Method, device and program product for real-time monitoring of wind-induced vibration displacement of long-span cable net structure

Abstract

The present disclosure provides a large-span cable net structure wind-induced vibration displacement real-time monitoring method, device and program product, relating to the technical field of computers. The wind-induced vibration displacement real-time monitoring method of the present disclosure constructs a fiber optic cable force sensor by laying out a plurality of key monitoring points on the surface of the bearing cable and the surface of the cable, and rigidly fixes a machine vision target at the top of the mast. Real three-dimensional spatial displacement data of the top of the mast is obtained in a short period of time using the constructed binocular camera vision system, and a deep learning network is trained with cable force as input and three-dimensional spatial displacement data as output. The trained network can identify the three-dimensional spatial displacement of the top of the mast in real time based on the cable force, without the need to construct complex mechanical models and structural dynamics equations to accurately characterize the stress path of complex cable net structures, effectively solving the problem of difficulty in modeling by physical methods and difficulty in capturing nonlinear coupling relationships.

Description

Technical Field

[0001] This disclosure relates to the field of computer technology, and in particular to a method, device and program product for real-time monitoring of wind-induced vibration displacement of a long-span cable net structure. Background Technology

[0002] Long-span cable-net structures are widely used in large public buildings such as stadiums, convention centers, and airport terminals due to their advantages of light weight, strong span capacity, and aesthetic appeal. However, these structures typically exhibit high flexibility, low damping, and low natural frequencies, making them highly sensitive to wind loads. Under wind load excitation, cable-net structures are prone to significant vibration responses, especially the mast top, which plays a core supporting role. Its displacement response directly affects the safety and stability of the overall structure. Therefore, real-time and accurate monitoring of the three-dimensional spatial displacement of the mast top is of great significance for wind-induced vibration assessment and safety early warning of long-span cable-net structures.

[0003] Currently, methods for measuring structural displacement are mainly divided into two categories: contact and non-contact. Contact measurement methods primarily rely on traditional instruments such as linear variable differential transformers, potentiometer-type displacement sensors, and capacitive contact sensors. Although these instruments offer high measurement accuracy and resolution, they typically suffer from high cost, stringent installation conditions, susceptibility to mechanical wear, and limited measurement range, thus restricting their large-scale application in practical engineering.

[0004] Non-contact displacement measurement methods primarily utilize image processing technology, employing industrial cameras to track targets or natural features on structural surfaces, enabling remote and continuous structural displacement monitoring. These methods offer advantages such as flexible deployment and no structural disturbance, making them particularly suitable for locations where sensor installation is difficult. However, limitations imposed by external environmental factors such as changes in lighting, shading, and weather conditions like rain and fog hinder the stability and reliability of machine vision technology in long-term field monitoring, making it difficult to meet the demands of long-term engineering monitoring. Summary of the Invention

[0005] This disclosure provides a method, equipment, and program product for real-time monitoring of wind-induced vibration displacement of a long-span cable net structure.

[0006] According to one aspect of this disclosure, a method for real-time monitoring of wind-induced vibration displacement of a long-span cable net structure is provided to overcome the limitation that physical equations cannot adequately describe the complex nonlinear relationship between cable force and three-dimensional spatial displacement at the top of the mast, thereby enabling real-time identification of the three-dimensional spatial displacement of the mast. The cable net structure includes a roof structure formed by orthogonally arranged concave load-bearing cables and convex stabilizing cables, with ring beams and inclined columns serving as boundary conditions. The roof structure is connected to the mast via ring cables and edge cables, and the top of the mast is also connected to tension cables for balancing the forces. The mast is connected to the ground via spherical hinge supports. The real-time monitoring method for wind-induced vibration displacement includes: Fiber optic cable force sensors are deployed at multiple monitoring points of the cable net structure to form multiple monitoring channels, and the cable force of the cable net structure under actual wind load is collected through the multiple monitoring channels. The multiple monitoring points include the surface of the load-bearing cable and the surface of the tension cable. A machine vision target is rigidly fixed at the top of the mast. The machine vision target is observed by a constructed binocular camera vision system to obtain three-dimensional spatial displacement data of the top of the mast under actual wind load. Align the cable force and the three-dimensional spatial displacement data in time to construct a supervised learning dataset; The deep learning network model is trained based on the supervised learning data. The deep learning network model takes the cable force in the supervised learning data as input and the three-dimensional spatial displacement data corresponding to the cable force as labels to learn and establish a nonlinear mapping relationship between the cable force and the three-dimensional spatial displacement data. During the on-site monitoring phase, the cable forces collected in real time from multiple monitoring channels are input into a trained deep learning network model. The deep learning network then outputs the three-dimensional spatial displacement value of the top of the mast in real time, thereby enabling real-time monitoring of the wind-induced vibration displacement of the cable net structure.

[0007] According to one technical solution, fiber optic cable force sensors are deployed at multiple key monitoring points on the surface of the load-bearing cables and the tension cables. A machine vision target is rigidly fixed at the top of the mast. A binocular camera vision system is used to acquire real three-dimensional spatial displacement data of the mast top in a short period of time. Then, the cable force and three-dimensional spatial displacement data are aligned in time to construct a supervised learning dataset, and a deep learning network model is trained with cable force as input and three-dimensional spatial displacement data of the mast top as output. This method can accurately represent the force path of complex cable net structures without constructing complex mechanical models and structural dynamic equations, effectively solving the problems of difficult physical modeling and difficulty in capturing nonlinear coupling relationships.

[0008] Furthermore, after completing the network model training, the binocular cameras and targets, which are highly sensitive to environmental conditions, are removed. Relying solely on the cable force collected in real time by the long-term stable fiber optic cable force sensors, the trained deep learning network model can accurately and in real time output the three-dimensional spatial displacement of the mast top. This effectively overcomes the shortcomings of traditional contact displacement sensors, such as limited installation and easy wear, as well as the instability of non-contact machine vision technology under long-term monitoring conditions such as lighting, rain, fog, and obstruction. This ensures the accuracy and stability of long-term monitoring of large-span cable net structures under wind loads.

[0009] According to at least one embodiment of the present disclosure, a real-time monitoring method for wind-induced vibration displacement of a long-span cable net structure is used to observe the machine vision target through a constructed binocular camera vision system to obtain three-dimensional spatial displacement data of the mast top under actual wind load, including: Two industrial cameras are arranged in parallel to form a binocular camera vision system, and the internal and external parameters of the cameras and the baseline distance are calibrated. The two industrial cameras simultaneously acquire left and right image sequences containing the machine vision target, and perform sub-pixel precision matching on the machine vision target in the left and right images to obtain the pixel coordinates of the machine vision target in the left and right images. Based on the horizontal coordinates of the machine vision target in the left and right images and the camera focal length, calculate the angle between the machine vision target and the optical axes of the two industrial cameras. The three-dimensional spatial coordinates of the machine vision target are calculated based on the baseline distance and the angle between the machine vision target and the optical axes of the two industrial cameras. By tracking the changes in the three-dimensional spatial coordinates of the machine vision target in the left and right image sequences, the three-dimensional spatial displacement data of the mast top is obtained.

[0010] According to the technical solution of this embodiment, there is no need to install complex mechanical transmission devices at the top of the mast, thus avoiding problems such as mechanical wear, strict installation conditions, and limited measurement range. At the same time, the binocular vision system enables remote and continuous displacement monitoring without disturbing the structure, making it particularly suitable for high-altitude locations where sensor installation is difficult.

[0011] According to at least one embodiment of the wind-induced vibration displacement real-time monitoring method for a long-span cable net structure of the present disclosure, when aligning the cable force and the three-dimensional spatial displacement data in time to construct a supervised learning dataset, the wind-induced vibration displacement real-time monitoring method includes: Calculate the first correlation strength between the cable force collected by each monitoring channel and the three-dimensional spatial displacement data, and filter the deployed fiber optic cable force sensors according to the first correlation strength to obtain effective monitoring channels; Align the cable force collected by the effective monitoring channel with the three-dimensional spatial displacement data in time to construct a supervised learning dataset.

[0012] According to the technical solution of this embodiment, correlation analysis is used to eliminate monitoring channels where data interruption is caused by equipment failure, retaining only the cable force that is strongly correlated with the three-dimensional spatial displacement of the mast top for training. This effectively avoids interference from noisy and invalid data on model training. Even in the event of partial sensor failure, effective identification can still be achieved by relying on highly correlated data, ensuring the flexibility and fault tolerance of the data processing workflow.

[0013] According to at least one embodiment of the wind-induced vibration displacement real-time monitoring method for a large-span cable net structure of this disclosure, the correlation strength between the cable force collected by each monitoring channel and the three-dimensional spatial displacement data is calculated, and the deployed fiber optic cable force sensors are screened according to the correlation strength to obtain effective monitoring channels, including: Calculate the Pearson correlation coefficient between the cable force and three-dimensional spatial displacement data collected by each monitoring channel, and use it as the first correlation strength of the monitoring channel; The Pearson correlation coefficients of each monitoring channel are sorted in descending order of numerical value, and the top N monitoring channels are selected as effective monitoring channels, where N is the threshold for the number of effective monitoring channels.

[0014] According to the technical solution of this embodiment, the linear correlation between cable force and three-dimensional spatial displacement data at the top of the mast is quantified by the Pearson correlation coefficient, making the screening process completely data-driven, avoiding the subjectivity of human experience judgment, and providing a scientific and repeatable quantitative basis for sensor screening.

[0015] According to at least one embodiment of the wind-induced vibration displacement real-time monitoring method for a long-span cable net structure of the present disclosure, after determining the effective monitoring channel, the wind-induced vibration displacement real-time monitoring method further includes: Calculate the second correlation strength between the cable forces of each effective monitoring channel; When the second correlation strength between two valid monitoring channels is greater than the redundancy threshold, the two valid monitoring channels are determined to be a redundant channel pair. For each pair of redundant channels, retain the effective monitoring channels with high first correlation strength with the three-dimensional spatial displacement data, and remove the effective monitoring channels with low first correlation strength with the three-dimensional spatial displacement data.

[0016] According to the technical solution of this embodiment, while retaining the most representative fiber optic cable force sensor, information redundancy can be minimized, the input feature set can be optimized, and the efficiency and generalization ability of subsequent model training can be improved.

[0017] According to at least one embodiment of the wind-induced vibration displacement real-time monitoring method for a long-span cable net structure, multiple fiber optic cable force sensors with different initial center wavelengths are sequentially fused onto the same fiber to form a multi-point series sensing unit. The sensing unit is connected to a single fiber optic demodulator channel, and the wavelength division multiplexing principle is used to realize the simultaneous acquisition and identification of multiple fiber optic cable force sensors in the same monitoring channel.

[0018] According to the technical solution of this embodiment, a single demodulator channel can simultaneously acquire cable force from multiple sensors, realizing the reuse of monitoring channels, greatly reducing the complexity of on-site wiring and equipment channel requirements, and is especially suitable for distributed monitoring scenarios with multiple measurement points and long distances in large-span cable net structures.

[0019] According to at least one embodiment of the present disclosure, the method for real-time monitoring of wind-induced vibration displacement of a long-span cable net structure, wherein the deep learning network model is a network model based on the Transformer architecture, and the deep learning network model includes an encoder module and a regression head module; The encoder module is used to extract deep-level features of the cable force. Its data processing flow includes: constructing an M-dimensional cable force vector from the cable forces collected by multiple monitoring channels, where M is the number of monitoring channels; mapping the M-dimensional cable force vector to a multi-dimensional feature space through an embedding layer to generate M feature vectors; sequentially inputting the M feature vectors into multiple cascaded encoder modules, each encoder module containing a multi-head self-attention substructure and a feedforward neural network substructure; in the multi-head self-attention substructure, mining the correlation between cable forces in different monitoring channels by calculating the attention weights between each feature vector, and performing weighted fusion on each feature vector; in the feedforward neural network substructure, performing a nonlinear transformation on the weighted fused feature vectors to obtain a feature tensor. The regression head module is used to convert the feature tensor output by the encoder module into a three-dimensional spatial displacement prediction value of the mast top. Its data processing flow includes: flattening the feature tensor output by the last encoder module to obtain a one-dimensional feature vector; inputting the one-dimensional feature vector into multiple fully connected layers in sequence; performing a nonlinear transformation after each fully connected layer with an activation function; and outputting a two-dimensional displacement result from the last fully connected layer, which corresponds to the displacement values ​​of the mast top in two orthogonal directions in the horizontal plane.

[0020] According to the technical solution of this embodiment, by utilizing the multi-head attention mechanism of the Transformer network, the model can adaptively explore the complex dependencies between input cable forces and between them and the three-dimensional spatial displacement of the top of the mast, effectively solving the significant nonlinear coupling problem between cable forces and the three-dimensional spatial displacement of the top of the mast in the cable net structure, and improving the recognition accuracy.

[0021] According to another aspect of this disclosure, a real-time monitoring device for wind-induced vibration displacement of a large-span cable net structure is provided. The cable net structure includes a roof structure formed by orthogonally arranged concave load-bearing cables and convex stabilizing cables. The roof structure uses ring beams and inclined columns as boundary conditions. The roof structure is connected to a mast by ring cables and edge cables. The top of the mast is also connected to a tension cable for balancing the force. The mast is connected to the ground by a spherical hinge support. The real-time monitoring device for wind-induced vibration displacement includes: The cable force acquisition module is used to deploy fiber optic cable force sensors at multiple monitoring points of the cable net structure to form multiple monitoring channels, and to acquire the cable force of the cable net structure under actual wind load through the multiple monitoring channels. The multiple monitoring points include the surface of the load-bearing cable and the surface of the tension cable. The displacement data acquisition module is used to rigidly fix a machine vision target at the top of the mast, observe the machine vision target through a constructed binocular camera vision system, and acquire three-dimensional spatial displacement data of the top of the mast under actual wind load. A construction module is used to align the cable force and the three-dimensional spatial displacement data in time to construct a supervised learning dataset; The training module trains the deep learning network model based on the supervised learning data. The deep learning network model takes the cable force in the supervised learning data as input and the three-dimensional spatial displacement data corresponding to the cable force as labels to learn and establish a nonlinear mapping relationship between the cable force and the three-dimensional spatial displacement data. The processing module is used to input the cable force collected in real time from multiple monitoring channels into the trained deep learning network model during the on-site monitoring phase. The deep learning network then outputs the three-dimensional spatial displacement value of the top of the mast in real time, thereby realizing real-time monitoring of the wind-induced vibration displacement of the cable net structure.

[0022] According to another aspect of this disclosure, an electronic device is provided, comprising: a memory storing execution instructions; and a processor executing the execution instructions stored in the memory, such that the processor performs a real-time monitoring method for wind-induced vibration displacement of a large-span cable net structure according to any embodiment of this disclosure.

[0023] According to another aspect of this disclosure, a readable storage medium is provided, wherein executable instructions are stored therein, which, when executed by a processor, are used to implement a real-time monitoring method for wind-induced vibration displacement of a large-span cable net structure according to any embodiment of this disclosure.

[0024] According to another aspect of this disclosure, a computer program product is provided, including a computer program that, when executed by a processor, implements a method for real-time monitoring of wind-induced vibration displacement of a large-span cable net structure according to any embodiment of this disclosure.

[0025] The beneficial effect of this disclosure is that it can accurately characterize the force path of complex cable net structures without constructing complex mechanical models and structural dynamic equations, effectively solving the problems of difficult physical modeling and difficulty in capturing nonlinear coupling relationships, and ensuring the accuracy and stability of long-term monitoring of long-span cable net structures under wind loads. Attached Figure Description

[0026] The accompanying drawings illustrate exemplary embodiments of the present disclosure and, together with the description thereof, serve to explain the principles of the present disclosure. These drawings are included to provide a further understanding of the present disclosure and are incorporated in and constitute a part of this specification.

[0027] Figure 1 This is a flowchart illustrating a method for real-time monitoring of wind-induced vibration displacement of a large-span cable net structure according to one embodiment of the present disclosure.

[0028] Figure 2 This is a flowchart illustrating the process of acquiring three-dimensional spatial displacement data of the mast top in a real-time monitoring method for wind-induced vibration displacement of a large-span cable net structure according to an embodiment of the present disclosure.

[0029] Figure 3 This is a flowchart illustrating step S130 of a method for real-time monitoring of wind-induced vibration displacement of a large-span cable net structure according to an embodiment of the present disclosure.

[0030] Figure 4 This is a flowchart illustrating step S131 of a method for real-time monitoring of wind-induced vibration displacement of a large-span cable net structure according to an embodiment of the present disclosure.

[0031] Figure 5 This is a flowchart illustrating the process of eliminating redundant monitoring channels in a real-time monitoring method for wind-induced vibration displacement of a large-span cable net structure according to an embodiment of this disclosure.

[0032] Figure 6 This is a schematic structural block diagram of a real-time monitoring device for wind-induced vibration displacement of a large-span cable net structure according to one embodiment of the present disclosure.

[0033] Figure 7 This is a schematic structural block diagram of an electronic device according to one embodiment of the present disclosure. Detailed Implementation

[0034] The present disclosure will now be described in further detail with reference to the accompanying drawings and examples. It should be understood that the specific examples described herein are for illustrative purposes only and are not intended to limit the scope of the disclosure. Furthermore, it should be noted that, for ease of description, only the parts relevant to the present disclosure are shown in the accompanying drawings.

[0035] It should be noted that, where there is no conflict, the embodiments and features described in this disclosure can be combined with each other. The technical solutions of this disclosure will now be described in detail with reference to the accompanying drawings and embodiments.

[0036] Existing technologies cannot guarantee the accuracy and stability of displacement measurements to meet the needs of long-term engineering monitoring.

[0037] To address this issue, the present disclosure proposes the following technical solution: fiber optic cable force sensors are deployed at multiple key monitoring points on the surface of the load-bearing cable and the tension cable. A machine vision target is rigidly fixed at the top of the mast. A binocular camera vision system is used to acquire real three-dimensional spatial displacement data of the mast top in a short period. The cable force and displacement data are then aligned temporally to construct a supervised learning dataset, training a deep learning network model with cable force as input and three-dimensional spatial displacement data of the mast top as output. This approach achieves accurate characterization of the force path of complex cable net structures without the need to construct complex mechanical models and structural dynamic equations, effectively solving the problems of difficult physical modeling and the inability to capture nonlinear coupling relationships.

[0038] Furthermore, after the model training is completed, the binocular camera and target, which are highly sensitive to environmental conditions, are removed. Relying solely on the cable force collected in real time by the long-term stable fiber optic cable force sensors, the trained deep learning network model can accurately and in real time output the three-dimensional spatial displacement of the mast top. This effectively overcomes the shortcomings of traditional contact displacement sensors, such as limited installation and easy wear, as well as the instability of non-contact machine vision technology under long-term monitoring conditions such as light, rain, fog, and obstruction. This ensures the accuracy and stability of long-term monitoring of large-span cable net structures under wind loads.

[0039] To facilitate description and make the technical solution of this disclosure easier to understand, the long-span cable net structure of this disclosure will be described first before describing the technical solution of this disclosure.

[0040] This large-span cable net structure includes a roof structure, which is formed by orthogonally arranged concave load-bearing cables and convex stabilizing cables. The load-bearing cables mainly bear the tensile force generated by the roof's self-weight and external loads, while the stabilizing cables form a self-balancing system with the load-bearing cables through pretensioning. The boundary conditions of the roof structure are provided by a ring beam (such as a crescent-shaped steel ring beam) and inclined columns. The ring beam acts as the main load-bearing component, bearing the horizontal thrust transmitted from the cable net, while the inclined columns act as vertical support components, working together with the ring beam to form a stable boundary constraint system.

[0041] The roof structure is connected to the mast via ring cables and edge cables. The ring cables are in tension, and their tension is transmitted to the ring beam through the orthogonally arranged roof structure, putting the outer ring beam in compression. At the same time, the ring cables are connected to the mast on both sides, transferring part of the load to the mast.

[0042] As a key supporting component of the cable net structure, the mast is connected to the ground at its bottom through a fixed spherical hinge support to cope with the coupling effect of multi-directional deformation and complex internal forces. At least one cable is also connected to the top of the mast to balance the force. The cable and the edge sealing cable work together to transmit the wind load, maintain the force balance of the mast, and prevent the mast from overturning or becoming unstable under the action of wind load.

[0043] Figure 1 This is a flowchart illustrating a method for real-time monitoring of wind-induced vibration displacement of a large-span cable net structure according to one embodiment of this disclosure. This method can be applied to servers or terminal devices.

[0044] like Figure 1 As shown, the real-time monitoring method for wind-induced vibration displacement preferably includes steps S110 to S150, which are described in detail below.

[0045] In step S110, fiber optic cable force sensors are deployed at multiple monitoring points of the cable net structure to form multiple monitoring channels. The cable force of the cable net structure under actual wind load is collected through the multiple monitoring channels. The multiple monitoring points include the surface of the load-bearing cable and the surface of the tension cable.

[0046] In this embodiment, the constructed fiber optic cable force sensor is a regionally distributed fiber optic sensor. This sensor develops traditional point-distributed FBG (fiber Bragg grating) sensing technology into a regionally distributed sensing system with adjustable gauge length. The gauge length can be set from tens of centimeters to several meters according to actual engineering needs. Furthermore, the sensor core wire uses an excimer laser phase mask method to directly etch the grating onto the fiber, eliminating the need to remove the fiber coating. By measuring the average cable force over a relatively long gauge length, this fiber optic cable force sensor can effectively reflect changes in physical quantities within a certain area of ​​the structure. The strain range of the fiber optic cable force sensor is -1000 to +10000 με, the dynamic measurement accuracy is 1 με, and the sampling frequency supports 0 to 200 Hz.

[0047] Preferably, in the fiber optic cable force sensor, basalt fiber composite material is used as an encapsulation layer, uniformly wrapped around the outside of the fiber grating area. Its coefficient of thermal expansion is highly matched with that of the optical fiber material, effectively eliminating measurement errors caused by inconsistent temperature deformation. Simultaneously, this encapsulation material possesses excellent corrosion resistance, creep resistance, and long-term stability, resisting humidity, salt spray, ultraviolet radiation, and temperature changes in the service environment. Furthermore, the encapsulation end employs a variable stiffness anchoring design, with stiffness reinforcement zones set in areas of concentrated anchoring stress, effectively reducing the anchoring stress of the sensing core wire. This solves the problems of slippage and brittle fracture during the anchoring process of ultra-thin optical fibers, ensuring that the sensor maintains long-term stable operation under conditions of large cable deformation, with a service life of over 20 years.

[0048] By analyzing the force transmission path of the cable structure, multiple monitoring points are identified in the cable net structure, and fiber optic cable force sensors are deployed at these monitoring points. Preferably, these multiple monitoring points include multiple points on the surface of the load-bearing cable and multiple points on the surface of the tension cable.

[0049] Specifically, fiber optic cable force sensors are installed on the surface of the load-bearing cable using mechanical anchoring to track strain changes (i.e., cable force) caused by prestress loss or wind load; fiber optic cable force sensors are also installed on the surface of the cable connected to the mast to acquire cable deformation data, thereby allowing the inference of the mast's three-dimensional spatial displacement characteristics.

[0050] Next, the fiber optic cable force sensors deployed at different locations were divided into multiple monitoring channels according to the area. A high-performance fiber Bragg grating demodulator was used to connect each monitoring channel, featuring system temperature self-compensation and error self-calibration technology. Under actual wind load, the sampling frequency was set from 20Hz to 100Hz to collect wavelength change data from each monitoring channel in real time and demodulate it into cable force. During the acquisition process, the accuracy of the cable force timestamps was ensured for subsequent time alignment with the three-dimensional spatial displacement data of the mast top.

[0051] Please continue to refer to this. Figure 1 In step S120, a machine vision target is rigidly fixed at the top of the mast, and the machine vision target is observed through a binocular camera vision system to obtain three-dimensional spatial displacement data of the top of the mast under actual wind load.

[0052] In this embodiment, the machine vision target is an infrared emitting target, which is rigidly fixed to the top of the mast (e.g., by bolts), ensuring that there is no relative degree of freedom between the machine vision target and the mast.

[0053] Next, a binocular camera vision system is set up on the ground. The binocular camera vision system continuously captures a sequence of target images at a certain sampling frequency (e.g., 20Hz, synchronized with the cable force sampling frequency). The target image sequence is then used to calculate and determine the coordinate changes of the machine vision target, thereby obtaining the three-dimensional spatial displacement data of the mast top under actual wind load.

[0054] In step S130, the cable force and the three-dimensional spatial displacement data are aligned in time to construct a supervised learning dataset.

[0055] In this implementation, all cable forces and three-dimensional spatial displacement data at the top of the mast are aligned according to the corresponding timestamps to form supervised learning samples with one-to-one correspondence between input features (cable forces) and output labels (three-dimensional spatial displacement data at the top of the mast), thereby constructing a supervised learning dataset.

[0056] In step S140, a deep learning network model is trained based on the supervised learning data. The deep learning network model takes the cable force in the supervised learning data as input and the three-dimensional spatial displacement data corresponding to the cable force as labels to learn and establish a nonlinear mapping relationship between the cable force and the three-dimensional spatial displacement data.

[0057] In this implementation, a deep learning network model is pre-constructed, with the cable force in the supervised learning dataset as the input feature and the corresponding three-dimensional spatial displacement data of the mast top as the output label. The deep learning network model is trained multiple times, enabling it to learn and establish a nonlinear mapping relationship between the cable force and the three-dimensional spatial displacement data of the mast top. The output displacement prediction value gradually approaches the true displacement label.

[0058] In step S150, during the on-site monitoring phase, the cable forces collected in real time from multiple monitoring channels are input into the trained deep learning network model, and the deep learning network outputs the three-dimensional spatial displacement value of the top of the mast in real time, thereby realizing real-time monitoring of the wind-induced vibration displacement of the cable net structure.

[0059] In this implementation, once the deep learning network model has completed training and passed verification, it is deployed in the on-site monitoring system, entering the on-site monitoring phase. At this time, the binocular camera vision system and machine vision target used for collecting three-dimensional spatial displacement data of the mast during the training phase are removed, leaving only the fiber optic cable force sensors and the monitoring channels they form, which have been deployed on the key monitoring points of the cable net structure (the surfaces of the load-bearing cables and the tension cables).

[0060] The on-site monitoring system collects cable forces in real time from each monitoring channel at a preset sampling frequency (e.g., 20Hz) and feeds these forces as input into a pre-trained deep learning network model. This deep learning network model performs forward computation and outputs the three-dimensional spatial displacement value of the mast top in real time, thus achieving efficient displacement identification and wind-induced vibration monitoring without the need for complex observation conditions.

[0061] Therefore, according to the real-time monitoring method for wind-induced vibration displacement of long-span cable net structures disclosed herein, fiber optic cable force sensors are deployed at multiple key monitoring points on the surface of the load-bearing cables, and a machine vision target is rigidly fixed at the top of the mast. A binocular camera is used to acquire real three-dimensional spatial displacement data of the mast top in a short period. Then, the cable force and three-dimensional spatial displacement data are aligned in time to construct a supervised learning dataset, and a deep learning network model is trained with cable force as input and three-dimensional spatial displacement data as output. This method achieves accurate characterization of the force path of complex cable net structures without the need to construct complex mechanical models and structural dynamic equations, effectively solving the problems of difficult physical modeling and the inability to capture nonlinear coupling relationships.

[0062] Furthermore, after the model training is completed, the binocular camera and target, which are highly sensitive to environmental conditions, are removed. Relying solely on the cable force collected in real time by the long-term stable fiber optic cable force sensors, the trained deep learning network model can output the three-dimensional spatial displacement value of the mast top in real time and accurately. This effectively overcomes the shortcomings of traditional contact displacement sensors, such as limited installation and easy wear, as well as the instability of non-contact machine vision technology under long-term monitoring conditions such as light, rain, fog, and obstruction. This ensures the accuracy and stability of long-term monitoring of large-span cable net structures under wind loads.

[0063] In some embodiments of this disclosure, a binocular camera vision system is constructed to observe the machine vision target and obtain the three-dimensional spatial displacement data of the mast top under actual wind load. Preferably, this includes steps S210 to S250. Please refer to [reference needed]. Figure 2 .

[0064] In step S210, two industrial cameras are arranged in parallel to form a binocular camera vision system, and the camera's internal and external parameters and baseline distance are calibrated.

[0065] In step S220, two industrial cameras are used to simultaneously acquire left and right image sequences containing the machine vision target, and sub-pixel precision matching is performed on the machine vision target in the left and right images to obtain the pixel coordinates of the machine vision target in the left and right images.

[0066] In step S230, the angle between the machine vision target and the optical axis of the two industrial cameras is calculated based on the horizontal coordinate of the machine vision target in the left and right images and the camera focal length.

[0067] In step S240, the three-dimensional spatial coordinates of the machine vision target are calculated based on the baseline distance and the angle between the machine vision target and the optical axes of the two industrial cameras.

[0068] In step S250, the changes in the three-dimensional spatial coordinates of the machine vision target in the left and right image sequences are tracked to obtain the three-dimensional spatial displacement data of the top of the mast.

[0069] In this embodiment, two industrial cameras are placed in parallel to form a binocular camera vision system. Before formal observation, the intrinsic and extrinsic parameters of the two industrial cameras are calibrated using a checkerboard calibration board. Intrinsic parameter calibration includes focal length, principal point coordinates, radial distortion coefficient, and tangential distortion coefficient, etc., by capturing images of the calibration board in various poses, and then solving the camera intrinsic parameter matrix (e.g., using the Zhang Youzheng calibration method). Extrinsic parameter calibration determines the relative position and pose relationship between the two industrial cameras, obtaining the baseline distance (i.e., camera spacing) and the rotation matrix and translation vector of the two industrial cameras. After calibration, the parameters are saved to the industrial control computer system for subsequent 3D coordinate calculations.

[0070] Two industrial cameras are connected to an industrial control computer system on the same local area network via a gigabit Ethernet (GigE) interface, and are controlled and processed by unified secondary development software. After adjusting the external hardware parameters of the cameras (such as aperture and focus), the secondary development software is used to set and identify the observation parameters of the machine vision target, and then the synchronous acquisition function is started to implement synchronous observation and recording throughout the entire process.

[0071] The machine vision system continuously captures left and right images containing the machine vision target at a sampling frequency of 20Hz (the image captured by the left industrial camera is the left image, and the image captured by the right industrial camera is the right image), obtaining left and right image sequences (including the left image sequence and the right image sequence). After acquiring the images, the feature points in the left and right images are matched with sub-pixel precision using the Digital Image Correlation (DIC) algorithm to obtain the pixel coordinates of the target point. Then, template matching search is used to determine the optimal matching position within a certain window using grayscale correlation measurement, thereby obtaining the image displacement, and then updating the pixel coordinates of the target point, finally obtaining the pixel coordinates of the machine vision target in the left and right images.

[0072] Specifically, using the Digital Image Correlation (DIC) algorithm, a gray-level feature sub-region centered on the initial target position is first selected in the reference image (e.g., the left image) as a template. Then, within a preset search window in the target image (e.g., the right image or the left image of the next frame), normalized cross-correlation (NCC) or zero-mean normalized cross-correlation (ZNCC) is used to measure the gray-level similarity between the template and the window sub-region pixel by pixel, determining the integer pixel position that maximizes the correlation coefficient as the coarse matching result. Based on this, the matching position is refined to sub-pixel precision (typically 0.01–0.05 pixels) through correlation coefficient surface fitting (e.g., a quadratic parabola) or gradient iteration, thereby obtaining the precise pixel coordinates of the target in the target image. Next, the current position of the target is updated using these precise pixel coordinates, and the matching result of the previous frame is used as the search center of the next frame, achieving frame-by-frame recursive tracking. Simultaneously, the above process is performed independently on the left and right image sequences to obtain the pixel coordinates of the machine vision target in the left and right images at the same time, which are then used for disparity calculation and subsequent 3D spatial coordinate solution.

[0073] Then, based on the pinhole imaging model, using the horizontal coordinates of the machine vision target in the left and right images and the camera focal length, the angles between the machine vision target and the optical axes of the left and right cameras are calculated respectively. , The calculation formula is as follows: , in, f The focal length of the camera; XL and XR These are the x-coordinates of the machine vision target in the left and right images, respectively.

[0074] Based on the principle of triangulation, and combining the baseline distance and the two included angles mentioned above, the three-dimensional spatial coordinates of the target (including depth coordinates and planar coordinates) are calculated. The formula for calculating the depth coordinates is as follows: .

[0075] Finally, by repeating the above calculations on a series of left and right images, the three-dimensional spatial coordinates of the machine vision target at each moment are tracked and compared with the initial time coordinates (determined based on the left and right images at the initial time) to obtain the real-time three-dimensional spatial displacement data of the mast top in three-dimensional space.

[0076] This eliminates the need for complex mechanical transmission devices at the top of the mast, avoiding problems such as mechanical wear, stringent installation conditions, and limited measurement range. Furthermore, the binocular vision system enables remote, continuous displacement monitoring without disturbing the structure, making it particularly suitable for high-altitude locations where sensor installation is difficult.

[0077] In some embodiments of this disclosure, step S130, aligning the cable force and the three-dimensional spatial displacement data in time to construct a supervised learning dataset, preferably includes steps S131 to S132. Please refer to [reference needed]. Figure 3 .

[0078] In step S131, the first correlation strength between the cable force collected by each monitoring channel and the three-dimensional spatial displacement data is calculated, and the deployed fiber optic cable force sensors are screened according to the first correlation strength to obtain effective monitoring channels.

[0079] In step S132, the cable force collected by the effective monitoring channel is aligned with the three-dimensional spatial displacement data in time to construct a supervised learning dataset.

[0080] In this implementation, based on the cable force and three-dimensional spatial displacement data of the mast top collected synchronously at each monitoring point during the training phase, the first correlation strength between the cable force of each monitoring channel and the three-dimensional spatial displacement data of the mast top is calculated (i.e., a quantitative index of the degree of linear correlation between cable force and three-dimensional spatial displacement data, such as the Pearson coefficient).

[0081] Next, based on the primary correlation strength between the cable force and three-dimensional spatial displacement data collected by each monitoring channel, the channels with higher primary correlation strengths are selected as valid monitoring channels. For example, monitoring channels with primary correlation strengths higher than a certain threshold are selected as valid monitoring channels, or a predetermined number or proportion of monitoring channels with higher primary correlation strengths are selected as valid monitoring channels.

[0082] The cable force and three-dimensional spatial displacement data collected from the selected effective monitoring channels are strictly aligned with the timestamps to form a supervised learning dataset in which the input features and output labels correspond one-to-one.

[0083] In this way, by eliminating monitoring channels with data interruptions due to equipment failure through correlation analysis, only cable forces with strong correlation to mast top displacement are retained for training, effectively avoiding interference from noisy and invalid data on model training. Even in the event of partial sensor failure, effective identification can still be achieved by relying on highly correlated data, ensuring the flexibility and fault tolerance of the data processing workflow.

[0084] In some embodiments of this disclosure, step S131, calculating the correlation strength between the cable force collected by each monitoring channel and the three-dimensional spatial displacement data, and filtering the deployed fiber optic cable force sensors according to the correlation strength to obtain effective monitoring channels, preferably includes steps S1311 to S1312. Please refer to [reference needed]. Figure 4 .

[0085] In step S1311, the Pearson correlation coefficient between the cable force and three-dimensional spatial displacement data collected by each monitoring channel is calculated, which is used as the first correlation strength of that monitoring channel; and In step S1312, the Pearson correlation coefficients of each monitoring channel are sorted in descending order of numerical value, and the top N monitoring channels are selected as effective monitoring channels, where N is the threshold for the number of effective monitoring channels.

[0086] In this embodiment, to objectively and quantitatively select the fiber optic cable force sensor most correlated with the mast top displacement from multiple monitoring channels, the Pearson correlation coefficient is used as the quantitative indicator of the first correlation strength. Specifically, firstly, using the cable force and three-dimensional spatial displacement data of the mast top collected synchronously during the training phase from each monitoring channel, the Pearson correlation coefficient between the cable force and the three-dimensional spatial displacement data of the mast top for each monitoring channel is calculated.

[0087] Then, all monitoring channels are sorted in descending order of the absolute value of the Pearson correlation coefficient, and the top N monitoring channels are selected as valid monitoring channels, where N is a preset threshold for the number of valid monitoring channels (e.g., N=8). This screening method is based on the linear correlation metric in statistics and has the advantages of simple calculation, strong interpretability, and no need for manual experience judgment. It can ensure that the retained cable force and the three-dimensional spatial displacement data of the mast top have the strongest linear correlation, thus providing the most effective input features for subsequent deep learning models.

[0088] It should be noted that in other embodiments, other correlation coefficient calculation methods may also be used, and this disclosure does not impose any special limitations on them.

[0089] In this way, by quantifying the linear correlation between cable force and three-dimensional spatial displacement data through the Pearson correlation coefficient, the screening process is made entirely data-driven, avoiding the subjectivity of human experience judgment, and providing a scientific and repeatable quantitative basis for sensor screening.

[0090] In some embodiments of this disclosure, after determining the effective monitoring channel, the real-time monitoring method for wind-induced vibration displacement preferably further includes steps S510 to S530, please refer to... Figure 5 .

[0091] In step S510, the second correlation strength between the cable forces of each effective monitoring channel is calculated.

[0092] In step S520, when the second correlation strength between two valid monitoring channels is greater than the redundancy threshold, the two valid monitoring channels are determined to be a redundant channel pair.

[0093] In step S530, for each pair of redundant channels, the effective monitoring channels with a high first correlation strength with the three-dimensional spatial displacement data are retained, while the effective monitoring channels with a low first correlation strength with the three-dimensional spatial displacement data are removed.

[0094] In this implementation, after selecting effective monitoring channels, there may be information redundancy among these channels. That is, the cable forces collected by multiple fiber optic cable force sensors are highly correlated, reflecting similar information about the same stress area of ​​the structure. To avoid redundant features causing overfitting of the deep learning model and increasing the computational burden, redundant channels are further eliminated.

[0095] Specifically, firstly, the second correlation strength between the cable forces of all effective monitoring channels is calculated (using the Pearson correlation coefficient), and a redundancy threshold is preset (e.g., 0.9). When the absolute value of the second correlation strength between two effective monitoring channels is greater than the threshold, they are determined to constitute a redundant channel pair.

[0096] For each pair of redundant channels, compare the first correlation strength between the two valid monitoring channels in the pair and the three-dimensional spatial displacement data of the mast top. Retain the valid monitoring channel with the higher first correlation strength and discard the valid monitoring channel with the lower first correlation strength. Repeat the above process until the second correlation strength between all retained channels is less than or equal to the redundancy threshold.

[0097] In this way, while retaining the most representative fiber optic cable force sensor, information redundancy can be minimized, the input feature set can be optimized, and the efficiency and generalization ability of subsequent model training can be improved.

[0098] In some embodiments of this disclosure, multiple fiber optic cable force sensors with different initial center wavelengths are sequentially fused onto the same fiber to form a multi-point series sensing unit. The sensing unit is then connected to a single fiber optic demodulator channel, and the wavelength division multiplexing principle is used to achieve simultaneous acquisition and identification of multiple fiber optic cable force sensors within the same monitoring channel.

[0099] In this embodiment, a batch of long gauge-length fiber optic force sensors with different initial center wavelengths are selected (e.g., wavelength range of 1510nm to 1590nm, with appropriate center wavelength intervals between each fiber optic force sensor). Then, these fiber optic force sensors are sequentially fused onto the same transmission fiber using a fiber optic fusion splicer in a preset order to form a multi-point series sensing unit.

[0100] One end of the sensing unit is connected to a single channel of the fiber optic demodulator. When the fiber optic demodulator emits broadband light into the fiber, each sensor reflects narrowband light corresponding to its center wavelength. By detecting the wavelength of the returned light signal, the fiber optic demodulator can distinguish between different fiber optic cable force sensors and analyze the cable force value of each sensor.

[0101] In this way, a single demodulator channel can simultaneously collect cable force from multiple sensors, realizing the reuse of monitoring channels, greatly reducing the complexity of on-site wiring and the demand for equipment channels, and is especially suitable for distributed monitoring scenarios with multiple measurement points and long distances in long-span cable net structures.

[0102] In some embodiments of this disclosure, the deep learning network model is a network model based on the Transformer architecture, and the deep learning network model includes an encoder module and a regression head module; The encoder module is used to extract deep-level features of the cable force. Its data processing flow includes: constructing an M-dimensional cable force vector from the cable forces collected by multiple monitoring channels, where M is the number of monitoring channels; mapping the M-dimensional cable force vector to a multi-dimensional feature space through an embedding layer to generate M feature vectors; sequentially inputting the M feature vectors into multiple cascaded encoder modules, each encoder module containing a multi-head self-attention substructure and a feedforward neural network substructure; in the multi-head self-attention substructure, mining the correlation between cable forces in different monitoring channels by calculating the attention weights between each feature vector, and performing weighted fusion on each feature vector; in the feedforward neural network substructure, performing a nonlinear transformation on the weighted fused feature vectors to obtain a feature tensor. The regression head module is used to convert the feature tensor output by the encoder module into a three-dimensional spatial displacement prediction value of the mast top. Its data processing flow includes: flattening the feature tensor output by the last encoder module to obtain a one-dimensional feature vector; inputting the one-dimensional feature vector into multiple fully connected layers in sequence; performing a nonlinear transformation after each fully connected layer with an activation function; and outputting a two-dimensional displacement result from the last fully connected layer, which corresponds to the displacement values ​​of the mast top in two orthogonal directions in the horizontal plane.

[0103] In this implementation, the deep learning network model adopts a regression architecture based on the Transformer encoder, which is mainly composed of an encoder module and a regression head module connected in series.

[0104] The encoder module is used to extract deep features from the multi-channel cable force. Its data processing flow is as follows: First, the current cable force values ​​collected by the M monitoring channels are organized into an M-dimensional cable force vector, with the cable force of each monitoring channel being treated as an independent term. Then, a trainable embedding layer maps the M-dimensional cable force vector to a high-dimensional feature space (e.g., 64-dimensional), generating M feature vectors. These M feature vectors are then sequentially input into multiple cascaded encoder layers (two layers in this embodiment). Each encoder layer contains a multi-head self-attention substructure and a feedforward neural network substructure.

[0105] In the multi-head self-attention substructure, query, key, and value matrices are calculated for the input feature vectors respectively, and the attention weight matrix is ​​obtained based on the dot product of the query and the key. This weight matrix reflects the correlation strength between the cable forces of different monitoring channels. Based on this, the model performs weighted fusion of each feature vector, so that the sensor features most relevant to displacement prediction receive higher weights.

[0106] In the feedforward neural network substructure, a nonlinear transformation (such as first increasing the dimensionality and then decreasing it) is performed on the weighted and fused feature vectors to extract a more abstract feature representation. Preferably, residual connections and layer normalization are introduced after each feedforward neural network substructure to stabilize training.

[0107] The encoder module ultimately outputs a feature tensor. The regression head module is responsible for converting this feature tensor into a two-dimensional displacement prediction value for the mast top. Specifically, the feature tensor is first converted into a one-dimensional feature vector through a flattening operation. Then, this one-dimensional vector is sequentially input into multiple fully connected layers (three fully connected layers in this embodiment, with output dimensions of 128, 64, and 2, respectively). Each fully connected layer is followed by a ReLU activation function for nonlinear transformation. The last fully connected layer directly outputs a two-dimensional vector, corresponding to the displacement values ​​of the mast top in the first and second directions in the horizontal plane (where the first and second directions are perpendicular). It should be noted that since the mast bottom is connected to the ground through a fixed spherical hinge support, this fixed spherical hinge support constrains the translational motion of the mast top in the vertical direction. In addition, the mast itself has a large axial stiffness, and the vertical displacement caused by wind load is several orders of magnitude smaller than the horizontal displacement, which can be ignored in engineering monitoring.

[0108] The entire deep learning network model takes the cable force at the current moment as input and the three-dimensional spatial displacement of the mast top at the corresponding moment as output. Through end-to-end supervised learning, a nonlinear mapping from the cable force to the three-dimensional spatial displacement of the mast top is established, which can realize real-time displacement reconstruction without the need for historical time series accumulation.

[0109] Thus, by utilizing the multi-head attention mechanism of the Transformer network, the model can adaptively uncover the complex dependencies between input cable forces and between them and output displacements, effectively solving the significant nonlinear coupling problem between cable forces in cable net structures and the three-dimensional spatial displacement of the mast top, thereby improving recognition accuracy.

[0110] Furthermore, this method eliminates the need for complex mechanical models and structural dynamic equations, enabling real-time mapping and identification of structural dynamic displacements directly through cable force response. This characteristic is particularly suitable for large-span spatial structures with complex force paths and unique boundary conditions, effectively circumventing the challenges of traditional modeling.

[0111] Figure 6 This is a schematic structural block diagram of a real-time monitoring device for wind-induced vibration displacement of a large-span cable net structure according to one embodiment of the present disclosure.

[0112] like Figure 6 As shown, the wind-induced vibration displacement real-time monitoring device preferably includes a cable force acquisition module 610, a displacement data acquisition module 620, a construction module 630, a training module 640, and a processing module 650.

[0113] The cable force acquisition module 610 is used to deploy fiber optic cable force sensors at multiple monitoring points of the cable net structure, forming multiple monitoring channels, and to acquire the cable force of the cable net structure under actual wind load through these channels. The multiple monitoring points include the surfaces of the load-bearing cables and the tension cables. The displacement data acquisition module 620 is used to rigidly fix a machine vision target at the top of the mast, and to observe the machine vision target using a constructed binocular camera vision system to obtain three-dimensional spatial displacement data of the top of the mast under actual wind load. The construction module 630 is used to align the cable force and the three-dimensional spatial displacement data in time. The system constructs a supervised learning dataset; the training module 640 is used to train a deep learning network model based on the supervised learning data. The deep learning network model takes the cable force in the supervised learning data as input and the three-dimensional spatial displacement data corresponding to the cable force as labels to learn and establish a nonlinear mapping relationship between the cable force and the three-dimensional spatial displacement data; and the processing module 650 is used to input the cable force collected in real time from multiple monitoring channels into the trained deep learning network model during the on-site monitoring stage, and the deep learning network outputs the three-dimensional spatial displacement value of the top of the mast in real time, so as to realize real-time monitoring of the wind-induced vibration displacement of the cable net structure.

[0114] In some embodiments of this disclosure, after determining the effective monitoring channels, the construction module 630 is further configured to calculate the second correlation strength between the cable forces of each effective monitoring channel; when the second correlation strength between two effective monitoring channels is greater than the redundancy threshold, the two effective monitoring channels are determined to be a redundant channel pair; and for each redundant channel pair, the effective monitoring channels with higher first correlation strength with the three-dimensional spatial displacement data are retained, and the effective monitoring channels with lower first correlation strength with the three-dimensional spatial displacement data are removed.

[0115] Figure 7 This is a schematic structural block diagram of an electronic device according to one embodiment of the present disclosure.

[0116] like Figure 7 As shown, the hardware structure of the electronic device 1000 can be implemented using a bus architecture. The bus architecture can include any number of interconnect buses and bridges, depending on the specific application and overall design constraints of the hardware. Bus 1100 connects various circuits including one or more processors 1200, memory 1300, and / or hardware modules. Bus 1100 can also connect various other circuits 1400 such as peripherals, voltage regulators, power management circuits, external antennas, etc. Bus 1100 can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Component (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of illustration, only one connection line is used in this figure, but this does not indicate that there is only one bus or one type of bus.

[0117] This disclosure also provides a readable storage medium storing a computer program that, when executed by a processor, is used to implement the methods described above. A "readable storage medium" can be any means that can contain a program for storage, communication, propagation, or transmission for use by or in conjunction with an instruction execution system, apparatus, or device. More specific examples of a readable storage medium include: an electrical connection with one or more wires (electronic device), a portable computer disk drive (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and programmable read-only memory (EPROM or flash memory), fiber optic devices, and portable read-only memory (CDROM), etc.

[0118] This disclosure also provides a computer program product, the methods of which can be implemented wholly or partially through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented wholly or partially as a computer program product. The computer program product includes one or more computer programs or instructions. When the computer program or instructions are loaded and executed, all or part of the processes or functions of this disclosure are performed.

[0119] Computer programs or instructions can be stored in a readable storage medium or transferred from one readable storage medium to another. For example, the computer program or instructions can be transferred from one website, computer, server, or data center to another website, computer, server, or data center via wired or wireless means. The readable storage medium can be any available medium capable of access, or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium, such as a floppy disk, hard disk, or magnetic tape; an optical medium, such as a digital video optical disc; or a semiconductor medium, such as a solid-state drive. The computer-readable storage medium can be a volatile or non-volatile storage medium, or it can include both volatile and non-volatile types of storage media.

[0120] Those skilled in the art will understand that embodiments of this disclosure can be provided as methods, systems, or computer program products. Therefore, this disclosure can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this disclosure can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0121] This disclosure is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus, and computer program products according to this disclosure. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0122] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0123] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0124] In the description of this specification, the references to terms such as "one embodiment / mode," "some embodiments / modes," "example," "specific example," or "some examples," etc., refer to specific features, structures, or characteristics described in connection with that embodiment / mode or example, which are included in at least one embodiment / mode or example of this disclosure. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment / mode or example. Moreover, the specific features, structures, or characteristics described may be combined in any suitable manner in one or more embodiments / modes or examples. Furthermore, without contradiction, those skilled in the art can combine and integrate the different embodiments / modes or examples described in this specification, as well as the features of different embodiments / modes or examples.

[0125] Those skilled in the art should understand that the above embodiments are merely for illustrating the present disclosure and are not intended to limit the scope of the disclosure. Those skilled in the art can make other changes or modifications based on the above disclosure, and these changes or modifications still fall within the scope of the present disclosure.

Claims

1. A real-time monitoring method for wind-induced vibration displacement of a long-span cable net structure, used to overcome the limitation that physical equations cannot adequately describe the complex nonlinear relationship between cable force and three-dimensional spatial displacement at the top of the mast, and to achieve real-time identification of the three-dimensional spatial displacement of the mast; characterized in that, The cable net structure includes a roof structure formed by orthogonally arranged concave load-bearing cables and convex stabilizing cables. The roof structure uses ring beams and inclined columns as boundary conditions. The roof structure is connected to the mast by ring cables and edge cables. The top of the mast is also connected to tension cables for balancing the force. The mast is connected to the ground by spherical hinge supports. The real-time monitoring method for wind-induced vibration displacement includes: Fiber optic cable force sensors are deployed at multiple monitoring points of the cable net structure to form multiple monitoring channels, and the cable force of the cable net structure under actual wind load is collected through the multiple monitoring channels. The multiple monitoring points include the surface of the load-bearing cable and the surface of the tension cable. A machine vision target is rigidly fixed at the top of the mast. The machine vision target is observed by a constructed binocular camera vision system to obtain three-dimensional spatial displacement data of the top of the mast under actual wind load. Align the cable force and the three-dimensional spatial displacement data in time to construct a supervised learning dataset; The deep learning network model is trained based on the supervised learning data. The deep learning network model takes the cable force from the supervised learning data as input and the corresponding three-dimensional spatial displacement data as labels to learn and establish a nonlinear mapping relationship between the cable force and the three-dimensional spatial displacement data; and During the on-site monitoring phase, the cable forces collected in real time from multiple monitoring channels are input into a trained deep learning network model. The deep learning network then outputs the three-dimensional spatial displacement value of the top of the mast in real time, thereby enabling real-time monitoring of the wind-induced vibration displacement of the cable net structure.

2. The method for real-time monitoring of wind-induced vibration displacement as described in claim 1, characterized in that, The machine vision target is observed using a constructed binocular camera vision system to obtain three-dimensional spatial displacement data of the mast top under actual wind load, including: Two industrial cameras are arranged in parallel to form a binocular camera vision system, and the internal and external parameters of the cameras and the baseline distance are calibrated. The two industrial cameras simultaneously acquire left and right image sequences containing the machine vision target, and perform sub-pixel precision matching on the machine vision target in the left and right images to obtain the pixel coordinates of the machine vision target in the left and right images. Based on the horizontal coordinates of the machine vision target in the left and right images and the camera focal length, calculate the angle between the machine vision target and the optical axes of the two industrial cameras. Based on the baseline distance and the angle between the machine vision target and the optical axes of the two industrial cameras, calculate the three-dimensional spatial coordinates of the machine vision target; and By tracking the changes in the three-dimensional spatial coordinates of the machine vision target in the left and right image sequences, the three-dimensional spatial displacement data of the mast top is obtained.

3. The method for real-time monitoring of wind-induced vibration displacement as described in claim 1, characterized in that, When aligning the cable force and the three-dimensional spatial displacement data in time to construct a supervised learning dataset, the real-time monitoring method for wind-induced vibration displacement includes: Calculate the Pearson correlation coefficient between the cable force and three-dimensional spatial displacement data collected by each monitoring channel, using it as the first correlation strength for that monitoring channel. Then, sort the Pearson correlation coefficients of each monitoring channel in descending order of value, selecting the top N monitoring channels as effective monitoring channels, where N is a threshold for the number of effective monitoring channels. Align the cable force collected by the effective monitoring channel with the three-dimensional spatial displacement data in time to construct a supervised learning dataset.

4. The method for real-time monitoring of wind-induced vibration displacement as described in claim 3, characterized in that, After determining the effective monitoring channel, the real-time monitoring method for wind-induced vibration displacement further includes: The Pearson correlation coefficient between the cable forces of each effective monitoring channel is calculated as the second correlation strength; When the second correlation strength between two valid monitoring channels is greater than the redundancy threshold, the two valid monitoring channels are determined to be a redundant channel pair; and For each pair of redundant channels, retain the effective monitoring channels with high first correlation strength with the three-dimensional spatial displacement data, and remove the effective monitoring channels with low first correlation strength with the three-dimensional spatial displacement data.

5. The method for real-time monitoring of wind-induced vibration displacement as described in claim 1, characterized in that, Multiple fiber optic force sensors with different initial center wavelengths are sequentially fused onto the same fiber to form a multi-point series sensing unit. The sensing unit is then connected to a single fiber optic demodulator channel. The wavelength division multiplexing principle is used to achieve simultaneous acquisition and identification of multiple fiber optic force sensors within the same monitoring channel.

6. The method for real-time monitoring of wind-induced vibration displacement as described in any one of claims 1-5, characterized in that, The deep learning network model is a network model based on the Transformer architecture, and the deep learning network model includes an encoder module and a regression head module; The encoder module is used to extract deep-level features of the cable force. Its data processing flow includes: constructing an M-dimensional cable force vector from the cable forces collected by multiple monitoring channels, where M is the number of monitoring channels; mapping the M-dimensional cable force vector to a multi-dimensional feature space through an embedding layer to generate M feature vectors; sequentially inputting the M feature vectors into multiple cascaded encoder modules, each encoder module containing a multi-head self-attention substructure and a feedforward neural network substructure; in the multi-head self-attention substructure, mining the correlation between cable forces in different monitoring channels by calculating the attention weights between each feature vector, and performing weighted fusion on each feature vector; in the feedforward neural network substructure, performing a nonlinear transformation on the weighted fused feature vectors to obtain a feature tensor. The regression head module is used to convert the feature tensor output by the encoder module into a two-dimensional displacement prediction value of the mast top. Its data processing flow includes: flattening the feature tensor output by the last encoder module to obtain a one-dimensional feature vector; inputting the one-dimensional feature vector into multiple fully connected layers in sequence; performing a nonlinear transformation after each fully connected layer with an activation function; and outputting a two-dimensional displacement result from the last fully connected layer, which corresponds to the displacement values ​​of the mast top in two orthogonal directions in the horizontal plane.

7. An electronic device, characterized in that, include: The memory stores execution instructions; as well as A processor that executes the execution instructions stored in the memory, causing the processor to perform the real-time monitoring method for wind-induced vibration displacement as described in any one of claims 1 to 6.

8. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the real-time monitoring method for wind-induced vibration displacement as described in any one of claims 1 to 6.

Images