Bridge inhaul cable force monitoring method and system based on hybrid beam control scanning radar

Abstract

The invention discloses a bridge inhaul cable force monitoring method and system based on a hybrid beam control scanning radar, and the method comprises the steps: achieving the large-range mechanical coarse adjustment of a radar beam through a servo holder, and covering a whole bridge inhaul cable; and then, a phased array or digital beam forming technology is utilized to carry out electronic fine tuning on beam pointing and adaptively optimize the beam shape, so that accurate aiming of the beam on the inhaul cable and vibration signal acquisition with an extremely high signal-to-noise ratio are realized. And finally calculating the cable force through spectrum analysis. According to the invention, high-precision, high-efficiency and automatic polling monitoring of a plurality of inhaul cables by a single radar is realized, the problems of high cost, low efficiency, limited precision and the like in the prior art are solved, and reliable technical guarantee is provided for safe operation of bridges.

Description

Technical Field

[0001] This invention relates to the field of bridge structural health monitoring technology, and in particular to a method and system for monitoring bridge cable tension based on hybrid beam control scanning radar. Background Technology

[0002] As the primary load-bearing components, the stress changes in bridge cables are a key indicator for assessing bridge health. Existing cable stress monitoring technologies suffer from the following limitations: Contact sensor methods, such as accelerometers and fiber Bragg gratings, require sensors to be installed on each cable, leading to installation and maintenance difficulties, high costs, and poor long-term stability. Traditional single-point radar methods, while achieving non-contact measurement, rely on manual aiming, resulting in low efficiency and inability to achieve automated multi-cable monitoring. Fixed multi-radar methods require equipping each cable with an independent radar, resulting in extremely high costs, system complexity, and lack of economic viability. Current phased array technology achieves electronic beam scanning by controlling the phase difference of each radiating element in the array antenna, offering advantages such as agility, lack of inertia, and high precision; however, its scanning angle is typically limited to ±60°. Digital beamforming (DBF) technology allows for more flexible synthesis and control of beam shape. Currently, these advanced radar technologies have not yet been applied to solve the problem of precise beam control and optimization in automated multi-cable monitoring of bridges. Summary of the Invention

[0003] The purpose of this invention is to solve the above-mentioned problems by designing a method and system for monitoring the cable tension of bridges based on hybrid beam control scanning radar.

[0004] To achieve the above objectives, the technical solution of the present invention further includes the following steps in the above-mentioned bridge cable tension monitoring method based on hybrid beam control scanning radar: Obtain the macroscopic coordinates of the target cable, return the gimbal to zero for system self-calibration, control the gimbal rotation, and coarsely align the radar line of sight to the macroscopic area where the target cable is located; Based on the improved YOLOv5 model as the cable target detection algorithm, an attention mechanism and multi-scale feature fusion are introduced to establish a cable target detection model; The macroscopic area is scanned using DBF technology and then input into the cable target detection model to obtain the precise location of the cable; The beamwidth is dynamically adjusted based on the precise position of the cable to generate the optimal beam pattern and maximize the signal-to-noise ratio, resulting in an optimized beam. Based on the optimized beam acquisition of the time-domain signal of cable vibration, spectrum analysis is performed and the cable force of the bridge is calculated; The changing trend of the cable force is analyzed using an LSTM (Long Short-Term Memory) network to generate early warning information.

[0005] Furthermore, in the aforementioned bridge cable tension monitoring method based on hybrid beam control scanning radar, the steps of acquiring the macroscopic coordinates of the target cable, returning the gimbal to zero for system self-calibration, controlling the gimbal rotation, and coarsely aligning the radar line of sight to the macroscopic region where the target cable is located include: A high-precision total station was used to measure the spatial coordinates of the bridge cables, and the horizontal and vertical angles of each cable relative to the radar installation point were obtained to establish a cable coordinate database. When the system starts up, the radar pan-tilt unit is controlled to the preset zero point position, and the system self-calibrates using a known reference point to keep the accuracy of angle measurement within ±0.05°. Based on the cable coordinate database, the dual-axis servo gimbal is controlled to rotate to the macroscopic angle range of the target cable, and the radar line of sight is coarsely aligned with the area where the target cable is located. During this stage, the gimbal rotation speed is controlled within 5° per second. The encoder provides real-time feedback on the gimbal position, and macro-area recognition is performed when the gimbal rotates to the target cable macro-area.

[0006] Furthermore, in the aforementioned bridge cable tension monitoring method based on hybrid beam control scanning radar, the improved YOLOv5 model is used as the cable target detection algorithm. An attention mechanism and multi-scale feature fusion are introduced to establish the cable target detection model, including: An improved YOLOv5 object detection model was constructed and optimized for the slender shape of the cable. In the backbone network CSPDarknet53, some convolutional layers were replaced with convolutional layers with self-attention mechanism to enhance the model's ability to capture long-distance dependencies of the cable. The standard convolutions in each convolutional block are replaced with self-attention convolutions, and the importance of features is dynamically adjusted by calculating the similarity between different locations in the feature map. An adaptive feature fusion layer is added to the FPN feature pyramid network, and a combination of channel attention mechanism and spatial attention mechanism is used to fuse feature maps of different scales.

[0007] Furthermore, in the aforementioned bridge cable tension monitoring method based on hybrid beam control scanning radar, the step of scanning the macroscopic area using DBF technology and inputting the data into the cable target detection model to obtain the precise location of the cable includes: Digital beamforming (DBF) technology is activated to perform high-resolution scanning of macroscopic areas. By weighting and phase-controlling the received signals, a digital beam pointing in a specific direction is formed, generating a high-resolution scan image. The scanned image is input into the cable target detection model, which performs feature extraction and target detection on the input image and outputs the precise coordinates and size parameters of the cable.

[0008] Furthermore, in the aforementioned bridge cable tension monitoring method based on hybrid beam control scanning radar, the step of dynamically adjusting the beamwidth according to the precise position of the cable to generate the optimal beam pattern and maximize the signal-to-noise ratio, thereby obtaining an optimized beam, includes: The beamwidth is dynamically adjusted according to the cable width. For cables with a width of less than 5cm, the beamwidth is set to 0.5°; for cables with a width of 5-10cm, the beamwidth is set to 1°; and for cables with a width greater than 10cm, the beamwidth is set to 1.5°. The DQN deep reinforcement learning algorithm transforms beam parameter adjustment into a Markov decision process. The state includes the current beam parameters, signal-to-noise ratio, and environmental interference level, while the actions include beamwidth, pointing angle, and shape adjustment, resulting in an optimized beam.

[0009] Furthermore, in the above-mentioned bridge cable tension monitoring method based on hybrid beam control scanning radar, the step of acquiring the time-domain signal of cable vibration based on the optimized beam, performing spectrum analysis, and calculating the cable tension of the bridge includes: The cable vibration time-domain signal was obtained by continuously acquiring signals from the cable for 120 seconds using an optimized beam and recording the signal at a sampling rate of 10 kHz. The acquired time-domain signal is preprocessed by digital signal processing, including filtering and detrending, and then subjected to spectrum analysis. The time-domain signal is converted into a frequency-domain signal using FFT (Fast Fourier Transform) to obtain the cable force.

[0010] Furthermore, in the aforementioned bridge cable tension monitoring method based on hybrid beam control scanning radar, the step of analyzing the changing trend of the cable tension using an LSTM long short-term memory network to generate early warning information includes: The long-term dependency of cable force changes is captured by LSTM units, and three warning levels are set according to the trend of cable force change: minor anomaly, moderate anomaly and severe anomaly. When abnormal changes are detected, an early warning message is automatically generated, including the abnormal cable number, the degree of abnormality, the change trend graph, and the recommended maintenance time.

[0011] Furthermore, in the bridge cable tension monitoring system based on hybrid beam control scanning radar, the bridge cable tension monitoring system includes the following modules: The radar sensor setting module is used to acquire the macroscopic coordinates of the target cable, return the gimbal to zero for system self-calibration, control the rotation of the gimbal, and coarsely align the radar line of sight to the macroscopic area where the target cable is located. The detection model building module is used to build a cable target detection model based on the improved YOLOv5 model as the cable target detection algorithm, by introducing an attention mechanism and multi-scale feature fusion. The cable location determination module is used to scan a macroscopic area using DBF technology and input the data into the cable target detection model to obtain the precise location of the cable. The beam optimization control module is used to dynamically adjust the beamwidth according to the precise position of the cable, generate the optimal beam pattern and maximize the signal-to-noise ratio to obtain an optimized beam. The cable force calculation module is used to collect the time-domain signal of cable vibration based on the optimized beam, perform spectrum analysis, and calculate the cable force of the bridge. The cable force signal processing module is used to analyze the changing trend of the cable force through an LSTM long short-term memory network and generate early warning information.

[0012] Furthermore, in the bridge cable tension monitoring system based on hybrid beam control scanning radar, the cable tension calculation module includes the following sub-modules: The acquisition submodule is used to continuously acquire signals from the cable for 120 seconds using an optimized beam, and to record the signal at a sampling rate of 10kHz to obtain the time-domain signal of the cable vibration. The conversion submodule is used to preprocess the acquired time-domain signal through digital signal processing, including filtering, detrending processing, and then performing spectrum analysis. The FFT (Fast Fourier Transform) is used to convert the time-domain signal into a frequency-domain signal to obtain the cable force.

[0013] Furthermore, in the bridge cable tension monitoring system based on hybrid beam control scanning radar, the cable tension signal processing module includes the following sub-modules: The processing submodule is used to capture the long-term dependency of cable force changes through LSTM units and set three warning levels based on the cable force change trend, including minor anomaly, moderate anomaly and severe anomaly. The early warning submodule is used to automatically generate early warning information when abnormal changes are detected, including the abnormal cable number, the degree of abnormality, the change trend graph, and the recommended maintenance time.

[0014] Its beneficial effects include high measurement accuracy: electronic fine-tuning eliminates mechanical positioning errors, and adaptive beamforming effectively suppresses background clutter, significantly improving the signal-to-noise ratio and accuracy of vibration frequency extraction. Before beamforming, radar indiscriminately monitors the environment, resulting in a large amount of noise in the recovered signal. Simultaneously, the signal strength in the target direction is low, which may prevent effective capture of vibration signals from distant bridge cables, hindering high-quality detection of all bridge cable vibrations. Introducing beamforming can, to some extent, avoid interference from false targets in the environment, allowing the detection signal to initially focus on the target bridge cable, providing a certain degree of spatial selectivity. However, limitations remain, such as a large sidelobe signal radiation range and the bridge cable size being smaller than the interference sources in the background. To address these shortcomings, adaptive beamforming can further enhance the spatial selectivity of the radar signal, effectively improve the main lobe signal strength, create a large-scale null effect, effectively filter out noise, and achieve accurate capture of the target bridge cable. High automation and efficiency: It achieves fully automated, unattended measurement; a single device can complete the monitoring of the entire bridge cable force, with efficiency far exceeding that of manual methods. High reliability: The hybrid architecture combines the wide coverage of mechanical scanning with the agility and precision of electronic scanning, resulting in higher system robustness and better long-term stability. Good cost-effectiveness: Compared to installing a large number of sensors or multiple fixed radars, this invention significantly reduces costs and simplifies maintenance. Attached Figure Description

[0015] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention.

[0016] Figure 1 This is a schematic diagram of the first embodiment of the bridge cable tension monitoring method based on hybrid beam control scanning radar in this invention. Figure 2 This is a schematic diagram of the second embodiment of the bridge cable tension monitoring method based on hybrid beam control scanning radar in this invention. Figure 3 This is a schematic diagram of the first embodiment of the bridge cable tension monitoring system based on hybrid beam control scanning radar in this invention. Figure 4 This is a schematic diagram of the second embodiment of the bridge cable tension monitoring system based on hybrid beam control scanning radar in this invention. Figure 5 This is a beamforming diagram of a bridge cable tension monitoring system based on hybrid beam control scanning radar in an embodiment of the present invention; Figure 6 This is a beamforming heat map of a bridge cable tension monitoring system based on a hybrid beam control scanning radar, as described in an embodiment of the present invention. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0018] Those skilled in the art will understand that, unless specifically stated otherwise, the singular forms "one," "an," and "this" used herein may also include the plural forms. It should be further understood that the terminology used in this specification includes the presence of features, integers, steps, operations, elements, and / or components, but does not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0019] The present invention will now be described in detail with reference to the accompanying drawings. Figure 1 As shown, a method for monitoring bridge cable tension based on hybrid beam control scanning radar includes the following steps: Step 101: Obtain the macroscopic coordinates of the target cable, return the gimbal to zero for system self-calibration, control the gimbal rotation, and coarsely align the radar line of sight to the macroscopic area where the target cable is located. Specifically, this embodiment also includes using a high-precision total station to measure the spatial coordinates of the bridge cables, obtaining the horizontal and vertical angles of each cable relative to the radar installation point, and establishing a cable coordinate database. When the system starts, the radar pan-tilt unit is controlled to a preset zero point position, and the system self-calibrates using a known reference point to ensure that the accuracy error of the angle measurement is controlled within ±0.05°. Based on the cable coordinate database, the dual-axis servo pan-tilt unit is controlled to rotate to the macroscopic angle range of the target cable, and the radar line of sight is coarsely aligned with the area where the target cable is located. During this stage, the pan-tilt unit rotation speed is controlled within 5° per second. The position of the pan-tilt unit is fed back in real time through an encoder, and macroscopic area recognition is performed after the pan-tilt unit rotates to the macroscopic area of ​​the target cable.

[0020] In the deployment phase of the bridge cable health monitoring system, accurately locating the macroscopic spatial position of the target cable is the foundation for all subsequent detection processes. The core objective of this step is to provide the radar system with a clear "initial aiming direction" to avoid subsequent detection from falling into invalid scans due to positioning deviations. First, a cable coordinate database needs to be established using high-precision measuring equipment. Considering that bridge cables are usually numerous, ranging from dozens to hundreds, widely distributed across hundreds of meters, and that there is a fixed spatial relationship between the radar installation point and the cables, the system selects a Leica TS60 high-precision total station with an angle measurement accuracy of ±0.5 seconds and a distance measurement accuracy of ±(0.6mm+1ppm×D) as the coordinate acquisition tool. Measurements must be taken under windless conditions with visibility ≥5km and ambient temperature fluctuations ≤2℃ to minimize interference from external factors on measurement accuracy. The measurement process uses the radar installation point as the origin to establish a three-dimensional Cartesian coordinate system, with the x-axis pointing longitudinally to the bridge, the y-axis pointing laterally, and the z-axis perpendicular to the ground. The spatial coordinates of the anchor points at both ends of each cable are measured. Then, the azimuth angle of the midpoint of each cable relative to the radar, the rotation angle around the z-axis (range 0°-360°), and the pitch angle around the x-axis (range -90°-90°) are calculated through spatial geometry. These coordinate data are entered into an encrypted cable coordinate database, and a monthly remeasurement and update are performed to compensate for minor coordinate shifts caused by temperature deformation or foundation settlement of the bridge.

[0021] After system startup, the primary task is to complete the zero-point return and self-calibration of the gimbal. The radar gimbal adopts a dual-axis servo structure, with the azimuth axis responsible for horizontal rotation and the pitch axis responsible for vertical rotation. Its "zero point" must simultaneously satisfy the consistency of both the mechanical and electronic zero points: the mechanical zero point is the mechanical limit position calibrated at the factory, while the electronic zero point is calibrated using a preset reference control point on the bridge, usually a stainless steel observation marker on the top of the bridge pier, with a coordinate stability error ≤0.1mm / year. During calibration, the gimbal first automatically rotates to the mechanical zero point, then the radar emits a low-power detection signal to align with the reference point. The angle of the actual received signal is compared with the theoretical angle of the reference point in the database, the deviation value is calculated and automatically compensated, ultimately controlling the angle measurement error within ±0.05°.

[0022] After calibration, the gimbal control module, based on an embedded STM32H743 processor with a response latency ≤10ms, calls the cable coordinate database and drives a dual-axis servo motor to rotate the radar. To avoid overshoot due to inertia, the motor speed is strictly controlled to within 5° per second. Simultaneously, a 1024-line photoelectric encoder with an angular resolution of 0.0879° collects the gimbal's current angle in real time, feeding back position data to the control module every 100ms, forming a closed-loop control. When the gimbal rotates into the target cable's "macro-angle range," i.e., the theoretical coordinate ±15° area, the control module triggers a stop command.

[0023] Step 102: Based on the improved YOLOv5 model as the cable target detection algorithm, an attention mechanism and multi-scale feature fusion are introduced to establish a cable target detection model; Specifically, this embodiment also includes constructing an improved YOLOv5 object detection model, optimizing it for the slender characteristics of the cable. In the backbone network CSPDarknet53, some convolutional layers are replaced with convolutional layers with self-attention mechanism to enhance the model's ability to capture long-distance dependencies of the cable. The standard convolutions in each convolutional block are replaced with self-attention convolutions, and the importance of features is dynamically adjusted by calculating the similarity between different positions in the feature map. An adaptive feature fusion layer is added to the FPN feature pyramid network, and a combination of channel attention mechanism and spatial attention mechanism is used to fuse feature maps of different scales.

[0024] In bridge cable detection scenarios, traditional target detection models, such as YOLOv3 and Faster R-CNN, often face two major challenges: First, cables are elongated and slender, with aspect ratios exceeding 10:1 in images, making them easily confused with other slender structures like bridge railings and steel trusses; second, cables occupy a small portion of radar scan images, with distant cables taking up only 2%-5% of the image pixels, and are easily affected by background elements such as sky, trees, and rain reflections, resulting in an accuracy rate of less than 85%. Therefore, this system is based on the YOLOv5s model and features targeted improvements.

[0025] One of the core improvements is the introduction of self-attention convolutional layers into the backbone network CSPDarknet53. The original CSPDarknet53 structure reduces computation through "cross-stage local connections," but it has a weak ability to capture long-range feature dependencies of slender targets. For example, the continuous features of a cable from the upper to the lower anchor point in an image are easily segmented into multiple "segments" due to background interference. The improved scheme replaces the standard convolutional layers 3, 5, and 7 of CSPDarknet53 with self-attention convolutional layers. This layer first splits the input feature map, such as a 128×128×64 dimension, into three matrices: Query and Key, and Value. It then calculates the cosine similarity between the Query and Key to obtain an attention weight matrix of dimension 128×128. The weight matrix is ​​then multiplied by the Value matrix to dynamically enhance the feature response of the cable region (weight value ≥ 0.8) while suppressing features in the background region (weight value ≤ 0.2). For example, when the cable in an image is partially obscured by leaves, the self-attention mechanism can "associate" the cable outline of the obscured area using features of the unobscured area, thus avoiding misjudging the complete cable as a multi-segment broken structure.

[0026] During the model training phase, a transfer learning strategy was employed to shorten the training cycle and improve generalization ability. First, YOLOv5s weights pre-trained on the ImageNet dataset were loaded. Then, fine-tuning was performed using a dedicated dataset of 1000 bridge cable images, covering three typical scenarios: 1) 600 images of cables from different bridge types (cable-stayed and suspension bridges); 2) 300 images of cables under different environmental conditions (sunny, cloudy, rainy, and nighttime lighting); and 300 images of cables under different occlusion conditions (leaf occlusion, billboard occlusion, and localized rust). All images were labeled with the cable bounding boxes using the LabelImg tool and divided into training, validation, and test sets in an 8:1:1 ratio. The training process was based on the PyTorch framework, using CIoULoss as the bounding box regression loss function. CIoULoss achieved 15% higher bounding box fitting accuracy for slender targets than IoULoss. The optimizer was Adam, with an initial learning rate of 0.001, warmed up to 0.01 in the first 10 epochs, and then decayed using cosine annealing. The batch size was set to 16, and the training run consisted of 100 epochs. The final training results showed that the improved YOLOv5 model achieved a 98.7% accuracy rate in identifying cables, with a false positive rate (misidentifying guardrails and steel cables as cables) of less than 1.3%. The single-frame processing speed remained at 18ms, fully meeting the requirements for real-time detection.

[0027] Step 103: Scan the macroscopic area using DBF technology and input it into the cable target detection model to obtain the precise location of the cable; specifically, this embodiment also includes activating digital beamforming (DBF) technology to perform high-resolution scanning of the macroscopic area, forming a digital beam pointing in a specific direction by weighting and phase control of the received signal, and generating a high-resolution scan image; input the scan image into the cable target detection model, perform feature extraction and target detection on the input image, and output the precise coordinates and size parameters of the cable.

[0028] After the gimbal is coarsely aligned, the system needs to use digital beamforming (DBF) technology to perform high-resolution scanning of the macroscopic area, providing high-quality input images for the cable target detection model. Compared with traditional mechanical scanning, DBF technology has the advantages of fast scanning speed, high resolution, and the ability to form multiple beams simultaneously. It can complete a full-coverage scan of the ±15° macroscopic area within 3 seconds, avoiding missed scans caused by the slight swaying of the cable and the wind-induced vibration amplitude, which is usually 5-10cm.

[0029] The core of the DBF system is a linear array of 16 antenna elements. Each antenna element is a microstrip patch antenna, operating at a frequency of 10 GHz with a bandwidth of 1 GHz. The element spacing is set to half a wavelength; since the wavelength of a 10 GHz signal is 3 cm, the element spacing is 1.5 cm. This spacing effectively avoids false beams in the array grating lobes and sidelobes, ensuring beam pointing accuracy. The 16 antenna elements are arranged in a straight line, with a total array length of 24 cm, and are connected to the radar transceiver via RF cables. During scanning, the radar transceiver first assigns different weighting coefficients to each antenna element, calculated from the beam pointing angle. Chebyshev weighting is used to suppress sidelobes. Then, a high-frequency pulse signal is transmitted synchronously with a pulse width of 100 ns and a repetition frequency of 1 kHz. When the signal encounters a target such as a cable, the resulting echo signal is synchronously received by each antenna element. After being amplified by a low-noise amplifier (noise figure ≤ 2 dB), the analog signal is converted into a digital signal by a 12-bit high-speed AD converter with a sampling rate of 500 MHz.

[0030] In the digital domain, the DBF processing unit, based on the FPGA chip XC7K325T, performs phase compensation and amplitude weighting on 16 digital signals with a parallel computing capability of 200Gbps: by adjusting the phase of each signal, the cable echo signals received by all antenna elements are superimposed in phase in a specified direction to form a "digital beam" pointing in that direction; at the same time, the sidelobe signal is suppressed by amplitude weighting, with the sidelobe level ≤-25dB, reducing background interference. The system performs step scanning with an angular resolution of 0.5°. This resolution is calculated using the antenna array resolution formula θ=λ / (N×d), where λ is the wavelength, N is the number of elements, and d is the spacing. A resolution of 0.5° means that one beam of data is generated for every 0.5° scan. A total of 61 beams are needed to cover a ±15° area. Each beam corresponds to one pixel in the scanned image. The pixel value is the signal strength received by the beam, with a grayscale value of 0-255. This results in a 61×41 pixel high-resolution grayscale image corresponding to a vertical ±10° scanning range. The cable area in the image exhibits a high grayscale value (180-255) due to strong metal reflection, while the background area has a low grayscale value (0-80).

[0031] After the scanned image is generated, it is immediately transmitted to the embedded terminal via Gigabit Ethernet and input into the improved YOLOv5 object detection model. The model first preprocesses the image: resizes the image to 640×640 pixels, applies Mosaic data augmentation, performs random cropping, flipping, and color gamut shifting to improve robustness, and then normalizes the pixel values ​​to the 0-1 range. Subsequently, the model extracts cable features through the backbone network, enhances the slender features through a multi-scale fusion module, and finally outputs the pixel coordinates of the cable, x, y, with the upper left corner of the image as the origin, the x-axis horizontally to the right, and the y-axis vertically downwards along the bounding box size, w, h. The system converts pixel coordinates into radar angles using a "camera calibration matrix." This matrix, obtained from previous calibration experiments, establishes a mapping relationship between pixel positions and radar azimuth and elevation angles. For example, the image center corresponds to the current angle of the gimbal, and each pixel corresponds to an angle offset of 0.01°. Therefore, the precise angle of the cable relative to the radar can be calculated using x and y, achieving a positioning accuracy of ±0.1. Simultaneously, the bounding box width w output by the model is converted into the actual width of the cable using a scale bar. For example, when the cable is 100 meters away from the radar, one pixel corresponds to 0.2 meters, and w=5 means the actual width is 1 meter. This provides key parameters for the next step of dynamically adjusting the beamwidth.

[0032] Step 104: Dynamically adjust the beamwidth according to the precise position of the cable to generate the optimal beam pattern and maximize the signal-to-noise ratio, thus obtaining the optimized beam. Specifically, this embodiment also includes dynamically adjusting the beamwidth based on the cable width. For cables with a width less than 5cm, the beamwidth is set to 0.5°; for cables with a width between 5-10cm, the beamwidth is set to 1°; and for cables with a width greater than 10cm, the beamwidth is set to 1.5°. The beam parameter adjustment is transformed into a Markov decision process using the DQN deep reinforcement learning algorithm. The states include the current beam parameters, signal-to-noise ratio, and environmental interference level, while the actions include beamwidth, pointing angle, and shape adjustment, resulting in an optimized beam.

[0033] Based on the precise location and size information of the cables, the system initiates an adaptive beam optimization algorithm. The algorithm models the beamwidth adjustment problem as an optimization problem, aiming to maximize the signal-to-noise ratio (SNR). The system dynamically adjusts the beamwidth according to the cable width: for cables less than 5cm wide, the beamwidth is set to 0.5°; for cables between 5-10cm wide, the beamwidth is set to 1°; and for cables greater than 10cm wide, the beamwidth is set to 1.5°. Simultaneously, the system calculates the optimal beam shape using deep reinforcement learning (DQN) algorithm. The algorithm transforms beam parameter adjustment into a Markov decision process, where the states include current beam parameters, SNR, and environmental interference level, and the actions include beamwidth, pointing angle, and shape adjustment. By dynamically adjusting the beam parameters in real-time based on SNR and interference levels, the system improves the SNR by more than 35%. The optimized beam is generated by a digital beamforming system, forming a beam pattern where the main lobe is aligned with the cable and the side lobes suppress environmental interference, ensuring that the radar signal is effectively focused on the cable target and providing high-quality input for subsequent signal acquisition.

[0034] Step 105: Based on the optimized beam acquisition of the time-domain signal of cable vibration, perform spectrum analysis and calculate the cable force of the bridge; Specifically, this embodiment also includes using an optimized beam to continuously acquire signals from the cable for 120 seconds, recording the signal at a sampling rate of 10kHz to obtain the time-domain signal of the cable vibration; the acquired time-domain signal is preprocessed by digital signal processing, including filtering and detrending processing, and then subjected to spectrum analysis; the time-domain signal is converted into a frequency-domain signal by using FFT (Fast Fourier Transform) to obtain the cable force.

[0035] The system uses optimized beamforming to continuously acquire signals from the cable for 120 seconds, obtaining the time-domain signal of the cable vibration. During the acquisition, the system records the signal at a sampling rate of 10kHz to ensure complete capture of the vibration frequency. The acquired time-domain signal undergoes preprocessing through digital signal processing, including filtering, high-frequency noise removal, and detrending. Subsequently, spectral analysis is performed using Fast Fourier Transform (FFT) to convert the time-domain signal into a frequency-domain signal, obtaining a spectrum. The system automatically identifies peaks in the spectrum to determine the cable's natural frequency. According to string vibration theory, the cable force calculation formula is: F = (π²L²f²m) / n², where F is the cable force (N), L is the cable length (m), f is the natural frequency (Hz), m is the mass per unit length of the cable (kg / m), and n is the vibration mode order, typically taken as 1. The system automatically extracts the cable length L from the initial coordinate database, the mass per unit length (m), the vibration mode order (n) from bridge design data, and combines this with the measured natural frequency (f) to calculate the cable force. The calculation accuracy is controlled within ±3%, which meets the needs of engineering applications.

[0036] Step 106: Analyze the changing trend of cable tension using an LSTM (Long Short-Term Memory) network to generate early warning information.

[0037] Specifically, this embodiment also includes capturing the long-term dependency of cable force changes through LSTM units, setting three warning levels based on the cable force change trend, including minor anomaly, moderate anomaly, and severe anomaly; when an abnormal change is detected, warning information is automatically generated, including the abnormal cable number, the degree of anomaly, the change trend graph, and the recommended maintenance time.

[0038] The system constructs an LSTM (Long Short-Term Memory) network model to analyze the temporal changes in cable force. The model input is historical cable force data, specifically the time series of force for each cable. The LSTM units capture the long-term dependencies in force changes. The model consists of two LSTM layers, each with 128 units, followed by a fully connected layer. The LSTM model is trained using cable force monitoring data from the past five years to learn normal patterns in force changes. During training, the Adam optimizer is used with a learning rate of 0.001 and a batch size of 32. The model optimizes parameters through backpropagation to minimize prediction error. In practical applications, the LSTM model analyzes the cable force data for each cable in real time, predicting the force change trend for the next 24 hours. The system sets three warning levels: minor anomaly (change rate 5-10%), moderate anomaly (change rate 10-20%), and severe anomaly (change rate >20%). When an abnormal change is detected, the system automatically generates warning information, including the abnormal cable number, the severity of the anomaly, a trend chart, and a suggested maintenance time. The early warning information is transmitted to the bridge maintenance center in real time via the 5G network. Simultaneously, the system generates a maintenance recommendation report, including an analysis of possible causes of the anomaly and a suggested inspection schedule. The early warning accuracy rate reaches 92.5%, providing ample time for bridge maintenance and significantly improving bridge safety management.

[0039] Its beneficial effects include high measurement accuracy: electronic fine-tuning eliminates mechanical positioning errors, and adaptive beamforming effectively suppresses background clutter, significantly improving the signal-to-noise ratio and accuracy of vibration frequency extraction. Before beamforming, radar indiscriminately monitors the environment, resulting in a large amount of noise in the recovered signal. Simultaneously, the signal strength in the target direction is low, which may prevent effective capture of vibration signals from distant bridge cables, hindering high-quality detection of all bridge cable vibrations. Introducing beamforming can, to some extent, avoid interference from false targets in the environment, allowing the detection signal to initially focus on the target bridge cable, providing a certain degree of spatial selectivity. However, limitations remain, such as a large sidelobe signal radiation range and the bridge cable size being smaller than the interference sources in the background. To address these shortcomings, adaptive beamforming can further enhance the spatial selectivity of the radar signal, effectively improve the main lobe signal strength, create a large-scale null effect, effectively filter out noise, and achieve accurate capture of the target bridge cable. High automation and efficiency: It achieves fully automated, unattended measurement; a single device can complete the monitoring of the entire bridge cable force, with efficiency far exceeding that of manual methods. High reliability: The hybrid architecture combines the wide coverage of mechanical scanning with the agility and precision of electronic scanning, resulting in higher system robustness and better long-term stability. Good cost-effectiveness: Compared to installing a large number of sensors or multiple fixed radars, this invention significantly reduces costs and simplifies maintenance.

[0040] Please see Figure 2 In the bridge cable tension monitoring method based on hybrid beam control scanning radar, the improved YOLOv5 model is used as the cable target detection algorithm. Attention mechanism and multi-scale feature fusion are introduced to establish the cable target detection model, which includes the following steps: Step 201: Construct an improved YOLOv5 object detection model, which is optimized for the slender shape of the cable. In the backbone network CSPDarknet53, some convolutional layers are replaced with convolutional layers with self-attention mechanism to enhance the model's ability to capture long-distance dependencies of the cable. Step 202: Replace the standard convolutions in each convolutional block with self-attention convolutions, and dynamically adjust the importance of features by calculating the similarity between different locations in the feature map; Step 203: Add an adaptive feature fusion layer to the FPN feature pyramid network, and use a combination of channel attention mechanism and spatial attention mechanism to fuse feature maps of different scales.

[0041] The above describes embodiments of the bridge cable tension monitoring method based on hybrid beam control scanning radar of the present invention. Please refer to [link / reference]. Figure 3 In a bridge cable tension monitoring system based on hybrid beam control scanning radar, the system includes the following modules: The radar sensor setting module is used to acquire the macroscopic coordinates of the target cable, return the gimbal to zero for system self-calibration, control the rotation of the gimbal, and coarsely align the radar line of sight to the macroscopic area where the target cable is located. The detection model building module is used to build a cable target detection model based on the improved YOLOv5 model as the cable target detection algorithm, by introducing an attention mechanism and multi-scale feature fusion. The cable location determination module is used to scan a macroscopic area using DBF technology and input the data into the cable target detection model to obtain the precise location of the cable. The beam optimization control module is used to dynamically adjust the beamwidth according to the precise position of the cable, generate the optimal beam pattern and maximize the signal-to-noise ratio to obtain an optimized beam. The cable force calculation module is used to perform spectrum analysis and calculate the cable force of the bridge based on the time-domain signal of cable vibration acquired by optimized beam acquisition. The cable force signal processing module is used to analyze the changing trend of cable force through an LSTM long short-term memory network and generate early warning information.

[0042] Specifically, the system of the present invention also includes the following units: Radar sensing unit: Employs a phased array antenna or an antenna array supporting digital beamforming (DBF), operating in the millimeter-wave band (such as 24 GHz or 77 GHz), and has high displacement resolution.

[0043] Hybrid Beam Control Unit: Mechanical scanning component (gimbal): High-precision dual-axis servo gimbal, carrying radar sensing unit, to achieve a wide range of azimuth and elevation angle rotation (first-level beam control).

[0044] Electronic scanning component: Composed of phase shifter network, DBF processor, etc., it is used to electronically fine-tune the beam pointing and adaptively optimize the beam shape (secondary beam control) based on mechanical scanning positioning.

[0045] Control and processing unit: Used for basic processing of the collected data.

[0046] Scanning path planning module: Pre-stores the macroscopic coordinates of each cable and generates the optimal rotation path of the gimbal. It then performs cyclic scanning according to the set scanning cycle.

[0047] Beam optimization control module: Calculates and issues beam pointing fine-tuning commands and beamforming parameters based on the distance, azimuth, and physical dimensions of the target cable.

[0048] Signal processing and cable force calculation module: Performs FFT and other spectral analysis on the acquired vibration signal to extract the natural frequency, and calculates the cable force based on the string vibration theory formula.

[0049] Auxiliary units include corner reflectors for system calibration, environmental sensors, power supply, and communication modules.

[0050] Specifically, the method of the present invention includes the following steps: (1) System initialization and calibration: calibrate the macroscopic coordinates of each cable, return the gimbal to zero, and perform system self-calibration.

[0051] (2) Mechanical coarse positioning: Control the gimbal to rotate and roughly align the radar line of sight to the macroscopic area where the target cable is located.

[0052] (3) Electronic fine-tuning and shaping: ① Precise beam pointing: Initiate electronic scanning, use DBF technology to quickly scan within the coarse aiming area, identify the precise position of the cable, and drive the beam center to precisely align with the cable without inertia. ② Beam adaptive optimization: Based on the current cable distance and diameter, dynamically adjust the beam width (e.g., narrow the beam) to generate the optimal beam pattern and maximize the signal-to-noise ratio.

[0053] (4) Data acquisition and processing: Under the optimized beam direction, the time domain signal of cable vibration is acquired, the spectrum analysis is performed and the cable force is calculated.

[0054] (5) Rotation monitoring: Determine whether it is the last cable. If not, control the gimbal to rotate to the macroscopic area of ​​the next cable and repeat steps 3-4.

[0055] Please see Figure 4 In a bridge cable tension monitoring system based on hybrid beam control scanning radar, taking a large cable-stayed bridge (with a total of 24 cables) as an example, Deployment: Install the system of this invention on the maintenance platform or crossbeam at the top of the bridge tower so that the radar field of view can cover the entire bridge.

[0056] Calibration: Use a total station or similar equipment to accurately measure the macroscopic azimuth and elevation angles of each cable relative to the radar mounting point, and input them into the system database. Simultaneously input the length L and linear density μ of each cable.

[0057] Scanning and monitoring: The system starts automatically at 2 AM every day.

[0058] The gimbal rotates along the optimal path, roughly aiming the radar at cable number 1.

[0059] The phased array system is activated, performs electronic scanning within a cone angle range of ±3°, quickly locks onto the center of cable 1, and dynamically narrows the beamwidth from 3° to 0.8°.

[0060] 120 seconds of vibration data were collected, processed in real time, and the cable force value T1 was obtained.

[0061] After completion, the gimbal is rotated to roughly aim at cable number 2, and the electronic fine-tuning and measurement process is repeated.

[0062] The entire process requires no human intervention and can complete the measurement of all 100 cords in about 3 hours.

[0063] Data Management: All cable data, spectrum diagrams, and system status information are uploaded to the cloud monitoring platform via the 5G network to generate trend reports and early warning information.

[0064] Please see Figure 5 This is a beamforming diagram of a bridge cable tension monitoring system based on hybrid beam control scanning radar.

[0065] Please see Figure 6 Beamforming heatmap in a bridge cable tension monitoring system based on hybrid beam control scanning radar.

[0066] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely preferred examples and are not intended to limit the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A bridge cable force monitoring method based on a hybrid beam control scanning radar, characterized in that, The bridge cable force monitoring method comprises the following steps: Obtain the macroscopic coordinates of the target cable, perform system self-calibration on the gimbal return zero point, control the gimbal rotation, and coarsely align the radar visual axis to the macroscopic region where the target cable is located; Based on the improved YOLOv5 model as the cable target detection algorithm, the attention mechanism and multi-scale feature fusion are introduced, and the cable target detection model is established; The macroscopic region is scanned by the DBF technology and input into the cable target detection model to obtain the accurate position of the cable; According to the accurate position of the cable, the beam width is dynamically adjusted, the best mode of the beam is generated, and the signal-to-noise ratio is maximized to obtain the optimized beam; Based on the optimized beam, the cable vibration time domain signal is collected, the frequency spectrum is analyzed, and the cable force of the bridge is calculated; The change trend of the cable force is analyzed by the LSTM long short-term memory network, and the warning information is generated. 2.The hybrid beam control scanning radar based bridge cable force monitoring method according to claim 1, wherein, The macroscopic coordinates of the target cable are obtained, the gimbal return zero point is calibrated, the gimbal is controlled to rotate, and the radar visual axis is coarsely aligned to the macroscopic region where the target cable is located, comprising: Use a high-precision total station to measure the spatial coordinates of the bridge cable, obtain the horizontal angle and vertical angle of each cable relative to the radar installation point, and establish a cable coordinate database; When the system starts, control the radar gimbal to the preset zero point position, perform system self-calibration through the known reference point, and control the accuracy error of angle measurement to be within ±0.05°; According to the cable coordinate database, control the double-shaft servo gimbal to rotate to the macroscopic angle range of the target cable, coarsely align the radar visual axis to the region where the target cable is located, and control the gimbal rotation speed to be within 5° per second in this stage; Real-time feedback of the gimbal position is realized through the encoder, and when the gimbal rotates to the macroscopic region of the target cable, the macroscopic region is identified. 3.The hybrid beam control scanning radar based bridge cable force monitoring method of claim 1, wherein, Based on the improved YOLOv5 model as the cable target detection algorithm, the attention mechanism and multi-scale feature fusion are introduced, and the cable target detection model is established, comprising: An improved YOLOv5 target detection model is constructed, which is optimized for the elongated feature of the cable. In the backbone network CSPDarknet53, some convolutional layers are replaced with convolutional layers with self-attention mechanism Self-Attention, enhancing the model's ability to capture long-distance dependencies of the cable; Replace the standard convolution in each convolution block with a self-attention convolution, dynamically adjust the importance of the features by calculating the similarity between different positions in the feature map; An adaptive feature fusion layer is added to the FPN feature pyramid network, and the channel attention mechanism and spatial attention mechanism are combined to fuse feature maps of different scales. 4.The hybrid beam control scanning radar based bridge cable force monitoring method of claim 1, wherein, The macroscopic region is scanned by the DBF technology and input into the cable target detection model to obtain the accurate position of the cable, comprising: Start the digital beam forming DBF technology to scan the macroscopic region at high resolution, form a digital beam pointing to a specific direction by weighting and phase control on the received signal, and generate a high-resolution scan image; The scanned image is input into a cable target detection model to perform feature extraction and target detection on the input image, and the precise coordinates and size parameters of the cable are output. 5.The hybrid beam control scanning radar based bridge cable force monitoring method of claim 1, wherein, The beam width is dynamically adjusted according to the precise position of the cable, the optimal mode of the beam is generated, and the signal-to-noise ratio is maximized to obtain an optimized beam, including: The beam width is dynamically adjusted according to the cable width, the beam width is set to 0.5° for a cable with a width less than 5 cm, the beam width is set to 1° for a cable with a width between 5-10 cm, and the beam width is set to 1.5° for a cable with a width greater than 10 cm; The beam parameter adjustment is converted into a Markov decision process through a DQN deep reinforcement learning algorithm, the state includes the current beam parameter, the signal-to-noise ratio, and the environmental interference level, the action includes the beam width, the pointing angle, and the shape adjustment, and the optimized beam is obtained. 6.The hybrid beam control scanning radar based bridge cable force monitoring method of claim 1, wherein, The cable vibration time domain signal is collected based on the optimized beam, frequency spectrum analysis is performed, and the cable force of the bridge is calculated, including: The cable is continuously signal collected for 120 seconds using the optimized beam, and the signal is recorded at a sampling rate of 10 kHz to obtain the time domain signal of the cable vibration; The collected time domain signal is preprocessed through digital signal processing, including filtering and detrending, and then frequency spectrum analysis is performed, the time domain signal is converted into a frequency domain signal through FFT fast Fourier transform, and the cable force is obtained.

7. The hybrid beam control scanning radar based bridge cable force monitoring method of claim 1, wherein, The change trend of the cable force is analyzed through an LSTM long short-term memory network, and warning information is generated, including: The LSTM unit captures the long-term dependence of the cable force change, three warning levels are set according to the change trend of the cable force, including slight abnormality, moderate abnormality, and severe abnormality; When an abnormal change is detected, warning information is automatically generated, including the abnormal cable number, the abnormality degree, the change trend graph, and the recommended maintenance time.

8. The bridge cable force monitoring system based on hybrid beam control scanning radar, characterized in that, The bridge cable force monitoring system includes the following modules: A radar sensor setting module is used to obtain the macro coordinates of the target cable, perform system self-calibration on the gimbal return zero point, control the gimbal rotation, and roughly align the radar visual axis to the macro area where the target cable is located. A detection model establishment module is used to establish a cable target detection model based on an improved YOLOv5 model as a cable target detection algorithm, introduce an attention mechanism and multi-scale feature fusion. A cable position determination module is used to scan the macro area through DBF technology and input into the cable target detection model to obtain the precise position of the cable. A beam optimization control module is used to dynamically adjust the beam width according to the precise position of the cable, generate the optimal mode of the beam, and maximize the signal-to-noise ratio to obtain an optimized beam. A cable force calculation module is used to collect the cable vibration time domain signal based on the optimized beam, perform frequency spectrum analysis, and calculate the cable force of the bridge. A cable force signal processing module is used to analyze the change trend of the cable force through an LSTM long short-term memory network and generate warning information.

9. The hybrid beam control scanning radar based bridge cable force monitoring system of claim 8, wherein, The cable force calculation module includes the following sub-modules: A collection sub-module is used to collect the cable for 120 seconds using the optimized beam, record the signal at a sampling rate of 10 kHz to obtain the time domain signal of the cable vibration, and perform frequency spectrum analysis on the preprocessed time domain signal. The conversion submodule is used for pre-processing the collected time domain signal through digital signal processing, including filtering and detrending, and then performing spectrum analysis, converting the time domain signal into a frequency domain signal by using FFT fast Fourier transform, and obtaining the cable force.

10. The hybrid beam control scanning radar based bridge cable force monitoring system of claim 8, wherein, The cable force signal processing module includes the following submodules: The processing submodule is used for capturing the long-term dependence of the cable force change through an LSTM unit, setting three early warning levels according to the change trend of the cable force, including slight abnormality, moderate abnormality and serious abnormality; The early warning submodule is used for automatically generating early warning information when detecting abnormal change, including the abnormal cable number, the abnormal degree, the change trend graph and the recommended maintenance time.

Images