Road bridge noise reduction device real-time monitoring method and system

Abstract

This application relates to the field of traffic infrastructure monitoring technology, and discloses a real-time monitoring method and system for road and bridge noise reduction devices. The method includes: acquiring and cleaning real-time environmental data from a sensor network surrounding the road and bridge, and extracting noise and vibration feature sets; generating a dynamic judgment benchmark based on real-time meteorological and traffic flow data through an environmental parameter fusion model, comparing it with the feature sets to identify high-risk scenarios and trigger alarms; verifying the status of the noise reduction device based on high-risk labels and generating an equipment status report; dynamically adjusting the frequency of the anti-phase sound wave and the barrier angle based on the report and environmental noise feedback, generating and verifying an adaptive control configuration to regulate the device in real time; continuously monitoring the effect after regulation, iteratively optimizing until the optimal noise suppression result is achieved, and storing the results. This invention achieves intelligent perception, precise early warning, and adaptive control of road and bridge noise, significantly improving the stability of noise reduction effects and system operating efficiency.

Description

Technical Field

[0001] This application relates to the field of traffic infrastructure monitoring technology, and in particular to a real-time monitoring method and system for road and bridge noise reduction devices. Background Technology

[0002] Roads and bridges, as crucial infrastructure in urban and high-speed transportation systems, have long been a focus of attention due to their surrounding environmental noise issues. With the continuous increase in traffic volume and the growing number of large bridge structures, tire noise, engine noise, wind noise, and the resulting structural vibrations generated by vehicles not only affect the comfort of nearby residents but may also adversely impact the structural safety of bridges. Therefore, the industry has explored various technical approaches for environmental noise control and monitoring in road and bridge scenarios.

[0003] Currently, the more mature technical means mainly include "passive" noise reduction measures such as setting up sound barriers, optimizing road surface materials, limiting vehicle speed, and arranging green belts. These measures can reduce noise transmission to a certain extent, but they are generally characterized by fixed structures, unadjustable parameters, or inflexible adjustments. They are difficult to respond to fluctuations in traffic flow, changes in weather conditions, and changes in bridge vibration status in a timely manner, and have limited ability to cope with sudden events or instantaneous high-noise scenarios.

[0004] With the development of sensor technology, the Internet of Things (IoT), and communication technology, noise monitoring devices, accelerometers, and traffic detection equipment have begun to be deployed in some road and bridge scenarios, forming preliminary environmental monitoring systems. These systems can typically collect and upload data online on noise levels, traffic flow, or structural vibrations for environmental quality assessment or compliance monitoring. However, existing systems are mostly geared towards "environmental monitoring," focusing on data display and alarms for exceeding limits, lacking dedicated real-time monitoring mechanisms for specific noise reduction devices (such as active noise cancellation devices and adjustable barriers), and their tracking, diagnosis, and feedback on the operational status of noise reduction devices are relatively weak.

[0005] On the other hand, active control technologies for road and bridge noise are gradually emerging. For example, devices that combine anti-phase sound wave generators with adjustable sound barriers can actively suppress noise within specific frequency bands. However, these technologies often focus on the structural design or control algorithms of the noise reduction device itself. In engineering applications, they often use preset or semi-static control parameters and rely solely on a single noise index or simple threshold for start-stop control. They lack comprehensive analysis of multi-source environmental information such as traffic flow, weather conditions, and bridge vibration, and also lack a closed-loop monitoring and iterative optimization mechanism for the operating parameters and noise suppression effect of the noise reduction device. This results in insufficient stability of the noise reduction effect, and the device's operating status is difficult to assess and adjust in a timely manner.

[0006] In summary, under the complex and ever-changing operating environment of roads and bridges, existing technologies have not yet developed a systematic real-time monitoring method for noise reduction devices. This method should be able to identify high-risk noise scenarios and continuously evaluate noise suppression effectiveness based on the collaborative perception of multi-source environmental data and device operating status, providing a reliable basis for subsequent adaptive control. Therefore, how to construct a real-time monitoring method for noise reduction devices under the complex operating environment of roads and bridges, achieving collaborative perception of multi-source environmental data and device operating status, identification of abnormal scenarios, and continuous evaluation of noise suppression effectiveness, has become a key technical problem that urgently needs to be solved in this field. Summary of the Invention

[0007] The purpose of this invention is to overcome the shortcomings of existing road and bridge noise control technologies, such as the lack of systematic real-time monitoring and closed-loop evaluation of noise reduction devices, and the inability to promptly identify high-risk noise scenarios and conduct targeted status checks and adaptive adjustments to noise reduction devices under complex traffic and environmental conditions. This application provides a real-time monitoring method and system for road and bridge noise reduction devices, enabling collaborative perception, dynamic analysis, and continuous evaluation of multi-source environmental data and the operating status of noise reduction devices in the road and bridge operating environment, providing an accurate and reliable monitoring foundation for subsequent adaptive control of noise reduction devices.

[0008] In a first aspect, this application provides a real-time monitoring method for road and bridge noise reduction devices, the method comprising:

[0009] S1. Obtain real-time environmental data from the sensor network deployed around the road and bridge, and perform preliminary cleaning to obtain a preliminary environmental set;

[0010] S2. Extract features from the preliminary environmental set, extract features of different frequency bands of noise signals and vibration features of road and bridge structures, and obtain the classified feature set.

[0011] S3. Generate a dynamic judgment benchmark based on real-time environmental parameters and environmental parameter fusion model, compare the classified feature set with the dynamic judgment benchmark, detect anomalies and trigger alarms, and determine high-risk scene labels.

[0012] S4. Based on the high-risk scenario labels, the status of the noise reduction device is checked in real time and the data storage is optimized to generate an equipment status report;

[0013] S5. Dynamically adjust the control parameters of the noise reduction device according to the equipment status report, generate an adaptive control configuration based on the adjustment results, and regulate the working status of the noise reduction device in real time through the adaptive control configuration.

[0014] S6. After regulation, continuously monitor environmental fluctuations and analyze the noise suppression effect. If the effect does not meet the preset target, generate optimization suggestions and iteratively adjust the noise reduction device operation strategy until the noise suppression effect meets the requirements.

[0015] Secondly, this application provides a real-time monitoring system for road and bridge noise reduction devices, the system comprising:

[0016] The data acquisition module is used to acquire real-time environmental data from the sensor network deployed around the road and bridge, and to perform preliminary cleaning to obtain a preliminary environmental dataset.

[0017] The feature extraction module is used to extract features from the preliminary environmental set, extract features of different frequency bands of noise signals and vibration features of road and bridge structures, and obtain a classified feature set.

[0018] The risk identification module is used to generate a dynamic judgment benchmark based on real-time environmental parameters and environmental parameter fusion model, compare the classified feature set with the dynamic judgment benchmark, detect anomalies and trigger alarms, and determine high-risk scene labels.

[0019] The status verification module is used to verify the status of the noise reduction device in real time based on high-risk scenario tags, optimize data storage, and generate equipment status reports.

[0020] The parameter control module is used to dynamically adjust the control parameters of the noise reduction device according to the equipment status report, generate an adaptive control configuration based on the adjustment results, and adjust the working status of the noise reduction device in real time through the adaptive control configuration.

[0021] The iterative optimization module is used to continuously monitor environmental fluctuations after regulation and analyze the noise suppression effect. If the effect does not meet the preset target, it generates optimization suggestions and iteratively adjusts the operation strategy of the noise reduction device until the noise suppression effect meets the requirements.

[0022] Compared with the prior art, the beneficial effects of the technical solution of this application are at least as follows:

[0023] 1. This invention enables the joint acquisition, cleaning, and feature extraction of multi-source environmental data such as traffic flow density, environmental noise signals, road and bridge structural vibration, and meteorological conditions in road and bridge scenarios. It also uses an environmental parameter fusion model to perform collaborative analysis of noise and vibration, thereby more accurately identifying high-risk noise scenarios with excessive noise intensity and abnormal structural vibration, providing a refined and scenario-based monitoring foundation for the real-time monitoring of road and bridge noise reduction devices.

[0024] 2. By using high-risk noise scene labels to drive the status verification of noise reduction devices, and introducing mechanisms such as equipment status self-inspection, vibration data compression, and cloud interaction minimization, key operating parameters of the noise reduction devices can be obtained in a timely manner, while effectively reducing data storage and transmission overhead. In road and bridge site environments with limited network and storage resources, this is conducive to achieving long-term and stable monitoring of the operating status of noise reduction devices.

[0025] 3. By analyzing the operating parameters in the equipment status report, the control parameters such as the output frequency of the anti-phase acoustic wave and the barrier angle are adaptively adjusted. Before and after the adjustment, environmental fluctuations are continuously monitored, historical data is compared, and the operating strategy is iteratively optimized. Thus, the appropriate noise reduction control configuration can be automatically matched under different traffic flow, weather and structural vibration conditions to achieve stable noise suppression effect of the road and bridge noise reduction device under complex working conditions, reduce manual intervention, and improve noise reduction efficiency and device operation reliability. Attached Figure Description

[0026] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0027] Figure 1 This is a flowchart of the real-time monitoring method for the road and bridge noise reduction device of this application;

[0028] Figure 2 This is a schematic diagram of the time-series effect of the adaptive control process in the technical effect characterization diagram of the embodiments of this application;

[0029] Figure 3 This is a schematic diagram illustrating the convergence process of the iterative optimization effect of the technical effect characterization diagram in the embodiments of this application;

[0030] Figure 4 This is a schematic diagram of the real-time monitoring system for road and bridge noise reduction devices in this application. Detailed Implementation

[0031] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a particular order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in a sequence other than that illustrated or described herein. Furthermore, the terms “comprising” or “having,” and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0032] For ease of understanding, the specific process of the embodiments of this application is described below. Figure 1 The diagram shows a flowchart of the real-time monitoring method for the road and bridge noise reduction device provided by the present invention. The flowchart specifically includes the following steps:

[0033] S1. Obtain real-time environmental data from the sensor network deployed around the road and bridge, and perform preliminary cleaning to obtain a preliminary environmental set.

[0034] In one specific embodiment, the process of performing step S1 may specifically include the following steps:

[0035] Real-time environmental data, including at least traffic flow density data, road and bridge structure vibration data, and environmental noise signals, is acquired from the sensor network at a preset acquisition frequency. A built-in dynamic adjustment mechanism for abnormal thresholds filters the traffic flow density data, road and bridge structure vibration data, and environmental noise signals, excluding data that does not meet the preset range. The filtered data is then integrated into a preliminary environmental dataset. For this preliminary environmental dataset, the acquisition time and sensor location information are recorded to form a complete data record. The complete data record is then verified for integrity. If data is missing, a preset interpolation algorithm is used to complete the data, obtaining the preliminary environmental dataset.

[0036] Specifically, the sensor network includes various types of sensor nodes deployed on the bridge deck, piers, guardrails on both sides of the road under the bridge, and nearby green belts. These nodes include at least: meteorological sensors (including at least wind speed, temperature, and humidity sensors) for collecting meteorological condition data, vehicle detection sensors for collecting traffic flow parameters, acceleration sensors for collecting road and bridge structure vibration data, and acoustic sensors for collecting environmental noise signals. Each sensor node establishes a data transmission connection with the edge computing unit through a wired bus or wireless communication, thereby forming a monitoring network covering the target road and bridge area.

[0037] In this embodiment, the edge computing unit acquires real-time environmental data from the sensor network according to a preset acquisition frequency. The acquisition frequency is set according to the traffic characteristics of the road and bridge scenario. For example, the acquisition period is set to 1 second during peak traffic hours in the daytime and 5 seconds during low traffic hours at night, so as to control the amount of data while ensuring data timeliness. During each acquisition, the meteorological sensor outputs real-time wind speed, temperature, and humidity values; the vehicle detection sensor outputs traffic flow parameters such as the number of vehicles passing through the monitoring section per unit time, instantaneous vehicle speed, and / or lane occupancy; the accelerometer outputs acceleration values ​​along the three directions of the structure, which are integrated or processed in the frequency domain to obtain vibration amplitude or vibration energy parameters; the acoustic sensor outputs the time-domain sampling sequence of the environmental noise signal, and the edge computing unit records the sampling values ​​and sound pressure level.

[0038] After acquiring the traffic flow parameters, the edge computing unit does not directly use the number of vehicles passing through the cross-section per unit time as the traffic flow density. Instead, it calculates the traffic flow density data based on a preset traffic flow conversion relationship (pre-calibrated according to historical monitoring data or a standard traffic flow model). Preferably, the edge computing unit first calculates the traffic flow rate based on the number of vehicles passing through the monitoring cross-section per unit time, and then combines the average driving speed or lane occupancy rate within the corresponding time window to obtain the traffic flow density according to the preset traffic flow conversion. The average driving speed preferably uses the spatial average speed consistent with the traffic flow rate statistical window, and the units of measurement for each physical quantity are unified when converting between traffic flow rate, speed, and density to ensure dimensional consistency. In one embodiment, when the traffic flow rate and spatial average speed are acquired, the traffic flow density can be calculated based on the basic traffic flow relationship; when speed data is missing or fluctuates significantly, the traffic flow density can also be estimated by combining the lane occupancy rate through a calibrated conversion relationship. The above method converts the raw traffic flow parameters output by vehicle detection sensors into traffic flow density data consistent with the input of the subsequent environmental parameter fusion model, avoiding object confusion and dimensional errors caused by mixing traffic flow, speed, and density. Traffic flow density is the number of vehicles per unit road length, while traffic flow rate is the number of vehicles passing through the monitoring section per unit time. The two are related but not identical through a preset traffic flow conversion relationship.

[0039] After acquiring real-time environmental data, the edge computing unit performs preliminary cleaning of the data through a built-in dynamic adjustment mechanism for anomaly thresholds. In one embodiment, this mechanism is based on a sliding time window to statistically analyze the distribution of historical data, calculating statistics for threshold updates for each type of data. For example, for noise sound pressure level data, the sound pressure level sequence of each sampling point within a sliding window of length M (a preset configurable parameter) is acquired, and the mean and standard deviation within the window are calculated. Then, based on the mean and standard deviation, and preset coefficients, a dynamic upper threshold and a dynamic lower threshold for noise are set. When a newly acquired noise sound pressure level exceeds this range, it is determined as an outlier and filtered. To distinguish between sensor malfunctions and genuine physical anomalies, a multi-source verification rule is introduced: if a data point exceeds a dynamic threshold, data from other related sensors (such as vibration, traffic flow, and weather) are simultaneously checked at the same time. If other sensors show no anomalies and the change trend is gradual, it is determined to be a sensor malfunction or occasional interference and is removed. If other sensors also show correlated anomalies (such as a sudden increase in noise and vibration simultaneously, or a sudden increase in noise accompanied by a sudden change in wind speed), it is determined to be a genuine high-risk event, retained, and marked as a "suspected anomaly" for subsequent in-depth analysis. This dynamic threshold mechanism based on statistical distribution can adapt to changes in noise levels caused by traffic flow at different times, solving the technical problem that fixed thresholds cannot cover both peak and off-peak periods, leading to numerous misjudgments and the discarding of a large amount of valid data. Traffic flow density data and road and bridge structure vibration data are also filtered using the same method. By applying dynamic threshold filtering mechanisms to different data sources, it is possible to remove instantaneous jumps caused by sensor malfunctions and occasional impact noise caused by external non-traffic events, reducing the interference of abnormal data on subsequent monitoring and analysis and improving the reliability of real-time monitoring results.

[0040] After dynamic threshold filtering, the edge computing unit integrates the filtered valid data to form a preliminary environmental dataset. This dataset includes at least meteorological condition data (wind speed, temperature, humidity), traffic flow density data, road and bridge structure vibration data, and environmental noise signals at the same timestamp. During the integration process, acquisition time and sensor location information are added to each record. The acquisition time can use a unified edge node clock timestamp, and the sensor location information can be represented by a combination of identifiers such as latitude and longitude coordinates and bridge span numbers. By creating an information structure containing time and spatial identifiers for each set of data, the technical problem of not being able to trace the environmental state at a specific time and location in subsequent analysis is solved, which is beneficial for locating anomalies to specific road sections or structural locations during the monitoring phase.

[0041] Integrity verification comprises two parts: temporal continuity checking and spatial coverage checking. In the temporal continuity checking, the theoretical time interval between adjacent records is calculated based on the acquisition frequency, and the difference between the timestamps of actual adjacent records is compared with this theoretical time interval. If the difference exceeds a preset allowable error, data loss is determined within this time interval. In the spatial coverage checking, a preset sensor deployment list is used to check whether all sensors at key locations have generated valid records within the current time slice. If a key sensor fails to generate valid records for multiple consecutive acquisition cycles, data loss or sensor malfunction is determined at that location. By performing integrity verification on the preliminary environmental dataset, data gaps caused by communication interruptions, sensor disconnections, or other issues during the acquisition process can be promptly identified. When data loss is detected, the edge computing unit uses a preset interpolation algorithm to complete the missing data to obtain the preliminary environmental dataset. In one implementation, linear interpolation algorithms are used for noise sound pressure levels and traffic flow density, while piecewise linear interpolation or frequency domain feature-based interpolation methods are used for vibration data. For vibration data, if the vibration response changes relatively smoothly (i.e., the root mean square value of the vibration signal changes less than a preset threshold within the sliding window, and the spectral energy distribution has no obvious main peak, indicating stable signal amplitude and no obvious periodic fluctuations), linear interpolation can also be used. For vibration signals with obvious periodic components, after obtaining the main frequency components by performing a short-time Fourier transform on the adjacent valid data, the vibration amplitude at the missing time can be constructed in the frequency domain, and then the time domain estimate can be obtained through inverse transform (for example, the main frequency components can be estimated by performing time-frequency analysis (such as short-time Fourier transform) on the adjacent valid data segments, and then the signal estimate at the missing time can be synthesized). By using interpolation algorithms to fill in the missing data, the initial environmental set maintains continuity in the time dimension and ensures that there is data available for analysis at the main monitoring locations in the spatial dimension, solving the technical problem that the working environment of road and bridge noise reduction devices cannot be continuously evaluated due to data loss during the monitoring process.

[0042] S2. Extract features from the preliminary environmental set, extract features of different frequency bands of noise signals and vibration features of road and bridge structures, and obtain the classified feature set.

[0043] In one specific embodiment, the process of performing step S2 may specifically include the following steps:

[0044] At edge computing nodes, spectral analysis is performed on environmental noise signals to extract noise features in different frequency bands;

[0045] Simultaneously, waveform analysis was performed on the vibration data of the road and bridge structure to obtain vibration characteristics;

[0046] By using threshold classification, noise features are divided into continuous and burst types to form a noise feature set, and vibration features are divided into normal and abnormal types to form a vibration feature set. The noise feature set and vibration feature set are standardized, and principal component analysis is used to reduce the dimensionality of the standardized features to obtain the classified feature set.

[0047] Specifically, S2 relies on edge computing nodes deployed at the road and bridge site to process the preliminary environmental set. Each data record in the preliminary environmental set contains at least a discrete sequence of environmental noise signal, a discrete sequence of road and bridge structure vibration signal, and the timestamp and sensor location information of the record at a certain sampling time. Therefore, the edge computing node can complete feature extraction without accessing the cloud, thereby meeting the processing latency requirements of the real-time monitoring method for road and bridge noise reduction devices.

[0048] In this embodiment, the feature extraction process for environmental noise signals is performed in batches on edge computing nodes. Each edge computing node selects a noise sampling sequence of length E1 as a frame signal, with a sampling frequency denoted as E2 and a frame length of E3 seconds. E1 and E2 satisfy E1 = E2 × E3. A window function (such as a Hamming window) is applied to each frame signal, multiplying the original sampled values ​​by the corresponding window function coefficients to reduce spectral leakage. Then, a discrete Fourier transform is performed on the windowed sequence to obtain the frequency domain spectral value sequence of the noise signal for that frame, with a frequency resolution of E2 / E1. Based on the characteristics of road and bridge noise sources, the entire frequency band is divided into several frequency segments. For example, 20 Hz to 200 Hz is classified as the low-frequency band, 200 Hz to 1000 Hz as the mid-frequency band, and 1000 Hz to 4000 Hz as the high-frequency band. The corresponding spectral amplitude or power spectral density is accumulated for each frequency band to obtain the energy characteristic values ​​of the noise in different frequency segments, thus forming a noise feature vector containing multi-dimensional frequency band energy characteristics. By performing spectral analysis and frequency band energy statistics, the original time-domain noise signal can be compressed into a feature quantity with lower dimensionality but containing the main frequency information. This solves the technical problem of excessive computational load and difficulty in distinguishing the frequency components of different noise sources when directly processing large-scale time-domain sampled data in real-time monitoring scenarios.

[0049] While performing spectral analysis on the noise signal, waveform analysis is performed on the vibration data of the road and bridge structure to obtain vibration characteristics. The edge computing node selects a vibration acceleration sampling sequence of length V1 as the analysis window, the vibration sampling frequency is denoted as V2, and the analysis window length is denoted as V3 seconds. V1 and V2 satisfy V1 = V2 × V3. For each analysis window, multiple time-domain features are calculated, such as the root mean square value of vibration acceleration V4, peak acceleration V5, and peak factor V6, where V6 is the ratio of V5 to V4. Simultaneously, the edge computing node can perform a discrete Fourier transform on the vibration data to obtain the dominant vibration frequency V7 and the corresponding frequency band energy, representing the modal characteristics of the structural vibration, thereby constructing a vibration feature vector containing multiple vibration descriptors. Through waveform analysis combining the time and frequency domains, key features reflecting the response intensity and vibration modes of the road and bridge structure can be extracted from the vibration data.

[0050] Threshold classification is performed separately for noise and vibration characteristics. Within a certain time window, the rate of change and peak characteristics of energy features in each frequency band are statistically analyzed. For example, for the low-frequency energy feature sequence F1(k), the energy difference between adjacent frames F2(k) = F1(k) - F1(k-1) is calculated, and the mean and peak values ​​of the absolute values ​​of the differences within this window are calculated. When the mean difference is small and the peak value is not obvious, the feature of this frequency band is classified as a persistent feature; when the peak value of the difference is significantly greater than the statistical mean at certain time points and the duration is short, it is marked as a burst feature. Burst features are often related to instantaneous noise events such as horn sounds and collision sounds. The same difference and peak value statistics are performed on the mid-frequency and high-frequency energy features, and persistent and burst types are classified according to preset thresholds, thus forming a noise feature set, where each element contains information about its frequency band and a persistent or burst type label. By classifying noise characteristics into continuous and sudden types through threshold classification, it is beneficial to distinguish between background traffic noise and short-term abnormal noise events in the real-time monitoring of road and bridge noise reduction devices. This solves the problem that different types of noise require different processing methods in control strategies, while traditional monitoring methods cannot distinguish noise types.

[0051] For vibration characteristics, edge computing nodes compare parameters such as the root mean square (RMS) value V4, peak ground acceleration (PGA) V5, peak factor V6, and dominant frequency V7 with historical statistical values ​​or structural design limits to classify them as normal or abnormal. In one embodiment, the edge computing nodes pre-store historical vibration statistical intervals or safety thresholds determined based on design specifications for a specific bridge. For example, the safe range for the RMS value is [W1, W2], and the safe range for the PGA is [W3, W4]. When V4 and V6 are both within their respective safe ranges in a certain analysis window, and the dominant frequency V7 does not show a significant shift, the vibration characteristic corresponding to that window is marked as normal. If V4 exceeds W2 or V6 exceeds W4, or the dominant frequency deviates from the historical dominant frequency by more than a preset offset threshold W5, the vibration characteristic is marked as abnormal. A vibration feature set is constructed accordingly, where each feature vector is accompanied by a normal or abnormal label and a corresponding sensor location identifier. This threshold classification method based on multiple indicators can quickly identify abnormal vibration states that may be related to structural damage, impact loads, or wind-induced vibrations during real-time monitoring of road and bridge noise reduction devices, thus solving the problem of misjudgment or missed judgment that may occur when relying on a single threshold.

[0052] After obtaining the noise and vibration feature vectors, the edge computing nodes standardize all feature vectors to obtain dimensionless standardized eigenvalues, eliminating the influence of different physical dimensions. Then, the edge computing nodes perform principal component analysis (PCA) dimensionality reduction on all standardized features to compress feature dimensions, remove redundant information, and reduce subsequent computational load. Specifically, a feature matrix Z6 is constructed, where each row corresponds to a feature vector of a sample (containing all standardized noise frequency band energy features and vibration description features), and each column corresponds to a feature dimension. The covariance matrix Z7 of this matrix is ​​calculated, and the eigenvalues ​​and eigenvectors of Z7 are solved. The top principal components with a cumulative variance contribution rate exceeding a preset proportion Z8 (e.g., 90%) are selected as new comprehensive feature vectors. Through PCA dimensionality reduction, the original high-dimensional feature set is compressed into a lower-dimensional comprehensive feature set, retaining the main information while reducing the computational burden. This solves the technical problem of excessive computational load and difficulty in online real-time response caused by directly using high-dimensional feature sets.

[0053] After PCA dimensionality reduction, the original high-dimensional feature vector of each sample is compressed into a low-dimensional composite feature vector. Simultaneously, the noise type label (persistent / burst) and vibration state label (normal / abnormal) obtained in step S2 are associated with this composite feature vector as auxiliary information. Each sample, consisting of its composite feature vector, noise type label, and vibration state label, constitutes a complete input unit, collectively referred to as the 'classified feature set'. The composite feature vector characterizes the numerical features of the sample, while the noise type label and vibration state label provide category indication information. Both serve as input to the subsequent risk identification module (S3), providing richer decision-making basis for high-risk scene identification.

[0054] After obtaining the noise and vibration feature vectors and before standardization, edge computing nodes can further perform correlation analysis to explore the coupling relationship between noise and vibration. Specifically, noise energy feature sequences and vibration feature sequences (such as low-frequency noise energy and vibration root mean square) in a specific frequency band are selected, and the Pearson correlation coefficient Z4 is calculated within a time window of length Z3. The correlation analysis results are used to characterize the coupling relationship between noise and vibration, and are not used as triggering conditions for feature optimization, but rather as reference information for subsequent interpretation of high-risk scenarios, tracing alarm causes, or assisting in the selection of control strategies. This analysis is performed before dimensionality reduction, which preserves the physical interpretability of the original features.

[0055] The monitoring system transforms the original noise and vibration signals into structured features and introduces noise type classification, vibration state discrimination, and cross-modal correlation characterization. This enables the real-time monitoring method for road and bridge noise reduction devices to make judgments based on a few key features in subsequent monitoring stages, thereby improving the accuracy and response speed of high-risk scene identification and reducing dependence on communication bandwidth and cloud computing resources.

[0056] S3. Based on real-time environmental parameters and environmental parameter fusion models, a dynamic judgment benchmark is generated. The classified feature set is compared with the dynamic judgment benchmark to detect anomalies and trigger alarms, thereby determining high-risk scene labels.

[0057] In one specific embodiment, the process of performing step S3 may specifically include the following steps:

[0058] An environmental parameter fusion model is established based on historical datasets, which include historical meteorological condition data, historical traffic flow density data, and historical noise feature sets and historical vibration feature sets that are collected and extracted simultaneously.

[0059] Real-time meteorological data and traffic density data are input into an environmental parameter fusion model to obtain the permissible noise intensity and permissible vibration range based on the current environment.

[0060] The current noise signal intensity is extracted based on the classified noise feature set, and the current vibration waveform accuracy features are extracted based on the classified vibration feature set.

[0061] The noise signal intensity and vibration waveform accuracy characteristics are compared with the output of the environmental parameter fusion model. If the noise signal intensity exceeds the allowable noise intensity and the vibration waveform accuracy characteristics exceed the allowable vibration range, an abnormal early warning response mechanism is triggered to generate an alarm signal. Based on the alarm signal and the environmental parameters at the time of triggering, the current environmental state is labeled to determine the high-risk noise scene label.

[0062] Specifically, step S3 is completed collaboratively by edge computing nodes and the backend server. The backend server is responsible for establishing and updating the environmental parameter fusion model based on historical datasets, thereby forming a dynamic judgment benchmark that changes with traffic flow and weather conditions in the real-time monitoring scenario of the road and bridge noise reduction device. The classified feature set is then compared with this dynamic judgment benchmark to detect anomalies and trigger alarms, thereby determining the label of high-risk noise scene. In this step, the environmental parameter fusion model is mainly used to characterize the influence of meteorological conditions and traffic load on noise propagation conditions and road and bridge structural response distribution, especially to depict the statistical change boundary of noise signal intensity and vibration waveform accuracy characteristics under normal operating conditions as the environment changes. Based on this, dynamic allowable boundaries for anomaly judgment are formed by combining standard limits and safety margins. The classified noise feature set and vibration feature set are used to characterize the noise state and structural vibration state obtained from real-time monitoring and are compared with the dynamic judgment benchmark to determine whether there are any anomalies in the current environmental state.

[0063] It should be noted that the "classified feature set" output by S2 includes the dimensionality-reduced comprehensive feature vector and its associated noise type and vibration state labels. During the online execution phase of S3, the current noise signal intensity and vibration waveform accuracy features are extracted from this feature set and compared with the dynamic allowable boundary output by the model. The environmental parameter fusion model of S3 itself uses historical environmental parameters and corresponding upper limits for noise signal intensity and vibration waveform accuracy features as supervision labels during training, aiming to establish a mapping relationship between environmental parameters and dynamic allowable boundaries, without directly using the dimensionality-reduced comprehensive feature vector.

[0064] The establishment of the environmental parameter fusion model uses historical datasets as input. The sources of the historical datasets are the same as those in steps S1 and S2. Each historical record in the historical dataset contains historical meteorological condition data (at least wind speed, temperature and humidity), historical traffic flow density data, historical environmental noise signals and historical road and bridge structure vibration data collected synchronously within the same time segment. The historical environmental noise signals and historical road and bridge structure vibration data undergo the same processing procedure as in step S2 to obtain historical noise feature sets and historical vibration feature sets (including historical vibration waveform accuracy features), thereby ensuring the consistency of model training data and online application data in terms of data structure and statistical distribution.

[0065] Preferably, the training samples in the historical dataset are selected from historical time segments where the equipment is operating normally, no structural abnormalities have occurred, and no high-risk alarms have been triggered. This allows the environmental parameter fusion model to learn a mapping relationship that reflects the variation law of allowable boundaries under normal operating conditions, reducing the interference of abnormal samples on the estimation results of allowable intensity and allowable range. Furthermore, the "variation law of allowable boundaries" does not refer to directly using the historical observation values ​​themselves as allowable values. Instead, the model first learns the mapping relationship between environmental parameters under normal operating conditions and the statistical upper bound of noise signal intensity and the statistical upper bound of vibration waveform accuracy characteristics. Then, the statistical upper bounds are combined with specification constraints and safety margins to generate the final allowable boundaries used for online judgment.

[0066] In this embodiment, the environmental parameter fusion model employs a supervised regression structure to output the allowable noise intensity and the allowable vibration range (i.e., the allowable threshold for vibration waveform accuracy). The model input consists of meteorological data and traffic flow density data, and the model output consists of the allowable noise intensity threshold and the allowable vibration waveform accuracy threshold under the given input conditions. For clarity and ease of separate optimization, the model can be constructed using two regression sub-models: one for predicting the allowable noise intensity and the other for predicting the allowable vibration waveform accuracy threshold. The two sub-models share input features but have different outputs. The input feature vector, denoted as G1, is obtained by concatenating wind speed components (horizontal and vertical wind speeds), humidity components, temperature components, and traffic flow density components. Each component is normalized and mapped to the [0, 1] interval to eliminate dimensional differences. The noise allowable intensity output is denoted as G2, and the vibration waveform accuracy threshold output is denoted as G3. During model training, the meteorological conditions and traffic density of each sample in the historical dataset are used as input. The dynamic allowable upper limit formed by the statistical value of the noise signal intensity corresponding to the sample after safety margin correction is used as the noise supervision label. The dynamic allowable boundary formed by the statistical upper limit of the vibration waveform precision feature corresponding to the sample after safety margin correction is used as the vibration supervision label. The noise supervision label can be the upper quantile value of the historical noise signal intensity or the upper limit of the mean plus the standard deviation. The vibration supervision label is the statistical upper limit (such as the upper quantile value) of the historical vibration waveform precision feature (calculated by the weighted combination of the dominant frequency drift and amplitude fluctuation rate). In combination with environmental protection limits, structural design requirements or engineering experience, a preset safety margin is introduced for boundary correction so that the model output represents the acceptable dynamic allowable intensity and dynamic allowable range under the current environmental conditions, rather than simply the historical average level.

[0067] More specifically, the model learning objective corresponds to "the estimated result of the statistical upper bound of the response under normal operating conditions under given environmental conditions," and the allowable noise intensity and vibration range are determined by the estimated result after boundary correction. Therefore, the model output semantically belongs to the integrated expression of the intermediate estimation result and the final judgment boundary used to determine the allowable boundary. Its essence is not a simple regression of the regulatory limit, nor a direct fit of the historical mean, but a representation of the dynamic judgment boundary after the statistical upper bound is corrected by safety constraints.

[0068] In one implementation, the safety margin is introduced by boundary correction of historical statistical upper limits. For the permissible noise intensity, the statistical upper limit of the noise signal intensity is first calculated based on historical samples within the corresponding meteorological conditions and traffic density range, denoted as G2_stat; then, combined with the standard permissible upper limit G2_norm under the corresponding scenario and the preset safety margin ST2, the permissible noise intensity is determined as: G2 = min(G2_stat + ST1, G2_norm - ST2), where ST1 is the statistical fluctuation compensation amount and ST2 is the safety margin retention amount. For the permissible threshold of vibration waveform accuracy, the statistical upper limit of the vibration waveform accuracy characteristics is first calculated, denoted as G3_stat; then, combined with the structural design permissible boundary G3_norm and the preset safety margin ST4, the permissible threshold of vibration waveform accuracy is determined as: G3 = min(G3_stat + ST3, G3_norm - ST4), where ST3 is the statistical fluctuation compensation amount and ST4 is the safety margin retention amount. By employing the above method, the permissible intensity and range output by the model reflect both the statistical regularity under historical normal operating conditions and the requirements of engineering specifications and safety boundaries. In other words, G2_stat and G3_stat reflect the statistical upper bound of the response corresponding to environmental conditions, while G2_norm and G3_norm reflect the upper limit of the allowable range specified by the specifications or design. The final G2 and G3 used for judgment are the convergence results of these two parameters after introducing a safety margin. Therefore, the relationship between the model and boundary judgment can be summarized as follows: first, estimate the upper bound of the environmentally relevant response; then, determine the dynamic permissible boundary; and finally, use this permissible boundary as the online anomaly judgment criterion.

[0069] It is important to clarify that the aforementioned safety margin correction process is completed during the construction of supervisory labels in the model training phase, not as a post-processing step after the model output. Specifically, for each historical training sample, its corresponding statistical upper bound G2_stat is first calculated. Then, the dynamic allowable upper limit after safety margin correction is calculated using the formula G2 = min(G2_stat + ST1, G2_norm - ST2), and this value is used as the supervisory label for that sample. The model learns the mapping from environmental parameters to this supervisory label and directly outputs the final dynamic allowable boundary G2 used for online judgment. Therefore, during online runtime, inputting real-time environmental parameters into the model yields G2, which is the dynamic allowable boundary that can be directly compared with the current noise signal intensity, without the need for additional correction. The vibration waveform accuracy allowable threshold G3 is calculated similarly.

[0070] In one implementation, both regression sub-models employ a multilayer perceptron structure to support nonlinear mapping. The input layer dimension is consistent with the G1 dimension. Two hidden layers are set, with the number of neurons in the first hidden layer denoted as G4 and the number of neurons in the second hidden layer denoted as G5. The output layers output G2 and G3 respectively, both using linear activation functions to ensure continuous and interpretable outputs. The training process includes sample normalization, parameter initialization, forward propagation, loss calculation, and backpropagation update. Sample normalization maps each input component to the [0, 1] interval to avoid training instability caused by differences in dimensions. The loss function adopts the mean squared error form, with the noise regression loss denoted as G6 and the vibration regression loss denoted as G7. The two are weighted and summed to obtain the total loss G8. The weight coefficient is denoted as G9, used to adjust the importance ratio of noise and vibration prediction. The training uses stochastic gradient descent or adaptive optimization algorithm to update parameters. The learning rate is denoted as G10, the number of training epochs is denoted as G11, and the batch size is denoted as G12. For example, G4 is set to 32, G5 to 16, G9 to 0.5, G10 to 0.001, G11 to 50, and G12 to 128, thereby controlling the model size to adapt to the update cycle of the road and bridge monitoring system while ensuring training convergence. Through the above model structure and training strategy, a precise nonlinear mapping relationship can be established between meteorological conditions, traffic flow, allowable noise intensity, and allowable thresholds for vibration waveform accuracy. This solves the technical problem of static threshold failure caused by the drift of noise and vibration stability baselines with environmental changes in road and bridge scenarios.

[0071] The phrase "establishing a nonlinear mapping relationship between meteorological conditions, traffic flow, noise allowable intensity, and vibration waveform accuracy allowable threshold" should be understood as establishing a comprehensive mapping relationship of "environmental parameters → upper bound of response statistics → dynamic allowable boundary". The first part is obtained by learning from historical normal samples, and the second part is determined by safety margin correction and standard boundary constraints.

[0072] In other words, the allowable thresholds for noise intensity and vibration waveform accuracy are not simply the result of calling fixed regulatory limits, nor are they the result of direct average prediction of historical observations. Instead, they are dynamic allowable boundaries determined by combining historical statistical distribution, environmental influencing factors, and safety margins.

[0073] Furthermore, the allowable thresholds for noise intensity and vibration waveform accuracy are not simply either "empirical boundaries" or "standard boundaries," but rather composite judgment boundaries obtained by the convergence of the environmental response statistical boundary and the standard safety boundary.

[0074] After the environmental parameter fusion model is trained and deployed, edge computing nodes input real-time meteorological condition data and traffic flow density data into the model during the online operation phase. This yields permissible thresholds for noise intensity and vibration waveform accuracy based on the current environment. The online input meteorological condition data is collected in real-time by meteorological sensors around the road and bridge, and the traffic flow density data is collected in real-time by vehicle detection sensors. These two data points are aligned with timestamps to form a data structure consistent with the model input. Then, the edge computing nodes call the model inference interface to output the permissible thresholds for noise intensity and vibration waveform accuracy, which are saved locally as dynamic judgment benchmarks. Through this online inference method, the dynamic judgment benchmarks are updated in real-time with the environment, solving the problem of not being able to adjust the judgment criteria in a timely manner when changes in traffic flow peaks and valleys or wind speeds cause changes in noise propagation conditions.

[0075] During the online operation phase, the results obtained by the edge computing node calling the model inference interface are essentially corresponding to the instantaneous estimate of the dynamic allowable boundary under the current environmental conditions. This instantaneous estimate is not an abstract threshold detached from the scenario, but a judgment benchmark driven by environmental parameters, constrained by statistical laws, and corrected by standardized boundaries.

[0076] Noise signal intensity is extracted and calculated from the classified noise feature set. The calculation method is to perform energy statistics on the effective sound pressure value of the noise time-domain signal within the current time window and convert it into a decibel value, or to obtain an equivalent intensity value by weighted summation of the energy of each frequency band in the frequency domain. This intensity value is consistent with the noise supervision label calculation caliber used in the model training phase. Vibration waveform accuracy features are extracted and calculated from the classified vibration feature set. Within a vibration window of length H1, the edge computing nodes calculate the dominant frequency drift H2 (the absolute value of the difference between the current dominant frequency and the historical dominant frequency (within a preset time)) and amplitude fluctuation rate H3 (the ratio of the standard deviation of the root mean square of vibration within the window to the mean, dimensionless). The two are then combined according to weights to obtain vibration waveform accuracy feature H4 (the weights are consistent with the vibration supervision label calculation weights during model training). H2 needs to be normalized first (e.g., divided by a reference frequency value, such as the median of the historical dominant frequency of the monitoring point, the center frequency of the frequency band of interest in the structural design, or a fixed frequency value preset according to the sensor range) to make it dimensionless. Then, it is weighted and summed with H3 to obtain H4. The weighting coefficients of the weighted summation are dimensionless constants, which can be preset through experimental calibration or engineering experience, and are consistent with the weights calculated by the vibration supervision labels during model training. Ensure that H4 completely matches the physical meaning and calculation method of the allowable threshold for the accuracy of the vibration waveform output by the model. By constructing vibration waveform accuracy features instead of just using the vibration peak value, it is possible to characterize whether the vibration waveform has unsteady changes, solving the problem of missed detection caused by changes in waveform shape when the road and bridge structure undergoes abnormal excitation or structural state changes, but the peak value may not immediately exceed the limit.

[0077] The current noise signal intensity is compared with the allowable noise intensity, and the vibration waveform accuracy feature is compared with the allowable vibration waveform accuracy threshold. The allowable noise intensity represents the acceptable upper limit of noise intensity under current weather and traffic conditions, while the allowable vibration waveform accuracy threshold represents the safety boundary of the road and bridge structure's vibration stability under current conditions (exceeding this threshold indicates non-steady-state changes in structural vibration, which may affect structural safety or the operational stability of noise reduction devices). Preferably, to avoid misinterpreting the model output as a pure statistical mean, both the "acceptable upper limit" and the "safety boundary" correspond to the allowable boundaries dynamically calculated by the environmental parameter fusion model and corrected for safety margins. Their determination criteria reflect both the normal range of variation under current environmental conditions and the constraints on noise control and structural safety in engineering applications. The comparison process employs a dual-condition triggering logic to reduce false alarms. If the current noise signal intensity exceeds G2 and the vibration waveform accuracy feature exceeds G3, an abnormal warning response mechanism is triggered to generate an alarm signal. The alarm signal includes the alarm type, timestamp, location identifier, and the noise intensity and vibration waveform accuracy feature values ​​at the time of triggering. The dual-condition triggering logic is used to solve the problem that noise exceeding the standard may be caused by occasional external sound sources without requiring high-risk treatment of noise reduction devices, and that vibration abnormality may be caused by structural maintenance work or short-term impact without requiring judgment as a high-risk noise scenario. In this way, the alarm target is limited to the coupled risk state of noise abnormality and vibration abnormality at the same time, thereby improving the directionality of the alarm.

[0078] In the above-described judgment logic, the current noise signal intensity and vibration waveform accuracy characteristics are compared with their corresponding dynamic allowable boundaries. Essentially, this compares the "real-time observation state" with the "allowable upper bound of the response under the current environmental conditions," rather than mechanically comparing the real-time observation state with a fixed statistical mean or a fixed regulatory threshold. Therefore, model output, regulatory boundaries, and historical statistical patterns can be uniformly incorporated into the same judgment framework.

[0079] In one embodiment, when the current noise signal intensity or vibration waveform accuracy characteristics reach a preset extreme over-limit level, a high-level alarm can also be directly triggered to avoid response delay caused by simultaneous judgment of two conditions in extreme risk scenarios.

[0080] After an alarm signal is generated, the edge computing node labels the current environmental state based on the alarm signal and the environmental parameters at the time of triggering to determine a high-risk noise scene label. The environmental parameters include at least the wind speed, humidity, temperature, and traffic flow density at the time of triggering, and can be combined with the alarm type to form a label field. The label adopts a structured format of "environmental condition-risk type". For example, when the traffic flow density exceeds a preset high flow threshold and the wind speed exceeds a preset wind noise threshold, the label can be set to "high traffic flow-high wind speed coupled noise risk". When the traffic flow is within the normal range but the humidity exceeds a preset humidity threshold and the noise intensity exceeds the allowable intensity, the label can be set to "high humidity propagation enhanced noise risk".

[0081] By associating alarm events with environmental parameters and forming high-risk noise scene labels, the problem that alarm signals alone cannot characterize the triggering background conditions and lead to a lack of basis for subsequent noise reduction device status verification and control strategy selection is solved. This enables subsequent steps to select different device verification cycles, different parameter adjustment strategies, or different data reporting strategies based on the label type.

[0082] S4. Based on high-risk scenario labels, perform real-time verification of the noise reduction device status and optimize data storage to generate an equipment status report.

[0083] In one specific embodiment, the process of performing step S4 may specifically include the following steps:

[0084] Based on the high-risk scenario labels, initiate a real-time verification process based on a preset monitoring cycle;

[0085] According to the preset equipment status self-test frequency, the key components of the noise reduction device are scanned and evaluated to obtain the operating parameters of the noise reduction device.

[0086] Vibration data was collected through road and bridge sensors during the verification process, and the vibration data was processed using a preset compression algorithm to reduce the data storage volume.

[0087] By employing a cloud-based interaction minimization strategy, the operating parameters and compressed vibration data are uploaded to the backend server.

[0088] The operating parameters and compressed vibration data are integrated to generate an equipment status report. If the equipment status report shows an abnormality, the abnormal time and corresponding parameters are recorded.

[0089] Specifically, after receiving a high-risk scenario label, the edge computing node writes the label into the event queue and initiates a real-time verification process based on a preset monitoring cycle. This solves the technical problem of invalid computing power occupation and invalid data accumulation caused by the continuous high-frequency self-check of road and bridge noise reduction devices under normal conditions, and links the frequency of equipment verification with the risk level, ensuring higher state observability when noise and vibration coupling is abnormal.

[0090] The real-time verification process based on a preset monitoring cycle includes parameters for the verification start time, verification duration, and verification interval. The verification start time is the time when the high-risk scenario label is generated. The verification duration is denoted as J1 minutes, and the verification interval is denoted as J2 seconds. J1 and J2 are set according to the label type. For example, when the label type is "high traffic flow-high wind speed coupled noise risk," J1 is set to 30 minutes and J2 is set to 2 seconds to intensively collect equipment status during the risk duration. When the label type is "sudden noise event accompanied by abnormal vibration," J1 is set to 10 minutes and J2 is set to 1 second to capture changes in device status during short-term strong disturbances. The edge computing node creates a verification task based on the above parameters and issues a verification command to the noise reduction device control unit. The verification command carries the task number, verification cycle parameters, and a list of target components, thus forming a repeatable verification scheduling mechanism in the road and bridge site. This method of configuring the verification cycle driven by scenario labels solves the problem of insufficient monitoring during high-risk periods or wasted resources during low-risk periods caused by using a fixed verification cycle in road and bridge scenarios with varying risk intensities. After the verification process is initiated, the edge computing nodes periodically perform status scans and evaluations of the key components of the noise reduction device at a verification interval of J2 seconds. Within each scan cycle, the operating parameters of each component are sampled multiple times at a self-test frequency of J3 Hz, and the statistical values ​​of the sampled values ​​(such as mean and instantaneous maximum) are recorded as the operating parameters for that cycle. The key components include at least an anti-phase acoustic wave generator module, a power amplifier module, a power supply and energy storage module, an angle actuator module, an angle position sensing module, a communication module, and a heat dissipation module, with each type of component corresponding to a set of quantifiable operating parameters. The operating parameters of the anti-phase acoustic wave generator module include output frequency setting, output phase setting, and output amplitude setting; the operating parameters of the power amplifier module include output current, output voltage, and temperature rise; the operating parameters of the power supply and energy storage module include input voltage, battery percentage, and instantaneous power consumption; the operating parameters of the angle actuator module include target angle, actual angle, angle deviation, and motor drive current; the operating parameters of the communication module include link signal-to-noise ratio, packet loss rate, and retransmission count; and the operating parameters of the heat dissipation module include cooling fan speed and casing temperature. The self-test frequency J3 is used to control the sampling density within a single scan cycle, ensuring that the acquired parameters reflect the device status at that moment. At the end of each scan cycle, the edge computing node writes the processed parameters into the operating parameter cache according to component number, timestamp, and location identifier. By performing status scanning and evaluation on key components and outputting quantifiable operating parameters, the problem of road and bridge noise reduction devices only monitoring external noise in high-risk noise scenarios and being unable to determine whether the device is in an abnormal operating state is solved, providing device-side basis for subsequent adaptive control.

[0091] While performing the status scan, the verification process collects vibration data through road and bridge sensors and compresses the vibration data to reduce data storage volume. The sensor locations for collecting vibration data are consistent with step S1, focusing on key locations on the bridge deck and piers. The vibration data compression process uses a preset compression algorithm and is executed at the edge computing node. Taking wavelet transform compression as an example, the edge computing node performs discrete wavelet decomposition on the vibration time-domain sequence of length K1 to obtain a multi-scale coefficient set, retains the coefficients with the highest absolute amplitude values ​​(K2), and sets the remaining coefficients to zero before encoding to form a compressed package. K2 is less than K1 to control the compression ratio. For example, if K1 is 2048 points and K2 is 256 points, the retention ratio is 12.5%. Run-length encoding and quantization encoding are used for the retained coefficients to reduce storage. Quantization encoding uses a step size K3 for uniform quantization. For example, if K3 is 0.001, each coefficient is quantized to an integer and stored in a variable-length encoding manner. By performing wavelet compression on the vibration data at the edge and outputting the compressed vibration data, the problem of excessively large volume of original vibration data generated by high-frequency sampling in road and bridge monitoring, leading to local storage overflow or excessive communication bandwidth consumption, is solved. Furthermore, the main energy coefficients are preserved so that the compressed data can still reflect the abnormal vibration characteristics.

[0092] After obtaining the operating parameter cache data and compressed vibration data, the necessary data is uploaded to the backend server through a cloud interaction minimization strategy. This strategy includes upload trigger conditions, data selection rules, and upload frequency control rules. The upload trigger condition is the persistent presence of a high-risk scenario label or an abnormal flag appearing in the equipment status assessment. The data selection rules limit the upload to only key operating parameters and compressed vibration data, excluding the full original waveform. Key operating parameters include at least the output frequency setpoint, angle deviation, power supply, and temperature rise. The upload frequency control rules restrict uploads to batches at preset times within the verification period to avoid frequent handshakes. This cloud interaction minimization strategy solves the problem of unstable real-time monitoring data uploads due to limited network bandwidth or fluctuating communication quality at road and bridge sites, and reduces the cloud storage pressure caused by continuously uploading large amounts of data during high-risk periods.

[0093] After receiving the uploaded frame, the backend server integrates the operating parameters and compressed vibration data to generate a device status report. The integration process includes data unpacking, timestamp alignment, component-level aggregation, and anomaly detection field generation. The operating parameters in each uploaded frame are categorized by component and arranged chronologically. The compressed vibration data is decoded into a vibration coefficient sequence or a reconstructed approximate vibration waveform that can be used for feature calculation. Then, the operating parameter sequence and vibration sequence within the same verification period are associated and stored using the verification task number as an index, forming the structured fields of the device status report. The device status report includes at least the verification start and end times, a summary of the trigger tag content, statistical values ​​of the operating parameters of each key component, and a summary index of the compressed vibration data. The operating parameter statistics include the maximum, minimum, average, and number of times each parameter exceeds the threshold. The compressed vibration data summary index includes the vibration energy change rate, the dominant frequency drift, and the number of times abnormal peaks occur. The anomaly detection field is obtained by comparing the operating parameter statistics with preset threshold rules. For example, when the shell temperature exceeds the upper temperature limit or the angle deviation continuously exceeds the upper angle deviation limit for a preset number of times, the device status report is marked as abnormal. After the device status report is generated, it is stored in the backend database. If the device status report shows an anomaly, the backend server records the anomaly timestamp and the corresponding set of parameter values ​​in the report and writes the anomaly event to the alarm log at the same time.

[0094] By generating equipment status reports that include operating parameters and vibration summaries, the problem of the lack of a unified status profile for road and bridge noise reduction devices under complex working conditions, which makes it impossible to judge the reliability of the device operation, is solved. Furthermore, by recording abnormal times and corresponding parameters, anomalies can be traced, enabling subsequent adaptive control to make targeted adjustments based on specific abnormal indicators rather than relying on blind adjustments triggered by a single noise indicator. This improves the stable operation capability of the noise reduction device in high-risk scenarios and reduces the risk of misadjustment.

[0095] S5. Dynamically adjust the control parameters of the noise reduction device according to the equipment status report, generate an adaptive control configuration based on the adjustment results, and regulate the working status of the noise reduction device in real time through the adaptive control configuration.

[0096] In one specific embodiment, in step S5, the control parameters of the noise reduction device are dynamically adjusted according to the equipment status report, and an adaptive control configuration is generated based on the adjustment result. This includes: analyzing the current noise suppression effect feedback based on the real-time collected current environmental noise signal; if the analysis shows that the current environmental noise signal intensity has not dropped to the preset target range, an adjustment mechanism is triggered to adjust the output frequency of the anti-phase sound wave generator in the noise reduction device; simultaneously, the angle parameters of the sound barrier in the noise reduction device are adjusted based on the vibration characteristics in the equipment status report and the current environmental data; an adaptive noise reduction control configuration is generated based on the adjusted anti-phase sound wave output frequency and the barrier angle parameters; the adaptive noise reduction control configuration is matched and verified with the current environmental data, and if the verification result shows that the parameter adjustment meets the expected noise control target, the adaptive noise reduction control configuration is saved as the current optimal configuration.

[0097] Specifically, this step is achieved collaboratively by the edge computing node at the road and bridge site, the noise reduction device control unit, and the noise reduction device actuator. The edge computing node reads key operating parameters and abnormal flag fields from the equipment status report, and obtains the current environmental noise signal, traffic flow density data, meteorological condition data, and road and bridge structure vibration data from the sensor network during the control phase. This forms an input data set for dynamic adjustment under the same time reference, solving the technical problem in road and bridge scenarios where parameter adjustment based solely on a single noise index leads to an inability to perceive the device's own working status, resulting in invalid or overloaded parameter adjustment.

[0098] The noise suppression effect feedback, based on real-time acquired ambient noise signals, is achieved through equivalent sound pressure level (SPL) calculation performed at the edge computing node. Within a sliding window of length P1 seconds, the edge computing node performs energy statistics on the time-domain sampled values ​​output by the noise sensor and converts them to decibels (dB) to obtain the current noise signal intensity index P2. Simultaneously, this index is compared with a preset target range, which consists of an upper target limit P3 and a lower target limit P4, consistent with road and bridge noise control standards or on-site control targets. For example, the root mean square (RMS) sound pressure level of the noise sampled values ​​within the window is calculated and converted to an equivalent SPL P2. If P2 is greater than P3, it is determined that the noise suppression effect has not reached the target range. By calculating the equivalent SPL within a fixed time window instead of using single-point instantaneous values, the problem of feedback signal jitter caused by instantaneous vehicle horns or short-term sudden noises in road and bridge scenarios, triggering frequent parameter adjustments, is solved, thus ensuring that the parameter adjustment triggering logic is consistent with the continuous noise control target.

[0099] When the ambient noise signal intensity fails to drop below the preset target range, an adjustment mechanism is triggered, and a frequency adjustment command is sent to the noise reduction device control unit to adjust the output frequency of the anti-phase acoustic wave generator. The frequency adjustment command includes the target frequency value, phase compensation options, and execution time. The target frequency value is calculated based on the noise signal's spectral characteristics and the anti-phase acoustic wave control principle. The edge computing node performs a discrete Fourier transform on the current noise signal within a spectral analysis window of length P5 seconds to obtain the amplitude spectrum. It then searches for the main noise frequency P6 corresponding to the energy peak within the preset frequency band. The output frequency of the anti-phase acoustic wave generator is then set to the frequency corresponding to P6 to achieve anti-phase suppression targeting the main noise frequency band. For example, if P5 is 2 seconds, and an energy peak is detected at 450Hz within the 200Hz to 1000Hz range, the output frequency is set to 450Hz. To suppress frequent adjustments caused by frequency drift, edge computing nodes employ a step-limiting rule for frequency updates, with an upper limit denoted as P7. For example, P7 is set to 20Hz. Therefore, if the output frequency of the previous control cycle was 440Hz and the current main noise frequency is 480Hz, the target frequency is updated to 460Hz. To compensate for sound propagation delay and execution link delay, the frequency adjustment command also carries a phase compensation amount P8. P8 is estimated based on the sound wave propagation path length and sound velocity. The sound velocity is taken as the sound velocity model value under the current temperature conditions, and the propagation path length is calculated from the sensor position and the relative position of the sound source and the noise reduction device. By adjusting the anti-phase sound wave frequency based on the main noise frequency and setting step limits and phase compensation, the problem of the inability to continuously suppress fixed-frequency anti-phase sound waves due to changes in the noise spectrum with vehicle speed in road and bridge scenarios is solved. Simultaneously, frequent large-scale parameter adjustments that could cause device overload or control instability are avoided.

[0100] While adjusting the output frequency of the anti-phase acoustic wave generator, the angle parameters of the sound barrier are simultaneously adjusted based on the vibration characteristics in the equipment status report and current environmental data. The sound barrier angle parameters are used to change the blocking and reflection direction of the sound wave propagation path to enhance the attenuation effect on the target area, and to reduce the additional load on the structure when structural vibration is abnormal or to prevent the barrier mechanism from entering a resonance state during vibration amplification. The vibration characteristics in the equipment status report include at least the vibration energy change rate, dominant frequency drift, and number of abnormal peaks within the current inspection period. These vibration characteristics are combined with environmental data such as current traffic flow density and wind speed to determine the angle adjustment amount. The angle adjustment amount can be determined using a rule mapping method. The edge computing nodes pre-store an angle adjustment rule table, which takes traffic flow density level, wind speed level, and vibration status markers as input and outputs the target angle value. For example, when traffic density is high, wind speed is medium, and vibration is normal, the target angle is set to 30 degrees with the road normal to enhance shielding against lateral noise from the lanes. Alternatively, if the vibration is marked as abnormal and the number of abnormal peaks exceeds a preset threshold, the target angle is adjusted to 15 degrees to reduce the wind-exposed area of ​​the barrier mechanism and decrease the load on the actuator. Angle control also incorporates a rate-of-change constraint, setting an upper limit for the angle change rate to prevent rapid swaying of the barrier mechanism that could lead to mechanical impact when bridge vibration is high. By linking angle adjustment to vibration characteristics, the problem of solely pursuing acoustic shielding effects while ignoring structural vibration in road and bridge scenarios, which could lead to overload of the barrier actuator or positioning deviations under abnormal vibration, is solved. This ensures that the control process balances noise suppression and device operational safety.

[0101] After obtaining the adjusted anti-phase acoustic wave output frequency parameters and barrier angle parameters, an adaptive noise reduction control configuration is generated based on the adjustment results. The adaptive control configuration is stored in a structured manner as a set of control parameters that can be issued, including a configuration number, applicable scenario label, parameter fields, and validity period field. The parameter fields include at least the anti-phase acoustic wave output frequency setpoint, phase compensation amount, output amplitude setpoint, barrier angle target value, upper limit of angle change rate, and execution timing flag. For example, the configuration number can be generated by combining a timestamp and a scenario label hash; the applicable scenario label is taken from the high-risk scenario label type field output in step S3; and the validity period field is set in minutes to adapt to changes in traffic flow and weather. By generating the adaptive control configuration in a structured manner, the problem of untraceable and difficult-to-reproduce parameters caused by the reliance on instantaneous control commands in the parameter adjustment process of road and bridge noise reduction devices is solved, and a comparable object is provided for subsequent configuration verification and optimal configuration storage.

[0102] The adaptive control configuration is matched and verified against current environmental data to determine whether the parameter adjustments meet the expected noise control target. The matching verification employs a short-time closed-loop verification method. The configuration is sent to the noise reduction device control unit, and the changes in noise signal intensity (P11) and device operating parameters (P12) are monitored within a verification window of length P10 seconds. The noise signal intensity change (P11) is defined as the difference between the average noise intensity after adjustment and the average noise intensity before adjustment within the verification window. A preset reduction threshold is defined as a positive value (P13, dB), indicating an expected reduction of at least P13 dB. The noise reduction is considered to meet the expectation if and only if P11 ≤ -P13. The device operating parameter change (P12) is used to verify that the adjustment did not trigger any device abnormalities. P12 may include the increase in power amplifier module temperature rise, actuator current increase, and angle deviation increase, which are compared with the safety thresholds in the equipment status report. If any increment exceeds the allowable range, the configuration is deemed to have failed verification. By simultaneously verifying noise intensity reduction and device operation safety constraints, the problem of device failure caused by simply pursuing noise reduction in road and bridge scenarios may be solved, which may lead to power amplifier overheating, actuator stalling, or power supply overload. This makes the adaptive control configuration both usable and safe.

[0103] When the verification results show that the parameter adjustment meets the expected noise control target and the device operating parameters do not trigger abnormal constraints, the adaptive control configuration is saved as the current optimal configuration. The saving process writes the configuration number, scene label, noise intensity change before and after adjustment, vibration characteristic summary during adjustment, and corresponding meteorological and traffic flow environmental parameters into the configuration library and creates an index. This allows for direct retrieval of the verified optimal configuration when similar or identical environmental conditions occur later, eliminating the need to search for parameters from scratch. To achieve matching of similar environmental conditions, the configuration library index can be divided into traffic flow density and wind speed ranges, storing multiple candidate configurations in each range. When retrieved, the current traffic flow density and wind speed are mapped to the corresponding range, and the configuration with the best historical noise reduction effect and no abnormalities is selected as the priority configuration. By saving the optimal configuration and supporting reuse under similar conditions, the problem of response lag and increased energy consumption caused by repeated parameter adjustments when environmental conditions recur is solved. Simultaneously, the stability and predictability of real-time control are improved, and the adaptive control configuration can directly drive the noise reduction device into a noise reduction mode dynamically matched to the current environment.

[0104] In one specific embodiment, S5 involves adjusting the operating state of the noise reduction device in real time through adaptive control configuration, including:

[0105] An adaptive noise reduction control configuration is deployed to the noise reduction device to control the device to perform adjustment operations; based on the local data storage capacity, newly acquired environmental noise and vibration data after adjustment are saved; and the environmental fluctuations after adjustment are continuously monitored through a real-time alarm triggering mechanism.

[0106] If environmental fluctuations are detected to exceed the preset fluctuation range, a fluctuation record will be generated and a secondary alarm will be triggered.

[0107] The adjustment coefficient is calculated based on the fluctuation record, and the output frequency of the anti-phase sound wave of the noise reduction device and the barrier angle parameter are corrected according to the adjustment coefficient. Through continuous monitoring and parameter adjustment, the dynamic matching environmental noise suppression result is obtained.

[0108] Specifically, the adaptive control configuration for real-time adjustment of the noise reduction device's operating status is achieved collaboratively by edge computing nodes, the noise reduction device control unit, the anti-phase acoustic wave generator, the sound barrier actuator, and the sensor network. The edge computing node encapsulates the adaptive noise reduction control configuration into a downloadable configuration message and sends it to the noise reduction device control unit via a fieldbus or industrial wireless link, thereby enabling the deployment of the adaptive noise reduction control configuration to the noise reduction device and controlling the noise reduction device to perform adjustment operations.

[0109] The configuration deployment includes three stages: configuration consistency verification, parameter writing, and execution activation. Upon receiving the configuration message, the noise reduction device control unit verifies the configuration number and timestamp to prevent older configurations from overwriting new ones. It also verifies whether parameter fields are within the device's allowable range. For example, it checks if the anti-phase acoustic wave output frequency setting is within the generator's output frequency band, if the barrier angle target value is within the actuator's mechanical travel range, and if the upper limit of the angle change rate is less than the actuator's maximum allowable angular velocity. After successful verification, the anti-phase acoustic wave output frequency setting is written to the generator's frequency register, and the phase compensation amount is written to the phase control register. Simultaneously, the barrier angle target value and the upper limit of the angle change rate are written to the angle controller's target register and speed limit register. Execution is initiated at the activation time indicated by the execution timing parameters. By verifying configuration consistency and parameter range, the system resolves the problem of abnormal device operation caused by configuration disorder due to communication delays or message loss at the road and bridge site, and avoids generator distortion or actuator limit collisions caused by parameters exceeding limits.

[0110] After the noise reduction device enters the control state, the sensor network continuously collects newly generated environmental noise and vibration data after the control, and the edge computing nodes save the newly collected data according to their local data storage capacity. The local storage capacity is determined by the available storage space of the edge nodes and the preset saving strategy. The saving strategy is implemented in a hierarchical caching manner, with a high-priority cache area for storing data segments directly related to the control effect, and a low-priority cache area for storing statistical summary data. A coverage strategy is triggered when the available storage space is lower than the threshold L1. The data stored in the high-priority cache area includes the noise equivalent sound pressure level sequence and the vibration root mean square sequence within each verification window, while the data stored in the low-priority cache area includes statistics such as the noise mean, noise peak, vibration mean, and vibration peak per minute. The coverage strategy is implemented using a circular queue and covers the oldest data in chronological order. For example, if the original sampling rate of noise and vibration is high, causing the usage of the high-priority cache area to approach its threshold, only the most recent T1 minutes of original data are retained, and the earlier data are converted into statistical summaries and transferred to the low-priority cache area. By dynamically selecting the storage granularity based on storage capacity, the problem of edge node storage exhaustion and inability to continuously monitor is solved due to the large amount of raw data generated during long-term operation of road and bridge sites. It also ensures that traceable data exists during critical control periods for subsequent evaluation and fault location.

[0111] While data is being stored, a real-time alarm triggering mechanism continuously monitors environmental fluctuations after regulation. This mechanism utilizes both noise and vibration fluctuation indices and employs a sliding window for judgment. The noise fluctuation index is denoted as L2, defined as the difference between the maximum and minimum equivalent sound pressure levels within the sliding window. The vibration fluctuation index is denoted as L3, defined as the difference between the maximum and minimum root mean square vibration values ​​within the sliding window. The sliding window length is denoted as L4 seconds. The preset fluctuation range is defined by the upper limit of noise fluctuation L5 and the upper limit of vibration fluctuation L6. When L2 exceeds L5 or L3 exceeds L6, the environmental fluctuation is determined to exceed the preset range. By using a sliding window fluctuation index instead of a single-point threshold, the problem of false triggering caused by individual instantaneous spikes in road and bridge scenarios is solved, and the fluctuation judgment becomes more sensitive to continuous fluctuations, thus corresponding to the regulation stability target.

[0112] When environmental fluctuations exceed the preset fluctuation range, a fluctuation record is generated and a secondary alarm is triggered. The fluctuation record includes a fluctuation occurrence timestamp, fluctuation duration, fluctuation type field, fluctuation amplitude field, and control configuration number field. The fluctuation type field indicates whether noise fluctuations, vibration fluctuations, or both exceed the limit. The fluctuation amplitude field records the actual values ​​of L2 and L3 and the corresponding noise peak and vibration peak values ​​within the window. The secondary alarm is implemented by encapsulating the fluctuation record into an alarm message and sending it to the backend server and the field alarm terminal. The alarm message carries the triggered configuration number to associate with the current control strategy. By generating fluctuation records containing configuration numbers and triggering secondary alarms, the problem of the road and bridge noise reduction device becoming inapplicable due to sudden changes in environmental conditions during control, but lacking a fast feedback channel, is solved. This allows the backend and field to identify the "instability during control" state and prepare to enter the parameter correction process.

[0113] After the secondary alarm is triggered, the adjustment coefficient is calculated based on the fluctuation record. The frequency of the anti-phase acoustic wave output and the barrier angle parameters of the noise reduction device are then corrected according to the adjustment coefficient. The adjustment coefficient maps the fluctuation amplitude to a parameter correction amount and is constrained by parameter changes. The frequency adjustment coefficient is denoted as L7, the angle adjustment coefficient as L8, the frequency correction amount as L9, and the angle correction amount as L10. The frequency adjustment coefficient L7 is calculated by the ratio of the noise fluctuation amplitude L2 to the preset noise fluctuation upper limit L5, and is limited to the interval [0, 2] to prevent excessive amplification. The angle adjustment coefficient L8 is calculated by the ratio of the vibration fluctuation amplitude L3 to the preset vibration fluctuation upper limit L6, and is limited to the interval [0, 2]. The frequency correction amount L9 is calculated by L9 = L7 × L11, where L11 is the maximum frequency correction step (e.g., 15Hz). The angle correction amount L10 is calculated by L10 = L8 × L12, where L12 is the maximum angle correction step (e.g., 3 degrees). When determining the correction direction, the peak shift of the spectrum and the trend of the main noise frequency change in the fluctuation record are used. If the fluctuation record shows that the main noise frequency is drifting to higher frequencies, the output frequency is increased by L9; if the main noise frequency is drifting to lower frequencies, the output frequency is decreased by L9. When determining the angle correction direction, the wind speed direction and the direction of the sound source are estimated. If the noise peak comes from outside the road and the wind direction blows the noise towards the sensitive area, the barrier angle is increased by L10 in the blocking direction; if the vibration fluctuation exceeds the limit and the load on the actuator increases, the barrier angle is decreased by L10 in the load reduction direction. By calculating the adjustment coefficient based on the fluctuation amplitude and mapping it to the limited frequency and angle correction amount, the problem of insufficient response of fixed step size parameter tuning or control oscillation caused by excessive parameter tuning when the environment changes suddenly in the road and bridge scenario is solved. The parameter correction is proportional to the fluctuation intensity, thereby achieving smoother online fine-tuning.

[0114] After completing the frequency and angle corrections, the corrected parameters are written into the new online configuration and immediately deployed to the noise reduction device control unit. The edge computing node continues to monitor the corrected fluctuation indicators L2 and L3 through a real-time alarm triggering mechanism. When L2 does not exceed L5 and L3 does not exceed L6 within a consecutive sliding window of R1, the control result of this segment is marked as the dynamically matched environmental noise suppression result (the final configuration parameters used during the stable control period and their corresponding average noise suppression levels are recorded together as the 'dynamically matched environmental noise suppression result' in the current environment). R1 ​​is the stable window counting threshold.

[0115] By continuously monitoring and adjusting parameters to form an online closed loop, the noise reduction device can maintain its noise suppression output within the target range even under conditions of sudden changes in traffic density, sudden increases in wind speed, or changes in road and bridge vibration. This also solves the problem of rapid degradation of the effect after one-time configuration and deployment due to the highly dynamic road and bridge environment, thus obtaining environmental noise suppression results that are dynamically matched to the current environment.

[0116] S6. After regulation, continuously monitor environmental fluctuations and analyze the noise suppression effect. If the effect does not meet the preset target, generate optimization suggestions and iteratively adjust the noise reduction device operation strategy until the noise suppression effect meets the requirements.

[0117] In one specific embodiment, the process of performing step S6 may specifically include the following steps:

[0118] The system collects adjusted environmental data in real time through a sensor network. This data includes at least traffic flow density data, road and bridge structure vibration data, and environmental noise signals. The currently collected environmental data is compared and analyzed with historical environmental data for the corresponding scenario to determine whether the noise suppression effect has reached the preset target. If the noise suppression effect has not reached the preset target, the system records the current environmental data and the operating parameters of the noise reduction device to generate an effect evaluation record. Based on the effect evaluation record, the system generates optimization suggestions for the noise reduction device's operating strategy and adjusts the control parameters of the noise reduction device according to the optimization suggestions. Through multiple iterative adjustments, the system obtains the optimal noise suppression result that satisfies the preset target and stores the optimal noise suppression result in association with environmental fluctuation data.

[0119] Specifically, this step involves continuous monitoring and effect evaluation of the regulated environmental state. When the noise suppression effect fails to meet the preset target, closed-loop optimization is achieved by iteratively adjusting the noise reduction device's operating strategy, thereby solving the problem of the single regulation effect decaying over time due to continuous changes in environmental conditions in road and bridge scenarios.

[0120] In this embodiment, the adjusted environmental data is collected in real time through a sensor network. The edge computing node synchronously collects the environmental data at a preset collection period and forms an environmental data frame. Each environmental data frame contains the traffic flow density value, vibration characteristic value or vibration sequence summary, and noise equivalent sound pressure level or noise sequence summary at the same timestamp, and is associated with the currently executed adaptive control configuration number, thereby ensuring a one-to-one correspondence between the control effect evaluation and the specific operation strategy.

[0121] After collecting current environmental data, it is compared and analyzed with historical environmental data in the corresponding scenario to determine whether the noise suppression effect has reached the preset target. The corresponding scenario is determined by the high-risk scenario label and the applicable scenario field of the adaptive control configuration. Historical environmental data is selected from historical noise equivalent sound pressure level sequences and historical vibration response sequences within the same or similar traffic flow density range and meteorological condition range. The historical environmental data segments and the current environmental data frames are stored using the same index caliber. The historical environmental database can be updated regularly to include recent monitoring data that has met the target and is stable, reflecting the long-term slow changes in the road and bridge environment and structural status. The comparative analysis focuses on the noise suppression effect evaluation index, denoted as M2. M2 is calculated from the difference between the current noise equivalent sound pressure level and the historical baseline noise equivalent sound pressure level, which can be understood as the relative baseline difference. The historical baseline noise equivalent sound pressure level is taken as the statistical mean or quantile of historical data in the corresponding scenario, and a negative M2 value indicates a reduction in noise. The preset target consists of a noise reduction magnitude target M3 and a noise upper limit target M4. The noise reduction magnitude target is defined as a positive value M3 (unit: dB), indicating that the desired reduction is at least M3 dB. Therefore, the noise suppression effect is considered to have reached the preset target if and only if M2 ≤ -M3 and the current noise equivalent sound pressure level ≤ M4. By introducing a dual-target judgment of relative baseline difference and absolute upper limit, the problem of unfair evaluation caused by simply comparing absolute noise and ignoring the difference in scene baseline is solved, while avoiding the problem of only looking at the relative reduction when the actual noise still exceeds the standard.

[0122] Comparative analysis results show that when the noise suppression effect fails to meet the preset target, the edge computing node records the current environmental data and the operating parameters of the noise reduction device, and generates an effect evaluation record. The operating parameters include at least the inverse acoustic wave output frequency setting, phase compensation amount, output amplitude setting, barrier angle target value, actual barrier angle value, power amplifier module temperature rise value, power supply percentage, and communication packet loss rate. The effect evaluation record also includes the scene label at the time of failure, the current configuration number, M2 value, current equivalent sound pressure level, current root mean square vibration, and traffic flow density value. By simultaneously recording environmental input and device parameter output at the time of failure, the problem of only detecting poor performance but being unable to pinpoint the cause is solved.

[0123] Based on the effect evaluation records, optimization suggestions for the operation strategy of the noise reduction device are generated. This is achieved through rule inference and parameter correction calculation. The optimization suggestions should include at least the frequency adjustment direction, frequency adjustment amplitude, angle adjustment direction, angle adjustment amplitude, and control suggestion fields such as whether it is necessary to reduce the output amplitude or extend the control window.

[0124] Preferably, the rule set upon which the rule inference is based is pre-stored in edge computing nodes or a backend server. The rule set is jointly established by historical control experimental data calibration, simulation analysis results, and engineering experience parameters. Specifically, during the system deployment or maintenance phase, multiple sets of parameter scanning experiments are performed for different traffic flow density ranges, wind speed ranges, noise peak frequency bands, and vibration state ranges. The noise improvement, vibration change, and equipment load change are recorded under different anti-phase sound wave frequencies, output amplitudes, and barrier angles. Based on the experimental data, a mapping relationship of "environmental state - parameter adjustment amount - effect change amount" is established, and the parameter adjustment relationship that meets the preset noise reduction effect and does not trigger equipment safety constraints is solidified as rule entries. For working conditions with insufficient experimental coverage, the rule entries can be supplemented and corrected by combining sound propagation models, structural response models, or engineering experience. Further, the rule entries include at least the following fields: applicable scenario label, traffic flow density range, wind speed range, noise peak frequency range, vibration root mean square range, frequency adjustment direction, frequency correction coefficient, frequency correction upper limit, angle adjustment direction, angle correction coefficient, angle correction upper limit, and corresponding safety constraint conditions. When generating optimization suggestions, edge computing nodes retrieve matching rule entries from the rule set based on scene labels, noise spectrum characteristics, vibration characteristics, and device status in the current effect evaluation record, and generate corresponding optimization suggestions according to the matched rule entries. If multiple matching entries exist, the entry with the largest historical improvement and which has not triggered safety constraints is selected first.

[0125] The rule inference uses the error values ​​in the effect evaluation records as input. These errors include the noise excess M5 and the noise improvement deficiency M6. The noise excess M5 is the difference between the current equivalent sound pressure level and the target noise ceiling M4. When the current equivalent sound pressure level does not exceed the target noise ceiling M4, M5 is set to 0. The noise improvement deficiency M6 characterizes the degree of inadequacy of the current noise reduction relative to the target reduction. Since the noise suppression effect evaluation index M2 is the difference between the current equivalent sound pressure level and the historical baseline equivalent sound pressure level, and a negative M2 indicates a noise reduction, the current actual reduction is -M2. Therefore, the noise improvement deficiency M6 is defined as: M6 = max(0, M3 - (-M2)), equivalently, M6 = max(0, M3 + M2). For example, the frequency adjustment magnitude is obtained by mapping M5 and M6 to a frequency correction step size, with the upper limit of the frequency correction step size denoted as M7 Hz and the frequency correction coefficient denoted as M8 Hz / dB. The frequency correction amount M9 is then calculated by M9 = min(M7, M8 × max(M5, M6)). For instance, when the noise reduction target M3 is 10dB, and the current actual noise is reduced by 3dB relative to the historical baseline, M2 = -3dB. Then, M6 = max(0, 10 + (-3)) = 7dB, indicating that there is still a 7dB improvement shortfall. If M5 = 2dB at the same time, then max(M5, M6) = 7dB. For example, when M7 is 30Hz and M8 is 8Hz / dB, if max(M5, M6) = 2dB, then M9 is 16Hz; if max(M5, M6) = 7dB, then M9 = 30Hz is calculated as M9 = min(30, 8×7). The frequency correction coefficient M8 is not an arbitrary value, but is determined through a calibration process: under a preset scenario, the anti-phase sound wave frequency is scanned multiple times with a fixed step size, recording the correspondence between the frequency change and the noise improvement, and using the average frequency adjustment corresponding to a unit noise improvement as the initial frequency correction coefficient; subsequently, this coefficient is corrected by combining historical operating data to obtain a frequency correction coefficient table applicable to different scenario labels. In other words, M8 can be obtained from experimental calibration results, historical operating regression results, or a weighted fusion of both. Preferably, different M8 values ​​are set for different scenario labels to improve the specificity of rule inference.

[0126] The frequency adjustment direction is determined by the main peak offset of the noise spectrum in the effect evaluation record. The edge computing node calculates the main peak frequency M10 of the noise spectrum of the non-compliant window and compares it with the current anti-phase sound wave frequency setting value. If M10 is higher than the current setting frequency, the frequency adjustment direction is upward; if M10 is lower than the current setting frequency, the frequency adjustment direction is downward. The angle adjustment range is determined by the traffic flow density and wind speed conditions in the effect evaluation record, as well as the barrier angle deviation. The angle deviation is recorded as M11 and is equal to the absolute value of the difference between the barrier target angle and the actual barrier angle. If M11 exceeds the preset angle deviation threshold, the optimization suggestion includes a field suggesting that the actuator correct or reduce the angle change rate. If the root mean square of vibration exceeds the structural safety threshold, the optimization suggestion includes a field suggesting that the angle be adjusted in the load reduction direction.

[0127] Preferably, the angle correction coefficient and the upper limit of angle correction are determined in a similar way: under different wind speeds and traffic density conditions, the barrier angle is adjusted in stages and the noise improvement, vibration change and actuator load change are recorded. The angle adjustment that brings the greatest improvement in the overall score under the conditions of satisfying vibration constraints and actuator current constraints is selected to form an angle correction rule table.

[0128] The edge computing node adjusts the control parameters of the noise reduction device based on optimization suggestions and forms a new operating strategy. The adjustment process includes parameter out-of-bounds checks, step limits, and safety constraint verification. Step limits restrict the frequency correction amount in each iteration to no more than M7 and the angle correction amount to no more than the preset angle step size limit M12 degrees. Safety constraint verification reduces the output amplitude or extends the adjustment interval when the temperature rise of the power amplifier module exceeds the upper limit or the power supply is below the lower limit, thus avoiding blindly increasing the output under substandard conditions, which could lead to overheating or power depletion. After adjustment, the new control parameters are encapsulated into a new adaptive control configuration and sent to the noise reduction device control unit for execution. Simultaneously, the edge computing node writes the new configuration number and the iteration number into the association field of the effect evaluation record. By performing out-of-bounds checks and safety constraint verification in each iteration, the problem of potential operational risks caused by repeated parameter tuning is solved, ensuring that the iteration process is conducted within the device's tolerance range.

[0129] In one implementation, when no completely matching rule entry is found in the current scene, the edge computing node can select the closest rule entry as the initial suggestion based on scene label similarity, traffic flow density proximity, and wind speed proximity, and fine-tune the corresponding correction coefficient online in subsequent iterations based on the actual improvement effect; after the online fine-tuning result is verified to be effective in multiple consecutive evaluation cycles, it is written back to the rule base as a new scene rule.

[0130] By iteratively executing environmental data collection, effect evaluation, optimization suggestion generation, and control parameter adjustment, the operating strategy gradually approaches the noise suppression effect to meet the preset target. When the noise control index stably meets the preset target within multiple consecutive evaluation cycles, the corresponding operating strategy is marked as the optimal noise suppression result. Preferably, an upper limit is set for the number of iterations to prevent infinite looping during periods of continuous anomaly, and a maintenance prompt is triggered when no improvement is achieved in consecutive preset iterations or when the improvement is lower than the preset minimum improvement threshold. The optimal noise suppression result is selected using a comprehensive scoring method when multiple compliant configurations exist. The comprehensive score is denoted as M15, which is obtained by weighted summation of the noise reduction magnitude score, vibration constraint score, and energy consumption score. The noise reduction magnitude score is obtained by normalizing the absolute value of M2, the vibration constraint score is obtained by normalizing the margin of the vibration root mean square relative to the safety threshold, and the energy consumption score is obtained by normalizing the average power consumption or power consumption rate of the power amplifier module. The compliant configuration with the largest M15 is selected as the optimal noise suppression result. By introducing a comprehensive score for noise, vibration, and energy consumption, the problem of selecting configurations that consume too much energy or have too little vibration margin may be solved by using noise compliance as the sole criterion, making the optimal result more sustainable under long-term road and bridge operation conditions.

[0131] After determining the optimal noise suppression result, it is associated and stored with environmental fluctuation data. This is achieved by establishing an index key and data binding relationship. The index key includes at least scene label, traffic flow density range, wind speed range, temperature range, and time period type fields. Environmental fluctuation data includes noise fluctuation index sequences and vibration fluctuation index sequences monitored during the iteration process, as well as corresponding secondary alarm records. Edge computing nodes package the optimal noise suppression result record, the corresponding adaptive control configuration, and the environmental fluctuation data into a scene policy entry and upload it to the policy library of the backend server. By associating and storing the optimal result with environmental fluctuation data, the problem of response lag caused by the need for re-iteration search when similar working conditions recur is solved. This also allows subsequent environmental monitoring to use the historical optimal configuration as a reference baseline, thus forming a continuous closed loop from regulation, evaluation, iteration to knowledge solidification in the real-time monitoring method of road and bridge noise reduction devices.

[0132] To more intuitively demonstrate the technical effectiveness of the proposed real-time monitoring method for road and bridge noise reduction devices in terms of adaptive control, iterative optimization, and knowledge reuse, this application provides a technical effectiveness characterization diagram, such as... Figure 2 and Figure 3 As shown.

[0133] Figure 2The timing effect of the adaptive control process is demonstrated. As shown in the figure, during the control process, the noise level (blue curve) initially reached 72dB, went through stages such as anti-phase acoustic wave frequency adjustment, barrier angle adjustment, verification, and fine-tuning, and finally stabilized at 60dB, reaching and exceeding the preset noise target threshold (green dashed line). Meanwhile, the device operating temperature (red curve) remained within a safe range (below 55℃) throughout the entire control process, demonstrating that this application ensures device operational safety while pursuing noise reduction effects. Figure 3 The convergence process of the iterative optimization effect is illustrated in the figure. As can be seen from the figure, the noise suppression performance index (blue curve) gradually improves with the number of iterations, reaching -11 dB after 9 iterations from an initial -3 dB, exceeding the preset threshold (-10 dB). The overall score (red curve) also gradually increases with iterations, eventually stabilizing at a high level of 90 points. The key iteration points marked in the figure illustrate important nodes in the optimization process: after initial adjustment, a frequency mismatch problem was discovered, which was significantly improved by adjusting the anti-phase acoustic wave frequency; subsequently, a barrier angle deviation was discovered, which was further optimized through correction; finally, the optimal configuration was achieved in the 9th iteration. This figure verifies that the iterative optimization mechanism proposed in this application can effectively converge to the optimal noise suppression result that meets the preset target.

[0134] pass Figure 2 and Figure 3 The comprehensive demonstration verifies that the real-time monitoring method for road and bridge noise reduction devices proposed in this application can not only achieve effective noise suppression, but also significantly improve the stability of noise reduction effect and system operating efficiency while ensuring the safe operation of the device through adaptive regulation, iterative optimization and knowledge reuse mechanisms.

[0135] The above describes the real-time monitoring method for road and bridge noise reduction devices in the embodiments of this application. The following describes the real-time monitoring system for road and bridge noise reduction devices in the embodiments of this application. Please refer to [link / reference]. Figure 4 The present application provides a schematic diagram of the structure of a real-time monitoring system for road and bridge noise reduction devices. The system includes:

[0136] The data acquisition module 10 is used to acquire real-time environmental data from the sensor network deployed around the road and bridge, and to perform preliminary cleaning to obtain a preliminary environmental dataset.

[0137] The feature extraction module 20 is used to extract features from the preliminary environmental set, extract features of different frequency bands of noise signals and vibration features of road and bridge structures, and obtain a classified feature set.

[0138] The risk identification module 30 is used to generate a dynamic judgment benchmark based on real-time environmental parameters and an environmental parameter fusion model, compare the classified feature set with the dynamic judgment benchmark, detect anomalies and trigger alarms, and determine high-risk scene labels.

[0139] The status verification module 40 is used to perform real-time verification of the status of the noise reduction device based on high-risk scenario labels, optimize data storage, and generate equipment status reports.

[0140] The parameter control module 50 is used to dynamically adjust the control parameters of the noise reduction device according to the equipment status report, generate an adaptive control configuration based on the adjustment results, and regulate the working status of the noise reduction device in real time through the adaptive control configuration.

[0141] The iterative optimization module 60 is used to continuously monitor environmental fluctuations after regulation and analyze the noise suppression effect. If the effect does not meet the preset target, it generates optimization suggestions and iteratively adjusts the operation strategy of the noise reduction device until the noise suppression effect meets the requirements.

[0142] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A real-time monitoring method for a road-bridge noise reduction device, characterized in that, The method includes: S1. Obtain real-time environmental data from the sensor network deployed around the road and bridge, and perform preliminary cleaning to obtain a preliminary environmental set; S2. Extract features from the preliminary environmental set, extracting different frequency band features of the noise signal and vibration features of the road and bridge structure to obtain a classified feature set; S3. Generate a dynamic judgment benchmark based on real-time environmental parameters and environmental parameter fusion model, compare the classified feature set with the dynamic judgment benchmark, detect anomalies and trigger alarms, and determine high-risk scene labels; S4. Based on the high-risk scenario labels, the status of the noise reduction device is checked in real time and the data storage is optimized to generate a device status report; S5. Dynamically adjust the control parameters of the noise reduction device according to the device status report, generate an adaptive control configuration based on the adjustment results, and regulate the working status of the noise reduction device in real time through the adaptive control configuration. S6. After regulation, continuously monitor environmental fluctuations and analyze the noise suppression effect. If the effect does not meet the preset target, generate optimization suggestions and iteratively adjust the noise reduction device operation strategy until the noise suppression effect meets the requirements.

2. The method of claim 1, wherein, S1 includes: acquiring real-time environmental data from the sensor network according to a preset acquisition frequency, including at least traffic flow density data, road and bridge structure vibration data, and environmental noise signals; The built-in abnormal threshold dynamic adjustment mechanism filters traffic flow density data, road and bridge structure vibration data and environmental noise signals to exclude data that does not meet the preset range; the filtered data is integrated into a preliminary environmental dataset, and the collection time and sensor location information are recorded for the preliminary environmental dataset to form a complete data record. The complete data records are verified for integrity. If data is missing, the data is completed using a preset interpolation algorithm to obtain the preliminary environment set.

3. The method of claim 1, wherein S2 include: At edge computing nodes, spectral analysis is performed on environmental noise signals to extract noise features in different frequency bands; Simultaneously, waveform analysis was performed on the vibration data of the road and bridge structure to obtain vibration characteristics; Noise features are classified into continuous and burst types by threshold classification to form a noise feature set, and vibration features are classified into normal and abnormal types to form a vibration feature set. The noise feature set and the vibration feature set are standardized, and principal component analysis is performed on the standardized features to reduce dimensionality and obtain the classified feature set.

4. The method according to claim 1, characterized in that, S3 include: An environmental parameter fusion model is established based on historical datasets, which include historical meteorological condition data, historical traffic flow density data, and historical noise feature sets and historical vibration feature sets that are collected and extracted simultaneously. Real-time meteorological data and traffic flow density data are input into the environmental parameter fusion model to obtain the allowable noise intensity and allowable vibration range based on the current environment. The current noise signal intensity is extracted based on the classified noise feature set, and the current vibration waveform accuracy features are extracted based on the classified vibration feature set. The noise signal intensity and vibration waveform accuracy features are compared with the output of the environmental parameter fusion model. If the noise signal intensity exceeds the allowable noise intensity and the vibration waveform accuracy features exceed the allowable vibration range, an abnormal early warning response mechanism is triggered to generate an alarm signal. Based on the alarm signal and the environmental parameters at the time of triggering, the current environmental state is labeled to determine the high-risk noise scene label.

5. The method according to claim 1, characterized in that, S4 includes: Based on the high-risk scenario tags, initiate a real-time verification process based on a preset monitoring cycle; According to the preset equipment status self-test frequency, the key components of the noise reduction device are scanned and evaluated to obtain the operating parameters of the noise reduction device. Vibration data was collected through road and bridge sensors during the verification process, and the vibration data was processed using a preset compression algorithm to reduce the data storage volume. By employing a cloud-based interaction minimization strategy, the operating parameters and compressed vibration data are uploaded to the backend server. The operating parameters and compressed vibration data are integrated to generate an equipment status report. If the equipment status report shows an abnormality, the abnormal time and corresponding parameters are recorded.

6. The method according to claim 1, characterized in that, In S5, the control parameters of the noise reduction device are dynamically adjusted according to the equipment status report, and an adaptive control configuration is generated based on the adjustment result. This includes: analyzing the current noise suppression effect feedback based on the real-time collected current environmental noise signal; if the analysis shows that the current environmental noise signal intensity has not dropped to the preset target range, an adjustment mechanism is triggered to adjust the output frequency of the anti-phase sound wave generator in the noise reduction device; simultaneously, the angle parameters of the sound barrier in the noise reduction device are adjusted based on the vibration characteristics in the equipment status report and the current environmental data; an adaptive noise reduction control configuration is generated based on the adjusted anti-phase sound wave output frequency and the barrier angle parameters; the adaptive noise reduction control configuration is matched and verified with the current environmental data, and if the verification result shows that the parameter adjustment meets the expected noise control target, the adaptive noise reduction control configuration is saved as the current optimal configuration.

7. The method according to claim 6, characterized in that, In S5, the working state of the noise reduction device is adjusted in real time through adaptive control configuration, including: The adaptive noise reduction control configuration is deployed to the noise reduction device to control the noise reduction device to perform regulation operations; according to the local data storage capacity, the newly collected environmental noise and vibration data after regulation are saved; and the environmental fluctuations after regulation are continuously monitored through a real-time alarm triggering mechanism. If environmental fluctuations are detected to exceed the preset fluctuation range, a fluctuation record will be generated and a secondary alarm will be triggered. The adjustment coefficient is calculated based on the fluctuation record, and the anti-phase sound wave output frequency and barrier angle parameters of the noise reduction device are corrected according to the adjustment coefficient; through continuous monitoring and parameter adjustment, the dynamic matching environmental noise suppression result is obtained.

8. The method according to claim 1, characterized in that, S6 include: The adjusted environmental data is collected in real time through a sensor network. The environmental data includes at least traffic flow density data, road and bridge structure vibration data, and environmental noise signals. The currently collected environmental data is compared and analyzed with historical environmental data in the corresponding scenario to determine whether the noise suppression effect has reached the preset target. If the noise suppression effect does not reach the preset target, the current environmental data and the operating parameters of the noise reduction device are recorded to generate an effect evaluation record; Based on the effect evaluation records, optimization suggestions for the operation strategy of the noise reduction device are generated, and the control parameters of the noise reduction device are adjusted according to the optimization suggestions. Through multiple iterative adjustments, the optimal noise suppression result that satisfies the preset target is obtained, and the optimal noise suppression result is associated with and stored with environmental fluctuation data.

9. A real-time monitoring system for road and bridge noise reduction devices, used to implement the method as described in any one of claims 1 to 8, characterized in that, The system includes: The data acquisition module is used to acquire real-time environmental data from the sensor network deployed around the road and bridge, and to perform preliminary cleaning to obtain a preliminary environmental dataset. The feature extraction module is used to extract features from the preliminary environment set, extract features of different frequency bands of noise signals and vibration features of road and bridge structures, and obtain a classified feature set. The risk identification module is used to generate a dynamic judgment benchmark based on real-time environmental parameters and an environmental parameter fusion model, compare the classified feature set with the dynamic judgment benchmark, detect anomalies and trigger alarms, and determine high-risk scene labels. The status verification module is used to perform real-time verification of the status of the noise reduction device based on the high-risk scenario tags, optimize data storage, and generate a device status report. The parameter control module is used to dynamically adjust the control parameters of the noise reduction device according to the device status report, generate an adaptive control configuration based on the adjustment result, and adjust the working status of the noise reduction device in real time through the adaptive control configuration. The iterative optimization module is used to continuously monitor environmental fluctuations after regulation and analyze the noise suppression effect. If the effect does not meet the preset target, it generates optimization suggestions and iteratively adjusts the operation strategy of the noise reduction device until the noise suppression effect meets the requirements.

Images