalt fiber cloth, slowly spreading it from the center to the periphery to avoid wrinkles and air bubbles. Use a silicone roller to roll from the center to the periphery to squeeze out the excess glue and air bubbles, ensuring that the fiber cloth is tightly bonded to the UHPC board. After laying, apply a uniform pressure of 5 kPa to 10 kPa and let it stand and cure at room temperature (20°C to 25°C) for 24 hours to ensure that the bonding strength meets the design requirements. The purpose of this step is to tightly bond the basalt fiber cloth with excellent high-temperature resistance and tensile strength to the perforated UHPC board to form a composite structure and enhance the overall performance and durability of the sound barrier.

[0043] The specific implementation method of step S05 is to select a through-hole ceramic board with a porosity of 40% to 60%, and its density is 1.8 g / cm 3 to 2.2 g / cm 3 , with a compressive strength of not less than 30 MPa and the pore diameter distributed between 0.5 mm and 2 mm. Preheat the through-hole ceramic board in a 170°C environment for 2 hours to remove internal moisture, with the moisture content controlled below 0.2%. Select a one-component waterproof polyurethane adhesive with a bonding strength of not less than 1.5 MPa, an elongation rate of not less than 300% after curing, and water resistance meeting the requirements of GB / T 9753 standard. Adjust the adhesive to an appropriate viscosity (3000 MPa·s to 5000 MPa·s), and use a gluing machine to evenly coat the polyurethane adhesive on the surface of the UHPC board with the basalt fiber cloth already fixed, with the gluing thickness strictly controlled between 1 mm and 1.5 mm. Carefully place the through-hole ceramic board so that the board edge is precisely aligned with the UHPC board, and use a guiding and positioning device during the placement process to ensure the position accuracy. Use a pneumatic pressure device to apply a uniform pressure (8 kPa to 12 kPa) to the entire structure to squeeze out the excess glue and air bubbles. Keep it under pressure for 24 hours, with the ambient temperature controlled at 20°C to 25°C and the relative humidity controlled at 40% to 60% to ensure that the adhesive is fully cured. After curing, conduct a pull-out test, and the bonding strength must reach above 1.5 MPa. The purpose of this step is to firmly bond the through-hole ceramic board with excellent sound absorption characteristics to the composite structure formed in the previous process and further enhance the sound absorption performance of the sound barrier.

[0044] The specific implementation of step S06 is to prepare two non-porous UHPC plates with thicknesses of 20 mm to 25 mm and dimensions slightly larger than the previous combined structure by 10 mm to 15 mm to form a staggered joint design and enhance the overall rigidity of the structure. Select polyester fiber sound insulation felt with a density of 100 kg / m 3 to 120 kg / m 3 and a thickness of 30 mm to 40 mm, and its sound insulation amount is not less than 25 dB in the frequency band of 500 Hz to 4000 Hz. Use acrylate structural adhesive to paste the first non-porous UHPC plate on the back of the through-hole ceramic plate, and control the thickness of the adhesive layer within 0.8 mm to 1.2 mm, and the bonding strength is not less than 2.0 MPa. Lay the sound insulation felt on the back of the first non-porous UHPC plate, and arrange it in a staggered joint to avoid the formation of sound bridges. Use the same structural adhesive to paste the second non-porous UHPC plate on the back of the sound insulation felt to form a six-layer composite structure of "perforated UHPC plate - basalt fiber cloth - through-hole ceramic plate - first non-porous UHPC plate - sound insulation felt - second non-porous UHPC plate". The entire structure is treated with edge sealing, and polysulfide sealant is used to seal the edges of the structure to prevent moisture from seeping in. The thickness of each layer is strictly set according to the optimal plate layer thickness combination parameters calculated in step S02, and the total thickness is controlled between 180 mm and 250 mm. The purpose of this step is to achieve the gradual attenuation and blocking of sound waves through the multi-layer composite structure design, improve the overall sound insulation performance of the sound barrier, and at the same time enhance the structural strength and wind pressure resistance through the staggered joint design.

[0045] The specific implementation of step S07 is to place the completed composite structure in a constant temperature and humidity curing chamber that meets the requirements of GB / T50081 standard. The temperature control system adopts a PID control algorithm to accurately maintain the temperature at 22 ± 3 °C (within the range of 20 °C to 25 °C), and the temperature fluctuation range does not exceed ±0.5 °C. The humidity control system uses an ultrasonic humidification device and a dehumidification device to work together to keep the relative humidity at 90 ± 5% (within the range of 85% to 95%), and the humidity fluctuation range does not exceed ±2%. The curing time is set to be not less than 14 days. The high humidity (90% to 95%) is maintained in the first 7 days, and gradually reduced to 85% to 90% in the next 7 days to simulate the gradual adaptation process in the natural environment. During the curing process, inspections are carried out every 24 hours, and the temperature and humidity data are recorded to ensure the stability of the curing conditions. During the curing period, the sound barrier units are placed with spaced supports to ensure that each surface is evenly in contact with the moist air. After curing, the sound barrier units are slowly dried in the natural environment for 7 days to fully release the internal stress and avoid cracking in the later stage. The purpose of this step is to ensure the full curing of the bonding material through strictly controlled curing conditions, form a firm bonding interface between the layers of the composite structure, and improve the overall durability and structural stability of the sound barrier.

[0046] The specific implementation of step S08 is to establish an acoustic test environment in accordance with the GB / T20247 standard, including an anechoic chamber and a reverberation chamber. The impedance tube method is used to test the sound absorption coefficient of the sound barrier unit, and the test frequency range is from 100 Hz to 5000 Hz, with a focus on the frequency band from 500 Hz to 4000 Hz. It is required that the sound absorption coefficient in the main frequency band (1000 Hz to 2500 Hz) is not less than 0.9, and the average sound absorption coefficient in the full frequency band from 500 Hz to 4000 Hz is not less than 0.85. The transmission method is used to test the sound insulation quantity. According to the GB / T19889 standard, the average sound insulation quantity of the sound barrier unit in the frequency band from 500 Hz to 4000 Hz is not less than 30 dB, and in the frequency band from 1000 Hz to 2000 Hz is not less than 35 dB. The sound attenuation value of the product is tested. According to the ISO10847 standard, a road noise source is simulated in the semi-anechoic chamber, and the sound pressure level difference before and after the sound barrier is measured. It is required that the A-weighted sound attenuation value is not less than 20 dB. The mechanical property tests include: the flexural strength test is carried out in accordance with the GB / T50081 standard, and the requirement is not less than 35 MPa; the compressive strength test is carried out in accordance with the GB / T50081 standard, and the requirement is not less than 150 MPa; the shear strength test is carried out in accordance with the GB / T228 standard, and the requirement is not less than 15 MPa. In addition, the weather resistance test is carried out, including ultraviolet light aging test (1000 hours), freeze-thaw cycle test (50 cycles) and salt spray test (1000 hours). After the test, the attenuation of each performance index does not exceed 10%. The purpose of this step is to verify whether the sound barrier unit meets the design requirements through comprehensive performance tests, and ensure the acoustic performance and structural safety of the product.

[0047] The specific implementation method of step S09 is to design the sound barrier installation foundation according to the actual situation of the road. The foundation adopts a reinforced concrete structure, the concrete strength grade is not less than C30, and the foundation depth is not less than 800mm. M20 bolts are embedded in the foundation, and the spacing is determined according to the size of the sound barrier unit, usually 1000mm to 1500mm. The bottom of the sound barrier unit is designed with an installation groove, and a rubber vibration damping pad is pre-installed in the groove, with a thickness of 10mm to 15mm and a hardness of 60±5 Shore A. The sound barrier unit is hoisted in place and fixed with the embedded bolts through the flange plate, and the tightening torque is controlled at 80N·m to 100N·m. EPDM rubber sealing strips are used to connect the units, the hardness of the sealing strips is 65±5 Shore A, the tensile strength is not less than 7MPa, and the compression permanent deformation rate does not exceed 30%. Use sealant to waterproof and dustproof the unit connection. The sealant uses neutral silicone sealant, and its tensile strength is not less than 0.8MPa and the elongation is not less than 300%. Finally, the overall anti-ultraviolet coating is sprayed. The coating material is acrylic modified polyurethane coating, the coating thickness is 60μm to 80μm, the ultraviolet reflectivity is not less than 85%, and the service life is not less than 15 years. After the installation is completed, the overall performance test is carried out, including air tightness test, water tightness test and wind pressure resistance test, to ensure that the overall structure meets the design requirements. The purpose of this step is to scientifically install qualified sound barrier units on the edge of the road to form a complete sound barrier system to achieve effective control of traffic noise. At the same time, through strict installation technology and protective treatment, the long-term stable operation of the sound barrier system is ensured.

[0048] The specific implementation of the structure of the NoiseSpecNet noise spectrum feature extraction model is to construct a hybrid architecture model based on deep learning, which includes five convolutional layers, two bidirectional long short-term memory network layers, and three fully connected layers. The convolutional layers use 2D convolution operations. The first layer uses 32 3×3 convolutional kernels with a stride of 1×1 and the ReLU activation function; the second layer uses 64 3×3 convolutional kernels with a stride of 2×2; the third layer uses 128 3×3 convolutional kernels with a stride of 1×1; the fourth layer uses 128 3×3 convolutional kernels with a stride of 2×2; the fifth layer uses 256 3×3 convolutional kernels with a stride of 1×1. After each layer of convolution, a batch normalization layer and the ReLU activation function are connected, and a max pooling layer with a kernel size of 2×2 is added after the second and fourth layers. The bidirectional long short-term memory network layers each contain 128 hidden units, which are used to capture the temporal features of the noise spectrum. The dropout technique is applied after the long short-term memory network layer with a probability set to 0.3 to prevent overfitting. The fully connected layers contain 512, 256, and the final output units, corresponding to the number of prediction parameters respectively. The last layer uses a linear activation function to output the predicted values of the acoustic parameters. The total number of parameters of the entire model is approximately 5.2 million. When training, the batch size is 64 and the number of training epochs is 200. This model can effectively extract the time-frequency features of traffic noise and accurately predict the optimal sound barrier structure parameters based on these features.

[0049] The specific implementation of the establishment of the training dataset for the NoiseSpecNet noise spectrum feature extraction model is to first set up microphone arrays on different levels of roads (expressways, urban arterial roads, secondary arterial roads, and branch roads), and collect noise samples during the peak traffic flow period (>2000 vehicles / hour), the off-peak period (1000 - 2000 vehicles / hour), and the low-traffic period (<1000 vehicles / hour) respectively. The samples are also collected under different meteorological conditions (sunny days, rainy days, windy days) to ensure the diversity of the dataset. The collected samples are transformed into spectrograms through the short-time Fourier transform with a time resolution of 10 ms and a frequency resolution of 10 Hz. The spectrograms are labeled with metadata such as traffic type, traffic flow, meteorological conditions, and vehicle composition ratio. According to the test results of the acoustic laboratory, the corresponding optimal sound barrier structure parameters and measured sound absorption coefficient data are labeled for each spectrogram. The total amount of the dataset reaches 50,000 samples, which are divided into a training set (35,000 samples), a validation set (10,000 samples), and a test set (5,000 samples) according to the ratio of 7:2:1. 10,000 samples of synthetic data are generated using acoustic finite element analysis software to simulate various complex road environment and vehicle combination scenarios, enhancing the generalization ability of the model. Finally, the prediction error of the sound absorption coefficient of the model on the test sets of different types of road noises is controlled within 5%, and the prediction accuracy of the structure parameters reaches more than 92%.

[0050] The mathematical models or calculation processes involved in the present invention will be described in detail below.

[0051] In steps S01 and S02, there are multiple calculation processes, which involve traffic noise spectrum analysis and sound barrier parameter optimization, and are specifically expressed as follows:

[0052] In step S01, the fast Fourier transform (FFT) is used to convert the time-domain noise signal into a frequency-domain representation, and its mathematical expression is:

[0053]

[0054] where X(k) is the frequency-domain signal; x(n) is the time-domain signal; N is the number of sampling points, set to 1024; k is the frequency index, with a range of 0 ≤ k ≤ N - 1; j is the imaginary unit, j 2 = -1; e -j2πkn / N is the Fourier transform kernel function.

[0055] The time-frequency spectrum generation uses the short-time Fourier transform (STFT), and its mathematical expression is:

[0056]

[0057] where STFT{x(n)}(m, k) is the time-frequency spectrum; x(n) is the time-domain signal; w(n - m) is the window function, using the Hanning window; m is the time index; k is the frequency index; N is the frame length, set to 1024.

[0058] The mathematical expression of the Hanning window function is:

[0059] w(n) = 0.5 - 0.5cos(2πn / (N - 1));

[0060] where w(n) is the window function value; n is the sample point index, with a range of 0 ≤ n ≤ N - 1; N is the window function length, set to 1024.

[0061] The frequency response function of the A-weighting network is:

[0062]

[0063] where A(f) is the A-weighting value; f is the frequency, in Hz.

[0064] The calculation formula for the weighted sound pressure level is:

[0065]

[0066] where L A is the A-weighted sound pressure level, in dB(A); L iis the unweighted sound pressure level of the i-th frequency band, in dB; A(f i ) is the A-weighting value at the center frequency of the i-th frequency band; n is the number of frequency bands.

[0067] In step S02, the resonance frequency calculation formula of the Helmholtz resonator theory is:

[0068]

[0069] In the formula, f0 is the resonance frequency, in Hz; c is the speed of sound, approximately 343 m / s (under the condition of 20 °C); S is the orifice area, in m 2 ; V is the cavity volume, in m 3 ; L is the orifice length (equivalent to the plate thickness), in m.

[0070] The sound absorption coefficient calculation formula of the porous material is based on the Delany-Bazley model:

[0071]

[0072] In the formula, α is the sound absorption coefficient; Z s is the surface acoustic impedance of the material, in Pa·s / m; ρ0 is the air density, approximately 1.21 kg / m 3 (under the condition of 20 °C); c0 is the speed of sound in air, approximately 343 m / s (under the condition of 20 °C).

[0073] Among them, the surface acoustic impedance Z s The calculation formula is:

[0074] Z s = ρ0c0coth(jkd(1 + jG));

[0075] In the formula, Z s is the surface acoustic impedance; j is the imaginary unit; k is the wave number, k = 2πf / c0; f is the frequency, in Hz; d is the material thickness, in m; G is the flow resistance parameter, related to the porosity and flow resistance of the material.

[0076] For the perforated plate structure, the following model is used to calculate its surface acoustic impedance:

[0077]

[0078] In the formula, Z perf is the surface acoustic impedance of the perforated plate; k is the wave number; d eff is the effective thickness, d eff = d + δ; d is the plate thickness; δ is the end correction, r is the hole radius; σ is the perforation rate; R s is the acoustic resistance, related to the frequency, aperture and perforation rate.

[0079] The total surface acoustic impedance of the composite structure is calculated using the transfer matrix method:

[0080]

[0081] where p1 and v1 are the sound pressure and particle velocity on the incident surface; p2 and v2 are the sound pressure and particle velocity on the transmission surface; k is the wave number; d is the layer thickness; Z c is the characteristic impedance.

[0082] The total transfer matrix of the multi-layer structure is the product of the transfer matrices of each layer:

[0083] T total = T1T2T3T4T5T6;

[0084] where T total is the total transfer matrix; T1 to T6 are the transfer matrices of the six-layer structure respectively.

[0085] Finally, the calculation formula for the total surface acoustic impedance is:

[0086]

[0087] where Z total is the total surface acoustic impedance; T 11 , T 12 , T 21 , T 22 are the elements of the total transfer matrix; Z0 is the impedance of the back cavity. For a rigid backplate, Z0 approaches infinity.

[0088] The perforation parameter optimization uses an adaptive grid search algorithm, and its objective function is:

[0089] F(x) = w1α avg + w2α peak - w3C(x) - w4M(x);

[0090] where F(x) is the objective function value; x is the parameter vector, including aperture, hole spacing, perforation rate, layer thickness, etc.; α avg is the average sound absorption coefficient in the frequency band from 500 Hz to 4000 Hz; α peak is the average sound absorption coefficient in the frequency band from 1000 Hz to 2500 Hz; C(x) is the cost penalty term; M(x) is the material weight penalty term; w1, w2, w3, w4 are the weight coefficients, which are 0.4, 0.4, 0.1, 0.1 respectively.

[0091] The loss function of the NoiseSpecNet noise spectrum feature extraction model is:

[0092] L total= λ1L CE + λ2L MSE ;

[0093] Wherein, L total is the total loss; L CE is the cross - entropy loss, used for classification tasks; L MSE is the mean squared error loss, used for regression tasks; λ1, λ2 are weight coefficients, being 0.6 and 0.4 respectively.

[0094] The cross - entropy loss function is defined as:

[0095]

[0096] Wherein, L CE is the cross - entropy loss; y i is the true label; is the predicted probability; C is the number of classes.

[0097] The mean squared error loss function is defined as:

[0098]

[0099] Wherein, L MSE is the mean squared error loss; y i is the true value; is the predicted value; n is the number of samples.

[0100] The learning rate is adjusted using the cosine annealing strategy:

[0101]

[0102] Wherein, η t is the current learning rate; η min is the minimum learning rate, set to 0.0001; η max is the maximum learning rate, set to 0.001; t is the current training epoch; T is the total number of training epochs.

[0103] The construction principles and meanings of these formulas are as follows:

[0104] The Fast Fourier Transform (FFT) and Short - Time Fourier Transform (STFT) formulas are based on Fourier analysis theory. By decomposing the time - domain signal into the superposition of sine waves with different frequencies, they can accurately reflect the distribution characteristics of traffic noise in the frequency domain. The Hanning window function is selected because it has a good balance in spectrum leakage suppression and frequency resolution, and is suitable for analyzing traffic noise with continuous spectrum characteristics.

[0105] The A-weighting network formula simulates the sensitivity differences of the human ear to sounds of different frequencies, making the measurement results closer to the human auditory perception. Among them, the fourth-power term of the frequency reflects the high sensitivity of the human ear in the mid- and high-frequency bands, while the terms in the denominator reflect the reduced sensitivity in the low-frequency and ultra-high-frequency bands.

[0106] The theoretical formula of the Helmholtz resonator describes the acoustic resonance characteristics of the perforated plate structure. By matching the resonance frequency with the target noise frequency band, the sound absorption effect can be significantly improved. The square root relationship is adopted in the formula because the resonance frequency is proportional to the square root of the ratio of the orifice area to the cavity volume, and this relationship stems from the natural vibration characteristics of the mass-spring system.

[0107] The Delany-Bazley model is based on the flow resistance characteristics of the material, and calculates the sound absorption coefficient through the complex acoustic impedance. The surface acoustic impedance formula uses the hyperbolic tangent function to describe the propagation behavior of sound waves in the sound-absorbing material. Among them, the imaginary term reflects the phase change, while the real term is related to the energy dissipation.

[0108] The transfer matrix method formula simplifies the complex multi-layer structure into matrix multiplication operations. The acoustic characteristics of each layer of material are represented by a 2×2 matrix, and finally the acoustic response of the entire system is obtained. This method takes into account the interaction between layers and reflects the acoustic performance of the composite structure more accurately than simple superposition.

[0109] The design of the parameter optimization objective function considers the balance between sound absorption performance and practicality, and adjusts the importance of different objectives through the weight coefficient. Among them, special attention is given to the frequency band from 1000 Hz to 2500 Hz because this frequency band has the greatest impact on the human ear and is also the main energy range of traffic noise.

[0110] The loss function of the NoiseSpecNet model combines a classification task (identifying the noise type) and a regression task (predicting the structural parameters). The cross-entropy loss is applicable to the classification task, while the mean squared error loss is applicable to the regression task. The combination of the two according to the weight can optimize these two types of tasks simultaneously.

[0111] The learning rate adjustment adopts the cosine annealing strategy. By periodically reducing the learning rate, it avoids the oscillation during the training process, and at the same time maintains a certain exploration ability, which helps to find a better local minimum. The cosine function provides a smooth transition from a high learning rate to a low learning rate, which is more conducive to the convergence of the model than stepwise decay.

[0112] Optionally, in step S02, the calculation formula for the perforation rate is:

[0113]

[0114] In the formula, σ is the perforation rate; d is the hole diameter, in mm; b is the hole spacing, in mm.

[0115] In the NoiseSpecNet noise spectrum feature extraction model, the calculation formula of the convolutional layer is as follows:

[0116]

[0117] In the formula, F(i, j) is the value at the (i, j) position of the output feature map; K(m, n) is the weight at the (m, n) position of the convolutional kernel; I(i + m, j + n) is the value at the corresponding position of the input feature map; M and N are the height and width of the convolutional kernel, both of which are 3.

[0118] Optionally, the calculation formula of the batch normalization layer is as follows:

[0119]

[0120] In the formula, y is the output after normalization; x is the input; E[x] is the mean within the batch; Var[x] is the variance within the batch; ∈ is a small constant to prevent division by zero, set to 10 -5 ; γ and β are learnable scaling and offset parameters.

[0121] Optionally, the calculation formula of the impedance tube method used in the actual sound absorption coefficient test is as follows:

[0122]

[0123] In the formula, α is the sound absorption coefficient; H 12 is the transfer function between two microphone positions; H i is the transfer function of the incident wave; H r is the transfer function of the reflected wave.

[0124] Specifically, the principle of the present invention is: The technical principle of the present invention is based on the integrated innovation of acoustic theory, the sound absorption mechanism of porous materials, and deep learning technology. First, the present invention adopts the Helmholtz resonance principle for the sound barrier perforated structure design. The Helmholtz resonator has a selective absorption effect on sound waves of specific frequencies, and its resonance frequency is determined by the cavity volume and the neck area. By precisely controlling the aperture, hole spacing, and perforation rate of the perforated UHPC board, a large number of micro Helmholtz resonator arrays are formed, which can achieve frequency-selective sound absorption for the main traffic noise frequency band from 500 Hz to 4000 Hz.

[0125] Secondly, the present invention adopts a multi-layer composite structure design, and utilizes the acoustic impedance matching principle and the sound absorption mechanism of porous materials to improve the sound absorption efficiency. During the propagation of sound waves, when encountering the interface of different materials, part of the energy will be reflected, and part of the energy will be transmitted and converted into heat energy in the porous materials. The perforated UHPC board provides primary sound wave guidance and resonance cavities; the basalt fiber cloth layer provides a microscopic porous structure, and absorbs high-frequency sound energy through friction and heat conversion mechanisms; the through-hole ceramic board provides medium-frequency sound absorption through its complex pore structure; the sound insulation felt mainly provides damping and absorption for low-frequency sound waves. This gradient multi-layer design enables the sound barrier to achieve efficient absorption of broadband noise.

[0126] Most innovatively, the NoiseSpecNet noise spectrum feature extraction model is introduced. Based on a hybrid architecture of convolutional neural network and long short-term memory network, this model can accurately extract time-frequency domain features from actual traffic noise samples and automatically optimize the sound barrier structure parameters. The convolutional layer is responsible for extracting local features and spatial correlations of the noise spectrum; the long short-term memory network layer captures the temporal variation features of the noise; the fully connected layer predicts the optimal structure parameters according to the extracted features. Through the hybrid training of large-scale real road noise data and acoustic simulation data, the model establishes a mapping relationship between noise features and optimal sound barrier structure parameters, realizing data-driven and intelligent design processes. This method breaks through the limitations of traditional empirical design, can automatically generate optimal sound absorption structure parameters according to the noise characteristics of specific road environments, and thus achieve efficient and accurate noise absorption.

[0127] A specific embodiment 1 of the present invention is provided below. The specific implementation manners of each step in this embodiment 1 are described in detail as follows.

[0128] The specific implementation manner of step S01 is to select a representative location at the edge of the road, and set up a high-precision microphone array according to the requirements specified in the acoustic survey specification GB / T4959. The microphone array uses 8 omnidirectional capacitive measurement microphones evenly distributed in a ring. The microphone sensitivity is not less than 50 mV / Pa, the frequency response range is from 20 Hz to 20000 Hz, and the signal-to-noise ratio is not less than 70 dB. The acquisition time covers different periods on weekdays and weekends, including the morning rush hour (7:00 - 9:00), the flat peak period (10:00 - 16:00), the evening rush hour (17:00 - 19:00) and the night (22:00 - 6:00). Each period is continuously acquired for not less than 60 minutes, the sampling rate is set to 48 kHz, and the quantization accuracy is 24 bit. The collected traffic noise samples are converted into time-frequency domain representations through fast Fourier transform (FFT), and its mathematical expression is:

[0129]

[0130] Wherein, X(k) is the frequency-domain signal; x(n) is the time-domain signal; N is the number of sampling points, set to 1024; k is the frequency index, with the range 0 ≤ k ≤ N - 1; j is the imaginary unit, j 2 = -1; e -j2πkn / N is the Fourier transform kernel function. The frame length is set to 1024 points, the frame shift is 512 points, and the window function selected is the Hanning window:

[0131] w(n) = 0.5 - 0.5cos(2πn / (N - 1));

[0132] Wherein, w(n) is the window function value; n is the sample point index, with the range 0 ≤ n ≤ N - 1; N is the window function length, set to 1024. The time-frequency spectrogram is generated using the short-time Fourier transform (STFT):

[0133]

[0134] Wherein, STFT{x(n)}(m, k) is the time-frequency spectrogram; x(n) is the time-domain signal; w(n - m) is the window function; m is the time index; k is the frequency index; N is the frame length. Through the analysis and processing of the NoiseSpecNet noise spectrum feature extraction model, the traffic noise spectrum features are extracted, and it is determined that the main noise frequency band is from 500 Hz to 4000 Hz, and the sound pressure level is concentrated between 70 dB and 85 dB. The A-weighting network is used to perform weighted calculation on the sound waves at different frequencies:

[0135]

[0136] Wherein, A(f) is the A-weighting value; f is the frequency, with the unit of Hz. The calculation formula for the weighted sound pressure level is:

[0137]

[0138] Wherein, L A is the A-weighted sound pressure level, with the unit of dB(A); L i is the unweighted sound pressure level of the i-th frequency band, with the unit of dB; A(f i ) is the A-weighting value of the center frequency of the i-th frequency band; n is the number of frequency bands. The degree of influence of the sound wave on the human hearing is evaluated, and the results show that the noise has the most significant influence on the human hearing in the frequency band from 1000 Hz to 2500 Hz. The purpose of this step is to accurately locate the traffic noise frequency bands that need to be key treated through scientific sampling and professional analysis, and provide accurate parameter basis for the subsequent sound barrier structure design.

[0139] The specific implementation of step S02 is to input the traffic noise spectrum distribution data processed by the NoiseSpecNet noise spectrum feature extraction model into the algorithm processing system to extract key acoustic parameters, including the spectral energy distribution characteristics, the direction of the main noise source and its incident angle range (usually ±60° on the horizontal plane and 0° to 45° on the vertical plane), the environmental temperature change range (-30°C to 50°C), and the humidity change range (30% to 100%). Set the target value of the expected sound absorption coefficient to be not less than 0.85 on average in the frequency band of 500 Hz to 4000 Hz and not less than 0.9 on average in the frequency band of 1000 Hz to 2500 Hz, considering the sound barrier thickness limit of 180 mm to 250 mm. Through the fully connected layer of the NoiseSpecNet noise spectrum feature extraction model, calculate the resonance frequency based on the Helmholtz resonator theory:

[0140]

[0141] In the formula, f0 is the resonance frequency, with the unit of Hz; c is the speed of sound, approximately 343 m / s (under the condition of 20°C); S is the orifice area, with the unit of m 2 ; V is the cavity volume, with the unit of m 3 ; L is the orifice length (equivalent to the plate thickness), with the unit of m. The sound absorption coefficient calculation of porous materials is based on the Delany-Bazley model:

[0142]

[0143] In the formula, α is the sound absorption coefficient; Z s is the acoustic impedance of the material surface, with the unit of Pa·s / m; ρ0 is the air density, approximately 1.21 kg / m 3 (under the condition of 20°C); c0 is the speed of sound in air, approximately 343 m / s (under the condition of 20°C). The formula for the surface acoustic impedance is:

[0144] Z s =ρ0c0coth(jkd(1 + jG));

[0145] In the formula, Z s is the surface acoustic impedance; j is the imaginary unit; k is the wave number, k = 2πf / c0; f is the frequency, with the unit of Hz; d is the material thickness, with the unit of m; G is the flow resistance parameter, related to the porosity and flow resistance of the material. The surface acoustic impedance calculation of the perforated plate structure is as follows:

[0146]

[0147] In the formula, Z perf is the surface acoustic impedance of the perforated plate; k is the wave number; d eff is the effective thickness, d eff= d + δ; where d is the plate thickness; δ is the end correction amount, r is the hole radius; σ is the perforation rate, and the calculation formula is:

[0148]

[0149] In the formula, σ is the perforation rate; d is the hole diameter, in mm; b is the hole pitch, in mm; R s is the acoustic resistance, which is related to the frequency, hole diameter, and perforation rate. The total surface acoustic impedance of the composite structure is calculated using the transfer matrix method:

[0150]

[0151] In the formula, p1 and v1 are the sound pressure and particle velocity at the incident surface; p2 and v2 are the sound pressure and particle velocity at the transmission surface; k is the wave number; d is the layer thickness; Z c is the characteristic impedance. The total transfer matrix of the multi-layer structure is the product of the transfer matrices of each layer:

[0152] T total = T1t2T3T4T5T6;

[0153] In the formula, T total is the total transfer matrix; T1 to T6 are the transfer matrices of the six-layer structure respectively. The calculation formula for the total surface acoustic impedance is:

[0154]

[0155] In the formula, Z total is the total surface acoustic impedance; T 11 , T 12 , T 21 , T 22 are the elements of the total transfer matrix; Z0 is the back cavity impedance. For a rigid back plate, Z0 approaches infinity. The perforation parameter optimization uses an adaptive grid search algorithm, and its objective function is:

[0156] F(x) = w1α avg + w2α peak - w3C(x) - w4M(x);

[0157] In the formula, F(x) is the objective function value; x is the parameter vector, including the hole diameter, hole pitch, perforation rate, layer thickness, etc.; α avg is the average sound absorption coefficient in the frequency band from 500 Hz to 4000 Hz; α peakis the average sound absorption coefficient in the frequency band of 1000 Hz to 2500 Hz; C(x) is the cost penalty term; M(x) is the material weight penalty term; w1, w2, w3, w4 are weight coefficients, which are 0.4, 0.4, 0.1, and 0.1 respectively. The optimal parameter combination of the perforated plate is finally output: the aperture size is 8 mm to 12 mm (the best value is 10 mm when the main frequency band is 1000 Hz to 2000 Hz), the hole spacing is 25 mm to 35 mm, the perforation rate is 25% to 30%, and the plate layer thickness configuration is: 15 mm to 20 mm for the perforated UHPC plate, 3 mm to 5 mm for the basalt fiber cloth layer, 30 mm to 40 mm for the through-hole ceramic plate, 20 mm to 25 mm for the first non-perforated UHPC plate, 30 mm to 40 mm for the sound insulation felt, and 20 mm to 25 mm for the second non-perforated UHPC plate. The purpose of this step is to determine the optimal structural parameters of each component layer of the sound barrier through scientific calculation according to the actual noise characteristics and environmental conditions, providing an accurate basis for subsequent production.

[0158] The specific implementation manners of steps S03 - S09 are the same as those described above and will not be elaborated here.

[0159] The specific implementation manner of the structure of the NoiseSpecNet noise spectrum feature extraction model is to construct a hybrid architecture model based on deep learning, including five convolutional layers, two bidirectional long short-term memory network layers, and three fully connected layers. The calculation formula of the convolutional layer is:

[0160]

[0161] In the formula, F(i, j) is the value at the (i, j) position of the output feature map; K(m, n) is the weight at the (m, n) position of the convolutional kernel; I(i + m, j + n) is the value at the corresponding position of the input feature map; M and N are the height and width of the convolutional kernel, both of which are 3. The first layer uses 32 3×3 convolutional kernels, the stride is 1×1, and the ReLU activation function is adopted; the second layer uses 64 3×3 convolutional kernels, the stride is 2×2; the third layer uses 128 3×3 convolutional kernels, the stride is 1×1; the fourth layer uses 128 3×3 convolutional kernels, the stride is 2×2; the fifth layer uses 256 3×3 convolutional kernels, the stride is 1×1. After each layer of convolution, a batch normalization layer is connected:

[0162]

[0163] In the formula, y is the normalized output; x is the input; E[x] is the mean within the batch; Var[x] is the variance within the batch; ∈ is a small constant to prevent division by zero, set to 10 -5 ; γ, β are learnable scaling and offset parameters. Each of the bidirectional long short-term memory network layers contains 128 hidden units, and the calculation formula is:

[0164] ft = σ(W f · [h t-1 , x t + b f );

[0165] i t = σ(W i · [h t-1 , x t + b i );

[0166]

[0167] o t = σ(W o · [h t-1 , x t + b o );

[0168] h t = o t × tanh(C t );

[0169] Where, f t is the forget gate; i t is the input gate; is the candidate memory cell; C t is the memory cell; o t is the output gate; h t is the hidden state; x t is the input at the current time step; W f , W i , W C , W o are weight matrices; b f , b i , b C , b o are bias vectors; σ is the sigmoid activation function; tanh is the hyperbolic tangent activation function. The fully connected layer contains 512, 256, and the final output units, corresponding to the number of prediction parameters respectively. The last layer uses a linear activation function to output the predicted values of acoustic parameters. The model is trained using a hybrid loss function:

[0170] L total = λ1L CE + λ2L MSE ;

[0171] Where, L total is the total loss; L CE is the cross-entropy loss, L MSE is the mean squared error loss, λ1 and λ2 are weight coefficients, which are 0.6 and 0.4 respectively. The learning rate adjustment adopts the cosine annealing strategy:

[0172]

[0173] In the formula, η t is the current learning rate; η min is the minimum learning rate, set to 0.0001; η max is the maximum learning rate, set to 0.001; t is the current training epoch; T is the total number of training epochs.

[0174] The specific implementation of establishing the training dataset for the NoiseSpecNet noise spectrum feature extraction model is to first set up microphone arrays on different levels of roads (expressways, urban arterials, secondary arterials, and branch roads), and collect noise samples during peak traffic hours (>2000 vehicles / hour), off-peak hours (1000 - 2000 vehicles / hour), and low-traffic hours (<1000 vehicles / hour) respectively. Collection is also carried out under different meteorological conditions (sunny days, rainy days, windy days) to ensure the diversity of the dataset. The collected samples are transformed into spectrograms through short-time Fourier transform, with a time resolution of 10 ms and a frequency resolution of 10 Hz. The spectrograms are labeled with metadata such as traffic type, flow, meteorological conditions, and vehicle composition ratio. According to the test results of the acoustic laboratory, the corresponding optimal sound barrier structure parameters and measured sound absorption coefficient data are labeled for each spectrogram. The total amount of the dataset reaches 50000 samples, which are divided into a training set (35000 samples), a validation set (10000 samples), and a test set (5000 samples) according to the ratio of 7:2:1. 10000 samples of synthetic data are generated using acoustic finite element analysis software to simulate various complex road environments and vehicle combination scenarios, enhancing the generalization ability of the model.

[0175] The specific implementation of training the NoiseSpecNet noise spectrum feature extraction model is to use a weighted combination of the cross-entropy loss function and the mean squared error loss function as the optimization objective:

[0176] L total = λ1L CE + λ2L MSE ;

[0177] In the formula, L total is the total loss function; λ1 is the cross-entropy loss weight, set to 0.6; λ2 is the mean squared error loss weight, set to 0.4; L CE is the cross-entropy loss, used for noise type classification; L MSE is the mean squared error loss, used for structure parameter prediction. The Adam optimizer is used for parameter update, with the initial learning rate set to 0.001 and dynamically adjusted using the cosine annealing strategy:

[0178]

[0179] where η t is the learning rate for the t-th training cycle; η min is the minimum learning rate, set to 0.0001; η max is the maximum learning rate, set to 0.001; t is the index of the current training cycle; T is the total number of training cycles, set to 200. The learning rate is reduced to 80% of the original every 50 training cycles. During the training process, a data augmentation technique is introduced, and Gaussian white noise is added to the input data:

[0180]

[0181] where x′ is the augmented input data; x is the original input data; is a Gaussian distribution with a mean of 0 and a variance of σ 2 , and the value range of σ is from 0.01 to 0.05. At the same time, random frequency masking is applied to randomly occlude 10% to 20% of the frequency intervals on the spectrogram to improve the model's robustness to frequency loss. During the training process, the batch size is set to 64, and the validation set is used to monitor the model performance, and the validation loss is calculated:

[0182]

[0183] where L val is the validation set loss; m is the number of samples in the validation set; L total (x i , y i ) is the total loss of the i-th validation sample. When the validation set loss does not decrease significantly (the improvement amplitude is less than 0.1%) for 10 consecutive training cycles, the training is terminated early to prevent overfitting. The prediction error of the sound absorption coefficient of the final model on the test sets of different types of road noises is controlled within 5%, and the prediction accuracy of the structural parameters reaches more than 92%.

[0184] The production method of the road sound barrier of the present invention realizes the accurate analysis and targeted treatment of traffic noise through the organic combination of deep learning technology and traditional acoustic theory. In step S01, the fast Fourier transform and the short-time Fourier transform are used to perform time-frequency analysis on the noise signal, and the A-weighting network is combined to evaluate its impact on human hearing; in step S02, based on the Helmholtz resonator theory, the Delany-Bazley model and the transfer matrix method, a multi-layer composite structure acoustic model is established, and the structure parameters of the sound barrier are optimized through an adaptive grid search algorithm; steps S03 to S06 convert the optimal parameters calculated theoretically into actual products through precise processing and assembly processes; the strict maintenance in step S07 ensures the release of internal stress and firm interface bonding of the composite structure; the comprehensive performance test in step S08 verifies the acoustic performance and mechanical performance of the product; the scientific installation process in step S09 ensures the long-term stable operation of the sound barrier system. The NoiseSpecNet noise spectrum feature extraction model adopts a hybrid architecture of a convolutional neural network and a long short-term memory network, which can effectively extract the time-frequency features of noise. Through the training of a large-scale diverse dataset, the generalization ability and prediction accuracy of the model are improved. The road sound barrier of the present invention has an absorption coefficient of not less than 0.9 and a sound insulation amount of not less than 35 dB in the main frequency band (1000 Hz to 2500 Hz), and at the same time has excellent characteristics such as light weight, high strength, and corrosion resistance, providing an efficient solution for urban traffic noise control.

[0185] To better understand and implement the present invention, an embodiment 2 of a specific application scenario of the present invention is provided below: In a highway expansion project, a high-performance sound barrier needs to be installed near a section adjacent to a high-end residential community to reduce the impact of traffic noise on residents. Researchers first used a high-precision microphone array to collect noise in this section and continuously monitored it for 7 days, respectively collecting noise data during the morning rush hour (7:00 - 9:00), flat peak period (10:00 - 16:00), evening rush hour (17:00 - 19:00), and night (22:00 - 6:00). Through the analysis of the NoiseSpecNet model, it was found that the traffic noise in this section mainly concentrated in the frequency band of 600 Hz to 3800 Hz, especially in the frequency band of 1200 Hz to 2200 Hz, where the sound pressure level was the highest, reaching 83.7 dB. Through the evaluation of the A-weighting network, the noise in this frequency band had the most significant impact on human hearing. The researchers decided to design a targeted light-weight, high-strength, corrosion-resistant, and highly sound-absorbing road sound barrier, with a structure of 10 mm of ordinary perforated UHPC board + basalt fiber cloth + 25 mm of through-hole ceramic board + 20 mm of non-perforated UHPC board + 2 mm of sound insulation felt + 20 mm of non-perforated UHPC board. The structure of this sound-absorbing barrier from the inside to the outside of the road is specifically a perforated UHPC board, basalt fiber cloth, through-hole ceramic board, the first non-perforated UHPC board, sound insulation felt, and the second non-perforated UHPC board.

[0186] First, the researchers optimized the perforation parameters through the NoiseSpecNet noise spectrum feature extraction model, inputting the measured noise spectrum data, the incident angle range (±55° in the horizontal plane, 0° to 40° in the vertical plane), and the environmental conditions (temperature from -25°C to 45°C, humidity from 35% to 95%) into the model. After calculation, it was determined that for the ordinary perforated UHPC board, the hole diameter is 5 mm, the hole spacing is 12 mm, and the perforation rate is 13%.

[0187] Subsequently, the researchers used a numerically controlled drilling machine to precisely perforate a 10-mm-thick UHPC board (compressive strength 165 MPa, flexural strength 32 MPa). The drill bit was made of diamond-coated cemented carbide, and the diameter accuracy of the drill bit was controlled within ±0.03 mm. During the drilling process, the rotational speed of the drill bit was set at 1350 rpm, the feed rate was 18 mm / min, and a water cooling system was used for cooling. The results of the perforation quality inspection are shown in Table 1:

[0188] Table 1 Results of the quality inspection of the perforated UHPC board

[0189] Test Items Design Value Measured Value Error Whether Qualified Aperture (mm) 5.00 5.03 +0.03 Qualified Hole Spacing (mm) 12.00 11.97 -0.03 Qualified Perforation Rate (%) 13.00 13.15 +0.15 Qualified Hole Position Error (mm) 0 0.11 +0.11 Qualified Hole Wall Roughness (μm) <30 25.6 - Qualified

[0190] A basalt fiber cloth (density 2.78 g / cm 3 , tensile strength 2250 MPa, fiber diameter 10 μm, areal density 350 g / m 2 ) was laid on the back of the perforated UHPC board. A two-component epoxy resin adhesive (shear strength 17.5 MPa, tensile strength 28.3 MPa) was selected and the main agent and curing agent were mixed in a ratio of 2:1. The thickness of the adhesive application was controlled at 0.9 mm, a uniform pressure of 7.5 kPa was applied, and it was cured for 24 hours in an environment of 22°C. The test result of the bonding strength reached 3.8 MPa, far exceeding the design requirement of 1.5 MPa.

[0191] Next, the through-hole ceramic board (density 2.05 g / cm 3 , porosity 52%, compressive strength 35 MPa) was placed corresponding to the basalt fiber cloth and bonded with a waterproof polyurethane adhesive (bonding strength 1.8 MPa, elongation at break 410%). The thickness of the adhesive layer was 1.2 mm, a uniform pressure of 10 kPa was applied, and it was cured for 24 hours in an environment of 23°C and relative humidity of 45%. The test result of the bonding strength was 2.1 MPa, exceeding the design requirement of 1.5 MPa.

[0192] Subsequently, a first non-perforated UHPC board (thickness 20 mm, compressive strength 170 MPa) and a sound insulation felt (thickness 2 mm, density 110 kg / m 3) The second non-porous UHPC board (with a thickness of 20 mm and a compressive strength of 168 MPa). Each layer is bonded with an acrylate structural adhesive (with a bonding strength of 2.5 MPa), and the thickness of the adhesive layer is 1.0 mm. Finally, the entire composite structure is sealed along the edges with a polysulfide sealant to prevent water infiltration.

[0193] The completed composite structure is placed in a constant temperature and humidity curing chamber. The temperature is set at 22.5 °C, the relative humidity is set at 92%, and the curing time is 18 days. The humidity is maintained at 92% for the first 10 days and gradually reduced to 87% in the following 8 days. After curing, it is slowly dried in the natural environment for 8 days to fully release the internal stress.

[0194] The acoustic performance of the cured sound barrier unit is tested, and the sound absorption coefficient is tested using the impedance tube method.

[0195] The test results are shown in Table 2:

[0196] Table 2 Test Results of Sound Absorption Coefficient of Sound Barrier

[0197] Frequency (Hz) Sound Absorption Coefficient Frequency (Hz) Sound Absorption Coefficient 125 0.26 1000 0.92 250 0.45 2000 0.97 500 0.85 4000 0.88 630 0.88 5000 0.82 800 0.90 Average Value (500 - 4000Hz) 0.90

[0198] The test results of sound insulation quantity are shown in Table 3:

[0199] Table 3 Test Results of Sound Insulation Quantity of Sound Barrier

[0200] Frequency (Hz) Sound Insulation Quantity (dB) Frequency (Hz) Sound Insulation Quantity (dB) 125 23.5 1000 38.2 250 26.8 2000 40.5 500 32.7 4000 36.8 630 34.5 5000 34.2 800 36.3 Average Value (500 - 4000Hz) 36.5

[0201] The test results of mechanical properties are shown in Table 4:

[0202] Table 4 Test Results of Mechanical Properties of Sound Barrier

[0203]

[0204] The test results of durability are shown in Table 5:

[0205] Table 5 Test Results of Durability of Sound Barrier

[0206]

[0207]

[0208] The completed sound barrier unit is installed beside the highway. The foundation is made of C35 reinforced concrete with a depth of 850 mm. M20 bolts are embedded with a spacing of 1200 mm. A 12-mm-thick rubber damping pad with a hardness of 62 Shore A is set in the installation groove at the bottom of the sound barrier unit. The units are connected with an EPDM rubber sealing strip with a hardness of 67 Shore A and a tensile strength of 8.3 MPa. The whole is coated with an anti-ultraviolet coating with a thickness of 75 μm and an ultraviolet reflectivity of 88%.

[0209] After installation, on-site acoustic performance tests were carried out, and the test results are shown in Table 6 as follows:

[0210] Table 6 On-site Acoustic Performance Test Results

[0211]

[0212] Six months after installation, performance re-tests were carried out. The sound barrier performance was stable, the noise reduction effect showed no obvious attenuation, and there were no obvious pollution and corrosion phenomena on the surface, fully verifying the practicability and durability of the sound barrier structure.

[0213] The road sound barrier developed by the present invention has obvious advantages compared with traditional sound barriers. Traditional road sound barriers mainly adopt the structural form of perforated metal plates + glass wool / rock wool. Although the initial sound absorption performance is good, there are many problems: the metal plates are prone to corrosion, and the surface area ash seriously affects the appearance; the performance of glass wool / rock wool drops significantly after rain penetration, and the actual service life is usually only 4 - 6 years; traditional sound barriers are heavy, generally reaching 40 - 50 kg / m 2 , increasing the burden on the foundation structure and the installation difficulty; the sound insulation performance decays rapidly over time, and the decay rate usually reaches 15% - 25% after 3 years. In addition, the design of traditional sound barriers often adopts a "one-size-fits-all" method and is not optimized for the noise characteristics of specific road sections, resulting in poor actual use effects.

[0214] In contrast, the road sound barrier of the present invention has the following advantages: First, the noise characteristics are accurately analyzed through the NoiseSpecNet noise spectrum feature extraction model, enabling the sound barrier structure to be precisely matched with the noise spectrum and improving the sound absorption efficiency; second, a composite structure of perforated UHPC plates + basalt fiber cloth + through-hole ceramic plates is adopted, which not only maintains excellent sound absorption performance but also solves the durability problem, and the service life can reach more than 15 years; third, a fully enclosed structure design is adopted to completely avoid the problem of rain penetration; fourth, the overall weight is reduced to 32 kg / m 2 , about 30% lighter than traditional sound barriers; fifth, it has strong environmental adaptability and stable performance in the temperature range of -30°C to 50°C; sixth, the maintenance cost is low, the surface is not easy to accumulate ash, and even if there is pollution, it is easy to clean. At the same time, the sound barrier of the present invention has a sound absorption coefficient as high as 0.92 - 0.97 and a sound insulation amount as high as 38.2 - 40.5 dB in the frequency band of 1000 Hz - 2000 Hz (the frequency band that has the greatest impact on the human ear), far superior to traditional sound barriers. These advantages make the present invention particularly suitable for the application scenarios of high-grade road sound barriers with high noise reduction requirements, long service life, and low maintenance cost.

[0215] It should be noted that the detailed explanations of the variables involved in the present invention are shown in Tables 7 and 8 as follows.

[0216] Table 7 Variable Explanation Table (First Part)

[0217]

[0218]

[0219] Table 8 Variable Explanation Table (Second Part)

[0220]

[0221] As described above, it 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 be covered within the protection scope of the present invention.

Claims

1. A production method of a lightweight, high-strength, corrosion-resistant and highly sound-absorbing road sound barrier, characterized in that, It includes the following steps: Set up a microphone array at the road edge to collect traffic noise samples, and extract the traffic noise spectrum features through the NoiseSpecNet noise spectrum feature extraction model; obtain the optimal parameter combination of the perforated panel according to the extracted traffic noise spectrum features; perform precise perforation on the ordinary perforated UHPC panel according to the optimized parameters; lay basalt fiber cloth on the back of the perforated UHPC panel; bond the through-hole ceramic panel with the basalt fiber cloth; paste the first non-perforated UHPC panel, sound insulation felt, and the second non-perforated UHPC panel on the back of the through-hole ceramic panel in sequence to form a composite structure; place the composite structure in a constant temperature and humidity curing chamber for curing; conduct acoustic performance tests and mechanical performance tests on the cured sound barrier units; install the qualified sound barrier units on the prefabricated foundation structure and perform waterproof, dustproof treatment and anti-ultraviolet coating spraying.

2. The production method of the lightweight, high-strength, corrosion-resistant and strong sound-absorbing road sound barrier according to claim 1, characterized in that The step of setting up a microphone array to collect traffic noise samples is specifically as follows: Set up multiple high-precision microphone arrays at the road edge, collect traffic noise samples at different times, input the collected traffic noise samples into the NoiseSpecNet noise spectrum feature extraction model, extract the traffic noise spectrum features, determine that the main noise frequency band is from 500 Hz to 4000 Hz, and calculate the influence degree of the sound wave superposition effect on human hearing at different frequencies.

3. The production method of the lightweight, high-strength, corrosion-resistant and highly sound-absorbing road sound barrier according to claim 2, characterized in that, The step of obtaining the optimal parameter combination of the perforated panel is specifically as follows: According to the processing results of the NoiseSpecNet noise spectrum feature extraction model, extract the traffic noise spectrum distribution data, incident angle range, expected sound absorption coefficient target value, sound barrier thickness limit value, and environmental temperature and humidity change range, and output the optimal aperture size, hole spacing, perforation rate, and plate layer thickness combination parameters of the perforated panel through the fully connected layer of the NoiseSpecNet noise spectrum feature extraction model.

4. The production method of the lightweight, high-strength, corrosion-resistant and highly sound-absorbing road sound barrier according to claim 3, characterized in that, The step of performing precise perforation on the ordinary perforated UHPC panel is specifically as follows: According to the optimized parameters, place the ordinary perforated UHPC panel in a mold, and use a special drill bit array to perform precise perforation on the UHPC panel according to the set hole spacing, so that the error of the optimal aperture size is controlled within ±0.1 mm, and the perforation rate reaches the calculated optimal perforation rate.

5. The production method of the lightweight, high-strength, corrosion-resistant and highly sound-absorbing road sound barrier according to claim 4, characterized in that, The step of laying basalt fiber cloth on the back of the perforated UHPC panel is specifically as follows: Lay basalt fiber cloth on the back of the perforated UHPC panel, and use epoxy resin glue to bond and fix the basalt fiber cloth to the perforated UHPC panel to ensure that the two are closely combined without gaps.

6. The production method of the lightweight, high-strength, corrosion-resistant and highly sound-absorbing road sound barrier according to claim 5, characterized in that, The step of bonding the through-hole ceramic panel with the basalt fiber cloth is specifically as follows: Place the through-hole ceramic panel formed by high-temperature sintering and the basalt fiber cloth correspondingly, and use waterproof polyurethane glue for bonding, and control the glue layer thickness between 1 mm and 1.5 mm to ensure that the bonding strength reaches more than 1.5 MPa.

7. The production method of the lightweight, high-strength, corrosion-resistant and highly sound-absorbing road sound barrier according to claim 6, characterized in that, The step of pasting the first non-perforated UHPC panel, sound insulation felt, and the second non-perforated UHPC panel on the back of the through-hole ceramic panel in sequence to form a composite structure is specifically as follows: Paste the first non-perforated UHPC panel, sound insulation felt, and the second non-perforated UHPC panel on the back of the through-hole ceramic panel in sequence to form a composite structure. A staggered joint design is adopted between the layers, and the plate layer thickness is set according to the calculated optimal plate layer thickness combination parameters to enhance the overall structural strength.

8. The production method of the lightweight, high-strength, corrosion-resistant and strong sound-absorbing road sound barrier according to claim 7, characterized in that, The steps of curing the composite structure in a constant temperature and humidity curing chamber are specifically as follows: Place the composite structure in a constant temperature and humidity curing chamber, control the temperature at 20°C to 25°C, keep the relative humidity at 85% to 95%, and the curing time is not less than 14 days to ensure that the bonding material is fully cured.

9. The production method of the lightweight, high-strength, corrosion-resistant and highly sound-absorbing road sound barrier according to claim 8, characterized in that, The NoiseSpecNet noise spectrum feature extraction model specifically refers to a hybrid architecture model based on a convolutional neural network and a long short-term memory network, including five convolutional layers for extracting local features of the noise spectrum, two bidirectional long short-term memory network layers for capturing temporal features, and three fully connected layers for parameter prediction.

10. The production method of the lightweight, high-strength, corrosion-resistant and strong sound-absorbing road sound barrier according to claim 9, characterized in that, The steps for establishing the training data set of the NoiseSpecNet noise spectrum feature extraction model include collecting noise samples under different road types, different traffic flows, and different meteorological conditions, performing time-frequency domain conversion on the noise samples to obtain spectrograms, and annotating the corresponding optimal sound barrier structure parameters and measured sound absorption coefficient data for each spectrogram.

Images