Wind environment and structure response cooperative monitoring and early warning method and system and storage medium

Abstract

The invention provides a wind environment and structure response cooperative monitoring and early warning method and system and a storage medium, and relates to the field of structure safety monitoring and early warning. The method comprises the following steps: acquiring historical wind field data and historical structure response data, and constructing a three-dimensional feature matrix which is classified in advance and is used for a wind environment by taking a wind field type as a row index, a structural component as a column index and the historical structure response data as a matrix element value; under the condition that the current wind field type is a conventional type, searching a structural member response threshold value corresponding to the current wind field type from the three-dimensional feature matrix, and generating early warning information according to the structural member response threshold value; and under the condition that the current wind field type is an unconventional type, determining a response prediction value of the structural member based on the target prediction model, and generating early warning information based on the response prediction value. According to the scheme, a cooperative monitoring system fusing a physical mechanism and deep learning is constructed, so that the accuracy of structural safety early warning in a complex wind environment is improved.

Description

Technical Field

[0001] The present application relates to the technical field of structural safety monitoring and early warning, and specifically to a method, system and storage medium for collaborative monitoring and early warning of wind environment and structural response. Background Art

[0002] As global climate change intensifies and natural disasters become more frequent, tall structures and long-span bridges are frequently damaged by wind-induced vibrations. Understanding key parameters of a structure's natural environment and its real-time response is crucial. Analyzing this data and promptly determining structural safety is crucial for structural operation and maintenance.

[0003] Existing monitoring of environmental parameters and structural responses has been integrated into health monitoring systems, which have been widely used in long-span bridges and tall structures. Health monitoring systems monitor wind, temperature, humidity, strain, stress, displacement, cracks, acceleration, and other structural data over a long period of time. These systems then use thresholds to perform real-time assessments of these parameters, thereby diagnosing the structural condition.

[0004] On the one hand, existing health monitoring systems provide real-time assessments, making it difficult to provide early warnings of impending disasters and unable to achieve advanced foresight, effectively ensuring structural safety. On the other hand, health monitoring systems focus on numerous factors, including force, response, environment, cracks, fatigue, and more, often integrating hundreds of sensors of various types. However, they fail to address the correlations between these data points, particularly the correlations between environmental parameters and structural responses. Furthermore, the probability of structural damage occurring is extremely low, yet existing sensor devices are always operating at full capacity, resulting in significant energy waste. Furthermore, the transmission of large amounts of data places a heavy burden on network bandwidth, easily causing data transmission delays and impacting the effectiveness of real-time analysis and early warning. Summary of the Invention

[0005] Based on this, this application provides a collaborative monitoring and early warning method for wind environment and structural response, which can improve the accuracy of structural safety early warning in complex wind environments by constructing a collaborative monitoring system that integrates physical mechanisms and deep learning.

[0006] In a first aspect, an embodiment of the present application provides a method for collaboratively monitoring and early warning of wind environment and structural response, comprising:

[0007] Historical wind field data and historical structural response data are obtained respectively, and the wind field type is used as the row index, the structural component as the column index, and the historical structural response data as the matrix element value to construct a three-dimensional feature matrix for advance classification of the wind environment; wherein, the wind field type is determined based on the historical wind field data; when the current wind field type is a conventional type, the structural component response threshold corresponding to the current wind field type is searched from the three-dimensional feature matrix, and if the current structural response data is close to or exceeds the structural component response threshold, an early warning message is generated; when the current wind field type is an unconventional type, the response prediction value of the structural component is determined based on the target prediction model, and an early warning message is generated based on the response prediction value; wherein, the target prediction model is obtained after training the initial prediction model according to the three-dimensional feature matrix.

[0008] In some embodiments, the manner of determining the wind farm type based on the wind farm data includes:

[0009] The historical wind farm data is input into a clustering model, and the wind farm type is determined based on the clustering model; wherein the step of constructing the clustering model includes: constructing a sample data set X={x1, x2…, x n}, where x i ={v i ,θ i ,T i ,RH i ,P i}, v i is the average wind speed, θ i is the wind direction angle, T i is the air temperature, RH i Relative humidity, P i is the atmospheric pressure; iterative optimization objective function The initial clustering model is trained to obtain the clustering model after the training is completed, where k is the number of clusters, n is the total number of samples, and ω ij For sample x i Membership to cluster j, μ j is the centroid of cluster j.

[0010] In some embodiments, the step of training the initial prediction model according to the three-dimensional feature matrix includes:

[0011] Constructing the network structure of the initial prediction model, the initial prediction model is a convolutional neural network model, and the network structure of the initial prediction model includes a convolution layer, a pooling layer and a fully connected layer. The convolution layer extracts local features between the wind field type and the component response from the three-dimensional feature matrix, the pooling layer uses a pooling operation to perform spatial dimensionality reduction on the feature map output by the convolution layer, and the fully connected layer outputs the response prediction value; the three-dimensional feature matrix is ​​converted into an image-like format through channel reorganization, and the three-dimensional feature matrix in the image-like format is used as the input of the initial prediction model to train the initial prediction model, and the target prediction model is obtained when the training is completed.

[0012] In some embodiments, the method provided in the present application may further include: driving the structural response monitoring sensor to collect current structural response data according to a preset low-frequency sampling frequency; and when it is determined that the wind speed change rate is greater than a preset wind speed change threshold, driving the structural response monitoring sensor to collect current structural response data according to a preset high-frequency sampling frequency.

[0013] In some embodiments, the method provided in the present application may further include: determining a risk coefficient based on the structural component response threshold and the structural response data; when the risk coefficient is greater than or equal to the first risk coefficient threshold, giving priority to a first warning method for warning prompts; when the risk coefficient is less than the first risk coefficient and greater than or equal to the second risk coefficient threshold, giving priority to a second warning method for warning prompts; when the risk coefficient is less than the second risk coefficient and greater than or equal to the third risk coefficient threshold, giving priority to a third warning method for warning prompts.

[0014] In a second aspect, an embodiment of the present application provides a wind environment and structural response collaborative monitoring and early warning system, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the method described in the above description when executing the computer program.

[0015] In some embodiments, the system architecture of the wind environment and structural response collaborative monitoring and early warning system includes a perception layer, an edge computing layer, a data storage and management layer, and an application layer;

[0016] The perception layer includes a wind environment monitoring sensor group and a structural response monitoring sensor group, which are used to obtain wind field data and structural response data;

[0017] The edge computing layer includes a gateway device for receiving raw data from the perception layer, performing data processing on the raw data, running a wind environment clustering algorithm to identify wind field types, adjusting sensor sampling frequencies, and generating warning information based on the three-dimensional feature matrix;

[0018] The data storage and management layer includes a local edge database and a cloud data center. The local edge database is used to record the structured data processed by the edge computing layer, and the cloud data center is used to regularly synchronize key data of edge nodes through a secure channel.

[0019] The application layer includes a monitoring center software platform, which is used to display wind field data and structural response data, receive early warning information sent by the edge computing layer, confirm the early warning information, and receive configuration management information of the wind environment and structural response collaborative monitoring and early warning system.

[0020] In some embodiments, at the edge computing layer, an algorithm for dynamically adjusting the sensor sampling frequency includes:

[0021]

[0022] Where f(t) is the sampling frequency, f high is the preset high-frequency sampling frequency, f low is the preset low-frequency sampling frequency, θ is the wind speed change rate threshold, v t is the average wind speed in the current time period, v t-1 is the average wind speed in the previous time period, and Δt is the interval between the current time period and the previous time period.

[0023] In some embodiments, the task offloading and allocation mechanism of the edge computing layer includes:

[0024] Assume that the CPU usage of the gateway device is U edge , the remaining memory is M edge , the network bandwidth is B edge , the corresponding parameter of the cloud server is U cloud , M cloud , B cloud , define the task complexity index C t The task complexity is calculated based on the computing resources and data transmission volume required for the task. If C t Greater than the task complexity threshold C t0 , and meets the conditions In the case of , the computationally intensive tasks are offloaded to the cloud server; where α is the utilization ratio threshold, β is the memory margin ratio threshold, and γ is the bandwidth ratio threshold

[0025] In a third aspect, an embodiment of the present application further provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the muscle function assessment method as described in the above description is implemented.

[0026] Compared with the existing technology, the beneficial effect of the present application is that by constructing a collaborative monitoring system that integrates physical mechanisms and deep learning, the accuracy of structural safety warnings in complex wind environments can be improved. In this application, the wind field-structure response correlation is encoded into a three-dimensional feature matrix, which not only retains the multi-dimensional coupling relationship between wind field characteristics such as wind speed and direction and response parameters such as component strain and displacement, but also provides a parsable input paradigm for convolutional neural networks through image-like reconstruction. This dual-driven framework of mechanism and data in this application can break through the limitations of traditional threshold warnings: for conventional wind field types, the empirical threshold established based on the historical feature matrix can quickly identify response anomalies; and the trained prediction model can independently explore the implicit mapping rules between wind field types and component responses, which can improve the accuracy of response deduction under unseen wind field conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 Schematic diagram of the steps of the method for collaborative monitoring and early warning of wind environment and structural response provided in an embodiment of the present application.

[0028] Figure 2 Schematic diagram of the overall architecture of the wind environment and structural response collaborative monitoring and early warning system provided in an embodiment of the present application. DETAILED DESCRIPTION

[0029] The present application is further described in detail below in conjunction with test examples and specific implementation methods. However, this should not be understood as limiting the scope of the above-mentioned subject matter of the present application to the following embodiments. All technologies implemented based on the content of the present application fall within the scope of protection of the present application.

[0030] Unless otherwise specified, in the description of the specific embodiments of this application, the terms indicating the orientation or position relationship such as "up", "down", "left", "right", "center", "inside", "outside", and "side" are based on the expression of the orientation or position relationship shown in the accompanying drawings, or the orientation or position relationship in which the product / device / apparatus is placed when it is usually used. These terms of orientation or position relationship are only for the convenience of describing the scheme of this application or simplifying the description in the specific embodiments to facilitate the technicians to quickly understand the scheme, and do not indicate or imply that a specific device / component / element must have a specific orientation, or be constructed and operated in a specific position relationship, and therefore should not be understood as limiting this application.

[0031] In the description of the embodiments of this application, the technical terms "first," "second," etc., merely distinguish one entity or operation from another and are not to be understood as indicating or implying relative importance or implicitly specifying the quantity, specific order, or primary-secondary relationship of the indicated technical features. In the description of the embodiments of this application, "plurality" means two or more, unless otherwise specifically defined.

[0032] References herein to "embodiments" mean that a particular feature, structure, or characteristic described in connection with the embodiments may be included in at least one embodiment of the present application. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor does it constitute an independent or alternative embodiment that is mutually exclusive of other embodiments. It is understood, both explicitly and implicitly, by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0033] Example 1

[0034] The following embodiments are examples of methods for implementing coordinated monitoring and early warning of wind environment and structural response.

[0035] Please see Figure 1 and Figure 2 , Figure 1 Schematic diagram of the steps of the method for collaborative monitoring and early warning of wind environment and structural response provided in an embodiment of the present application. Figure 2 The overall architecture diagram of the wind environment and structural response collaborative monitoring and early warning system provided in the embodiment of the present application. The wind environment and structural response collaborative monitoring and early warning method provided in the embodiment of the present application can be applied to the wind environment and structural response collaborative monitoring and early warning system.

[0036] The following first introduces the wind environment and structural response collaborative monitoring and early warning system provided by the embodiment of the present application. The system architecture of the wind environment and structural response collaborative monitoring and early warning system includes a perception layer 21, an edge computing layer 22, a data storage and management layer 23, and an application layer 24.

[0037] The perception layer 21 includes a wind environment monitoring sensor group and a structural response monitoring sensor group, which are used to obtain wind field data and structural response data.

[0038] The edge computing layer 22 includes a gateway device for receiving the raw data from the perception layer 21, performing data processing on the raw data, running a wind environment clustering algorithm to identify the wind field type, adjusting the sensor sampling frequency, and generating early warning information based on the three-dimensional feature matrix.

[0039] The data storage and management layer 23 includes a local edge database and a cloud data center. The local edge database is used to record the structured data processed by the edge computing layer 22, and the cloud data center is used to regularly synchronize key data of edge nodes through a secure channel.

[0040] The application layer 24 includes a monitoring center software platform for displaying wind farm data and structural response data, receiving and confirming the warning information sent by the edge computing layer 22, and receiving configuration management information of the wind environment and structural response collaborative monitoring and warning system.

[0041] The following will take the wind environment and structural response collaborative monitoring and early warning system to implement the wind environment and structural response collaborative monitoring and early warning method as an example to illustrate.

[0042] The overall process of realizing the coordinated monitoring and early warning method of wind environment and structural response is as follows:

[0043] Execute step S1 to obtain historical wind field data and historical structural response data respectively, use wind field type as row index, structural component as column index, and historical structural response data as matrix element value to construct a three-dimensional feature matrix for advance classification of wind environment.

[0044] Among them, historical wind field data can be obtained by the wind environment monitoring sensor group, and historical structural response data can be obtained by the structural response monitoring sensor group. In the system's perception layer 21, the wind environment monitoring sensor group includes wind speed and direction sensors, temperature and humidity sensors, and air pressure sensors. The wind speed and direction sensors use ultrasonic or mechanical wind speed measurement principles to accurately capture wind speed changes. The measurement range covers 0 to 60 m / s, with a resolution of 0.01 m / s; the wind direction is 360° all-round sensing with a resolution of within ±1°. The temperature and humidity sensors are used to monitor the ambient temperature and humidity, and assist in determining the potential impact of atmospheric stability on the wind field. The temperature measurement accuracy is ±0.1°C, and the humidity accuracy is ±0.1% RH. The air pressure sensor is used to measure atmospheric pressure, providing supplementary data for wind field analysis, with an accuracy of ±0.5 hPa. Historical structural response data can be acquired using a structural response monitoring sensor suite, which includes displacement sensors, accelerometers, and strain gauges. Displacement sensors utilize laser, wire, or capacitive principles to monitor displacements at key structural locations. Their range is set based on actual structural requirements, with an accuracy of ±1 mm. Accelerometers are used to capture structural vibration acceleration and understand the dynamic response characteristics of the structure. They have a measurement range of ±2 g and an accuracy of ±0.5 mg. Strain gauges, affixed to the surfaces of key structural components, measure the strain generated by applied forces to the components with an accuracy of ±1 με, and are used to assess changes in internal forces within the components.

[0045] In the process of acquiring historical wind field data and historical structural response data, the perception layer 21 uses numerical simulation software (such as OpenFoam) to simulate the process of different wind field types (changing parameters such as wind speed, wind direction, and turbulence intensity) acting on the target structure, and obtains the response data of the structural components (strain, displacement, internal force, etc.). During the specific simulation, according to the principles of fluid dynamics, a wind field calculation domain is established, boundary conditions (such as inlet wind speed profile, outlet pressure conditions, wall roughness, etc.) are set, and the Navier-Stokes equations are solved to obtain the wind field distribution. The wind field is then loaded onto the structural finite element model (based on the principles of structural mechanics, structural unit division, definition of material properties and constraints, etc.) to calculate the component response.

[0046] Combined with the field measured data, the wind environment and structural real response information under different working conditions are accumulated during the actual wind disaster event or long-term monitoring process to supplement and correct the simulation data. Suppose the simulation data set is S = {s1, s2…, s n}, the measured data set is M={m1,m2…,m n}, through weighted fusion algorithm (such as giving simulated data weight ω according to data reliability s , measured data weight ω m , and ω s +ω m =1, fused data F = ω m S+ω m M), making the data closer to reality. When determining the weights, a cross-validation method can be used. A portion of the data is used as a validation set, and the weight values ​​are continuously adjusted to minimize the prediction error of the fused data on the validation set.

[0047] The acquired historical wind field data and historical structural response data can be stored in the data storage and management layer 23. In the system's data storage and management layer 23, the local edge database stores the wind environment and structural response monitoring data processed by the edge computing gateway. It uses a lightweight database (such as SQLite, InfluxDB, etc.) and stores it in a time series format to facilitate quick query and analysis of recent data trends. The SQLite database is based on the file system and stores data in a tree structure. It has the advantages of small size and fast read and write speed, and is suitable for local storage of edge computing devices. InfluxDB is designed for time series data and uses an efficient TSDB storage engine. It can quickly process write and query operations on massive time series data and can be selected according to actual needs. The local database also saves key configuration information such as clustering models, feature matrices, and warning thresholds to ensure the continuous and stable operation of edge computing.

[0048] The local database is configured with a data backup policy, such as automatically backing up data to an external storage device (such as an SD card or USB drive) every hour to prevent local data loss. The backup process uses an incremental backup method, backing up only data added or modified since the previous backup, improving backup efficiency. Furthermore, a data recovery mechanism is established to quickly restore data from backups in the event of data loss or corruption, ensuring normal system operation.

[0049] Cloud data centers regularly synchronize key data from edge nodes via secure channels (such as VPNs and dedicated lines) to achieve data backup and long-term archiving. Equipped with data cleaning and preprocessing tools, cloud data centers perform a second data quality check after receiving synchronized data from edge nodes, removing any potential duplicates or errors to further improve data quality. At the same time, they utilize big data analysis tools (such as Spark and Hive) to mine and analyze massive amounts of historical data. For example, correlation analysis can identify potential patterns between wind field characteristics and long-term changes in the response of structural components, providing a basis for optimizing feature matrices and adjusting warning thresholds.

[0050] After acquiring historical wind farm data and historical structural response data, the edge computing layer 22 can first determine the wind farm type based on the historical wind farm data. It then constructs a three-dimensional feature matrix for pre-classifying the wind environment, using the wind farm type as the row index, the structural component as the column index, and the historical structural response data as the matrix element values.

[0051] The edge computing layer 22 may determine the wind farm type based on the wind farm data in the following manner:

[0052] Construct a sample data set X={x1,x2…,x n}, where x i ={v i ,θ i ,T i ,RH i ,P i}, v i is the average wind speed, θ i is the wind direction angle, T i is the air temperature, RH i Relative humidity, P i is the atmospheric pressure;

[0053] Iterative optimization of the objective function The initial clustering model is trained to obtain the clustering model after the training is completed, where k is the number of clusters, n is the total number of samples, and ω ij For sample x i Membership to cluster j, μ j is the centroid of cluster j.

[0054] After training, the clustering model can automatically identify typical wind types, such as daily thermally driven gales, gusts, monsoons, gusts, turbulent winds, and severe winds. The number of clusters can be determined empirically or by the elbow method (plotting a clustering error versus k curve and selecting the k value at the inflection point of the curve).

[0055] After determining the wind field type, the edge computing layer 22 uses the wind field type as the row index, the structural component as the column index, and the component response as the matrix element value to construct a three-dimensional feature matrix. For example, for a bridge, the wind field type can be divided into daily thermal-driven strong winds, gusts, monsoons, sudden winds, turbulent winds, and violent winds. The structural components include piers, main beams, suspenders, etc. The corresponding component responses are the values ​​of the stress, displacement extremes, etc. of each component under different wind fields. Assuming there are k types of wind field types, l types of structural components, and r types of component responses, the feature matrix C can be expressed as C k,l,r .

[0056] The hardware configuration of the edge computing gateway uses industrial-grade gateway devices with high-performance processors, large-capacity memory, and rich communication interfaces. The edge computing gateway receives wind environment and structural response data from the perception layer 21 in real time, performs preliminary data cleaning, removes outliers (such as sensor readings that clearly exceed the reasonable range, which is determined by setting the upper and lower limits of the normal value range of each parameter; if exceeded, it is considered an anomaly) and filters noise (using the Butterworth low-pass filter algorithm, with the cutoff frequency set according to the wind field and structural response frequency characteristics. Generally, for wind environment monitoring, the cutoff frequency of wind speed and direction data is set to 10Hz, and the cutoff frequency of structural response data is adjusted according to the natural frequency of the structure, such as 0.5Hz for bridge structures) to ensure data quality.

[0057] When constructing the three-dimensional feature matrix, the edge computing layer 22 can introduce active learning strategies from machine learning to proactively select and collect the most valuable data for constructing the feature matrix. For example, based on the uncertainty sampling principle, wind field and structural response data combinations with significant differences between simulation results and measured results are prioritized for re-simulation or field measurement to fill knowledge gaps and optimize the feature matrix. Specifically, uncertainty indicators, such as prediction variance or information entropy, are calculated for each data point, and points with high uncertainty are prioritized for collection.

[0058] Furthermore, the edge computing layer 22 can also use association analysis technology (such as the variant of the Apriori algorithm and the FP-Growth algorithm for continuous numerical data association) to explore the deep relationship between wind field characteristics and component responses, optimize the feature matrix, and improve its accuracy in reflecting the actual situation. Assume that the association rule mined is A→B (A is the wind field feature subset, B is the component response subset), and by calculating the support (#(A∪B) is the number of data sets that contain both A and B, N is the total number of data sets), confidence (#(A) is the number of data sets containing A), strong association rules are screened out, and the element values ​​of the feature matrix are adjusted accordingly to make it more consistent with actual physical laws.

[0059] After the edge computing layer 22 completes the construction of the three-dimensional feature matrix, if the current wind farm type is a conventional type, step S2 is executed. When the current wind farm type is a conventional type, the structural component response threshold corresponding to the current wind farm type is searched from the three-dimensional feature matrix. If the current structural response data is close to or exceeds the structural component response threshold, an early warning message is generated.

[0060] Among them, historical wind field data and historical structural response data can be stored in the local edge database. According to the three-dimensional feature matrix of wind field type-structural component response established in step S1, the potential response risk of the structural component corresponding to the current wind field type can be quickly found.

[0061] In the edge computing layer 22, the structural component response threshold can be set to y th (Set according to structural design specifications and historical experience data, such as the bridge main beam displacement threshold is set to 10mm, the bridge pier strain threshold is set to 100με, etc.) When the monitored structural response data y approaches or exceeds the warning threshold, an early warning message is immediately generated.

[0062] If the current wind farm type is an unconventional type, step S3 is executed. In the case that the current wind farm type is an unconventional type, a response prediction value of the structural component is determined based on the target prediction model, and warning information is generated based on the response prediction value.

[0063] The target prediction model is obtained by training the initial prediction model according to the three-dimensional feature matrix, and the steps of training the initial prediction model include:

[0064] Construct a network structure of the initial prediction model, which is a convolutional neural network model. The network structure of the initial prediction model includes a convolution layer, a pooling layer and a fully connected layer. The convolution layer extracts local features between the wind field type and the component response from the three-dimensional feature matrix, the pooling layer uses a pooling operation to perform spatial dimensionality reduction on the feature map output by the convolution layer, and the fully connected layer outputs the response prediction value.

[0065] The three-dimensional feature matrix is ​​converted into an image-like format through channel reorganization, and the three-dimensional feature matrix in the image-like format is used as the input of the initial prediction model to train the initial prediction model, and the target prediction model is obtained when the training is completed.

[0066] After the training is completed, the response prediction value of the structural component can be determined based on the target prediction model. If the response prediction value is close to or exceeds the response threshold, an early warning message is generated.

[0067] When the current wind farm type is an unconventional type, the edge computing layer 22 can use the convolutional neural network (CNN) in deep learning to perform feature extraction and pattern recognition on the feature matrix. By treating the feature matrix as image data and inputting it into the CNN model, the convolution kernel of the convolution layer is used to extract the wind farm type and the local features between the wind farm type and the component response. The pooling layer performs dimensionality reduction and the connection layer performs classification or regression prediction. The CNN model is trained with a large amount of historical data so that it can automatically identify complex wind farm-component response patterns, further improving the effectiveness of the feature matrix. For example, a trained CNN model can quickly determine the abnormal response that may occur in a specific structural component under a certain new wind farm mode and issue an early warning.

[0068] Example 2

[0069] The following embodiments are examples of implementing sensor sleep and wake-up control.

[0070] The perception layer 21 can drive the structural response monitoring sensor to collect the current structural response data according to the preset low-frequency sampling frequency. When it is determined that the wind speed change rate is greater than the preset wind speed change threshold, the perception layer 21 can drive the structural response monitoring sensor to collect the current structural response data according to the preset high-frequency sampling frequency.

[0071] For example, the sensing layer 21 can drive the structural response monitoring sensor to collect data at a preset low-frequency sampling frequency, such as 0.1Hz or 0.01Hz. At the same time, the wind speed data at time t is continuously compared with the wind speed data at time t-1 and earlier, and the wind speed change rate is calculated. Δt is the time interval, which can be set to 1s or longer. Once a clear trend of increasing wind speed is detected, such as when Δv>5m / s (this threshold is set based on the sensitivity of the structure to wind loads and historical wind disaster data), the relevant sensors are immediately awakened and switched to a high sampling frequency, such as 1Hz or 10Hz, to ensure that subtle changes in the structural response are captured.

[0072] In addition, considering the impact of ambient temperature and humidity on sensor performance, when the temperature is higher than 40°C or the humidity is greater than 80% RH, the low-frequency sampling time interval can be appropriately extended, for example, from 0.1Hz to 0.05Hz, to reduce the risk of increased power consumption and measurement errors caused by harsh environments.

[0073] Example 3

[0074] The following embodiments are examples of strategies for adjusting the sampling frequency at the perception layer.

[0075] In the system’s perception layer 21, a low-power sensor chip is selected and equipped with an intelligent power management module. This module can receive control instructions from the edge computing gateway to achieve fast power switching (from sleep to work, working frequency switching). Assume that the power consumption of the sensor in sleep mode is Psleep , the power consumption when working at low frequency is P low , when working at high frequency, the power consumption is P high , obtain the power consumption values ​​through experimental testing or chip manuals. The edge computing layer 22 implements the algorithm for dynamically adjusting the sensor sampling frequency, including:

[0076]

[0077] Where f(t) is the sampling frequency, f high is the preset high-frequency sampling frequency, f low is the preset low-frequency sampling frequency, θ is the wind speed change rate threshold, v t is the average wind speed in the current time period, v t-1 is the average wind speed in the previous time period, and Δt is the interval between the current time period and the previous time period.

[0078] In addition to judging based on wind speed changes, this application also comprehensively considers the impact of factors such as ambient temperature and humidity on sensor power consumption. Assuming the ambient temperature is T and the humidity is RH, the sensor power consumption model is established. (P base is the basic power consumption, k T 、k RH is the temperature and humidity influence coefficient, P i (f i ) is the frequency of the i-th sensor at f i Energy consumption is estimated based on sensor type, sampling frequency, working time and other parameters. On the premise of meeting the monitoring accuracy requirements, the sampling frequency strategy is continuously optimized to maximize energy saving.

[0079] Example of the above algorithm:

[0080]

[0081] In addition, reinforcement learning algorithms can be introduced to allow the edge computing gateway to autonomously learn the optimal sensor sleep and wake-up strategies. The various sensor states (such as sleep, low-frequency operation, and high-frequency operation) are considered different actions, and the environmental state (including wind speed, temperature, humidity, etc.) is used as the state space. The reward function is to minimize system energy consumption and maximize monitoring accuracy. Through continuous trial and error, the gateway gradually learns to select the most appropriate sensor actions under different environmental conditions, achieving a long-term balance between energy conservation and accurate monitoring. For example, in the initial stage, the gateway may frequently wake up the sensor, but as learning deepens, it can accurately control the sensor according to environmental changes, reducing unnecessary energy consumption.

[0082] The edge computing gateway can also introduce an adaptive learning mechanism, enabling the system to automatically optimize the wind environment clustering model and feature matrix based on the continuously accumulated monitoring data. At regular intervals (such as one week), the clustering algorithm is incrementally learned using the newly added data to redefine the clustering center and membership function of the wind farm type to adapt to the seasonal and long-term changes in the wind farm environment. At the same time, the feature matrix is ​​dynamically updated. Based on the association rule mining algorithm (such as the improved Apriori algorithm), new potential relationships between wind farm characteristics and component responses are discovered from the new data, and the feature matrix element values ​​are adjusted in a timely manner to improve the accuracy of the early warning. For example, if a new wind farm pattern is found in a certain area in a specific season, and the wind farm produces unique response changes to a certain type of structural component, the system can automatically identify and update the relevant model parameters without manual intervention.

[0083] Example 4

[0084] The following embodiments are examples of optimizing edge computing technology in the edge layer.

[0085] Edge computing layer 22 enables optimization of edge computing technologies, including task offloading and allocation, as well as caching mechanisms. Task offloading and allocation in edge computing layer 22 can be tailored to complex data processing tasks (such as real-time analysis of large-scale wind farm simulation data and frequent querying and updating of feature matrices). Task offloading strategies can be rationally designed based on the real-time resource status of edge computing gateways and cloud servers (including CPU usage, remaining memory, network bandwidth, etc.).

[0086] For example, suppose the CPU usage of the gateway device is U edge , the remaining memory is M edge , the network bandwidth is B edge , the corresponding parameter of the cloud server is U cloud , M cloud , B cloud , define the task complexity index C t The task complexity is calculated based on the computing resources and data transmission volume required for the task. If C t Greater than the task complexity threshold C t0 , and meets the conditions In this case, the computationally intensive tasks are offloaded to the cloud server; wherein α is the utilization ratio threshold, which can be taken as 0.8, β is the memory margin ratio threshold, which can be taken as 2, and γ is the bandwidth ratio threshold, which can be taken as 3.

[0087] The cache mechanism of the edge computing layer 22 is to establish a data cache area in the edge computing gateway to cache frequently accessed data (such as recent wind environment classification results and commonly used feature matrix subsets) to reduce repeated data queries and transmissions and improve processing efficiency. Advanced cache replacement algorithms (such as improved versions of LRU and LFU) are used to ensure that the cache area always stores the most valuable data. Let the data access frequency be f and the most recent access time be t, and define the data value index (λ is the weight coefficient, ranging from 0 to 1, determined according to the importance of data timeliness), and the cache replacement order is determined according to the size of the V value.

[0088] Example 5

[0089] The following embodiment is an example of introducing a risk factor to comprehensively evaluate the urgency of an early warning and thereby determine an early warning prompt method.

[0090] The application layer 24 can introduce a risk coefficient to comprehensively evaluate the urgency of the warning. When the risk coefficient is greater than or equal to the first risk coefficient threshold, the first warning method is used for warning prompting. When the risk coefficient is less than the first risk coefficient and greater than or equal to the second risk coefficient threshold, the second warning method is used for warning prompting. When the risk coefficient is less than the second risk coefficient and greater than or equal to the third risk coefficient threshold, the third warning method is used for warning prompting.

[0091] For example, the risk factor Among them, y is the structural response data, y th is the structural response threshold. When R≥0.95, it is marked as a high-risk warning, and relevant personnel are notified by strong prompts such as sound and light alarms. When 0.85≤R≤0.95, it is marked as a medium-risk warning, and notifications are made by SMS, pop-up windows, etc. When 0.75≤R≤0.85, it is marked as a low-risk warning, and displayed as a prompt message on the monitoring platform for reference by operation and maintenance personnel.

[0092] Example 6

[0093] The following embodiments are examples for realizing data visualization, early warning management, and system management.

[0094] The system's monitoring center software platform can realize data visualization, early warning management and system management functions.

[0095] Among them, data visualization uses intuitive charts (such as wind speed and direction rose diagrams, structural displacement time history diagrams, strain distribution cloud diagrams, etc.) to display real-time wind environment and structural response data, allowing managers to quickly understand the on-site situation. The wind speed and direction rose diagram is presented in polar coordinates by counting the frequency of wind speeds in different wind direction ranges, intuitively reflecting the relationship between the wind direction distribution and wind speed in the wind field; the structural displacement time history diagram uses time as the horizontal axis and displacement as the vertical axis to draw the displacement curve of key parts of the structure over time, clearly showing the dynamic displacement changes of the structure; the strain distribution cloud diagram uses color gradients to represent the strain distribution on the surface of structural components, making it easy to quickly locate areas with larger strains. The software platform uses professional visualization libraries (such as Echarts, D3.js, etc.) to draw these charts, providing a smooth visualization experience.

[0096] Early warning management involves receiving and aggregating pre-warning information from edge computing gateways, and then finalizing the warning information based on the results of cloud-based big data analysis. Once the risk is confirmed, precise early warning notifications are issued to relevant personnel (including structural maintenance teams, safety managers, and surrounding residents) via SMS, email, and sound and light alarms. The notifications contain information about the wind farm type, dangerous structural locations, estimated risk levels, and recommended measures. The early warning notification system is integrated with third-party SMS platforms and email servers to ensure the timeliness and reliability of notifications. At the same time, a warning history record library is established on the monitoring platform to archive all issued warning information, including the warning time, warning content, and handling status, to facilitate subsequent tracing and analysis.

[0097] System management is the configuration management of the entire monitoring and early warning system, including operations such as sensor parameter settings, clustering model updates, and early warning threshold adjustments to ensure the system's adaptability and reliability. System administrators can remotely log in to the edge computing gateway through the monitoring platform's management interface to retrain its internal clustering model (when significant changes in wind farm characteristics are found, such as when new high-rise buildings in the surrounding area change the wind farm environment), and adjust parameters such as sensor sampling frequency and early warning thresholds to adapt to new monitoring needs. At the same time, the system management module has user permission management functions, with different settings for different users. For example, ordinary operation and maintenance personnel can only view data and receive early warning notifications, while administrators can configure the system and adjust parameters to ensure the safe and stable operation of the system.

[0098] In summary, in the collaborative monitoring and early warning method for wind environment and structural response provided in the embodiment of the present application, the accuracy of structural safety early warning in complex wind environments can be improved by constructing a collaborative monitoring system that integrates physical mechanisms and deep learning. In this application, the wind field-structural response correlation is encoded into a three-dimensional feature matrix, which not only retains the multi-dimensional coupling relationship between wind field characteristics such as wind speed and direction and response parameters such as component strain and displacement, but also provides a parsable input paradigm for convolutional neural networks through image-like reconstruction. This dual-driven mechanism and data framework in the present application can break through the limitations of traditional threshold early warning: for conventional wind field types, the empirical threshold established based on the historical feature matrix can quickly identify response anomalies; and the trained prediction model can independently explore the implicit mapping rules between wind field types and component responses, which can improve the accuracy of response deduction under unseen wind field conditions.

[0099] It should be understood that when the various modules of the system provided in the above embodiments are working, they are only illustrated by the division of the various functional modules in the above description. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above.

[0100] The functional modules in the above embodiments may be integrated into a single processing unit, or each unit may exist physically separately, or two or more units may be integrated into a single unit. The above integrated units may be implemented in the form of hardware or software functional units. In addition, the specific names of the functional units and modules are only for the purpose of distinguishing them from each other and are not intended to limit the scope of protection of the embodiments of this application.

[0101] Based on the same application concept, an embodiment of the present application further provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the method described in the above description.

[0102] The above-described embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present application, and should all be included in the scope of protection of the present application.

Claims

1. A wind environment and structural response collaborative monitoring and early warning method, characterized in that: include: Acquire historical wind farm data and historical structural response data respectively, use wind farm type as a row index, structural component as a column index, and the historical structural response data as matrix element values ​​to construct a three-dimensional feature matrix for pre-classification of the wind environment; wherein the wind farm type is determined based on the historical wind farm data; When the current wind farm type is a conventional type, searching the structural component response threshold corresponding to the current wind farm type from the three-dimensional feature matrix, and generating a warning message if the current structural response data approaches or exceeds the structural component response threshold; When the current wind field type is an unconventional type, a response prediction value of a structural component is determined based on a target prediction model, and warning information is generated based on the response prediction value; wherein, the target prediction model is obtained by training an initial prediction model according to the three-dimensional feature matrix.

2. The method according to claim 1, characterized in that The method of determining the wind farm type based on the wind farm data includes: Inputting the historical wind farm data into a clustering model, and determining the wind farm type based on the clustering model; wherein the step of constructing the clustering model includes: Construct a sample data set X={x1,x2…,x n }, where x i ={v i ,θ i ,T i ,RH i ,P i }, v i is the average wind speed, θ i is the wind direction angle, T i is the air temperature, RH i Relative humidity, P i is the atmospheric pressure; Iterative optimization of the objective function The initial clustering model is trained to obtain the clustering model after the training is completed, where k is the number of clusters, n is the total number of samples, and ω ij For sample x i Membership to cluster j, μ j is the centroid of cluster j.

3. The method according to claim 1, characterized in that The step of training the initial prediction model according to the three-dimensional feature matrix includes: Constructing a network structure of the initial prediction model, wherein the initial prediction model is a convolutional neural network model, and the network structure of the initial prediction model includes a convolution layer, a pooling layer, and a fully connected layer, wherein the convolution layer extracts local features between the wind field type and the component response from the three-dimensional feature matrix, the pooling layer uses a pooling operation to perform spatial dimensionality reduction on the feature map output by the convolution layer, and the fully connected layer outputs the response prediction value; The three-dimensional feature matrix is ​​converted into an image-like format through channel reorganization, and the three-dimensional feature matrix in the image-like format is used as the input of the initial prediction model to train the initial prediction model, and the target prediction model is obtained when the training is completed.

4. The method according to claim 1, wherein The method further includes: driving the structural response monitoring sensor to collect current structural response data according to a preset low-frequency sampling frequency; and when it is determined that the wind speed change rate is greater than a preset wind speed change threshold, driving the structural response monitoring sensor to collect current structural response data according to a preset high-frequency sampling frequency.

5. The method according to claim 1, wherein The method also includes: determining a risk coefficient based on the structural component response threshold and the structural response data; when the risk coefficient is greater than or equal to a first risk coefficient threshold, giving priority to a first warning method for warning prompts; when the risk coefficient is less than the first risk coefficient and greater than or equal to a second risk coefficient threshold, giving priority to a second warning method for warning prompts; when the risk coefficient is less than the second risk coefficient and greater than or equal to a third risk coefficient threshold, giving priority to a third warning method for warning prompts.

6. A wind environment and structural response coordinated monitoring and early warning system, characterized in that: The wind environment and structural response collaborative monitoring and early warning system includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the method according to any one of claims 1 to 5 is implemented.

7. The system according to claim 6, characterized in that The system architecture of the wind environment and structural response collaborative monitoring and early warning system includes a perception layer, an edge computing layer, a data storage and management layer, and an application layer; The perception layer includes a wind environment monitoring sensor group and a structural response monitoring sensor group, which are used to obtain wind field data and structural response data; The edge computing layer includes a gateway device for receiving raw data from the perception layer, performing data processing on the raw data, running a wind environment clustering algorithm to identify wind field types, adjusting sensor sampling frequencies, and generating warning information based on the three-dimensional feature matrix; The data storage and management layer includes a local edge database and a cloud data center. The local edge database is used to record the structured data processed by the edge computing layer, and the cloud data center is used to regularly synchronize key data of edge nodes through a secure channel. The application layer includes a monitoring center software platform, which is used to display wind field data and structural response data, receive early warning information sent by the edge computing layer, confirm the early warning information, and receive configuration management information of the wind environment and structural response collaborative monitoring and early warning system.

8. The system according to claim 6, wherein: At the edge computing layer, the algorithm for dynamically adjusting the sensor sampling frequency includes: Where f(t) is the sampling frequency, f high is the preset high-frequency sampling frequency, f low is the preset low-frequency sampling frequency, θ is the wind speed change rate threshold, v t is the average wind speed in the current time period, v t-1 is the average wind speed in the previous time period, and Δt is the interval between the current time period and the previous time period.

9. The system according to claim 6, wherein: The task offloading and allocation mechanism of the edge computing layer includes: Assume that the CPU usage of the gateway device is U edge , the remaining memory is M edge , the network bandwidth is B edge , the corresponding parameter of the cloud server is U cloud , M cloud , B cloud , define the task complexity index C t The task complexity is calculated based on the computing resources and data transmission volume required for the task. If C t Greater than the task complexity threshold C t0 , and meets the conditions In the case of , the computationally intensive tasks are offloaded to the cloud server; wherein α is the utilization ratio threshold, β is the memory margin ratio threshold, and γ is the bandwidth ratio threshold.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

Images