A Health Monitoring Method for Transmission Tower Structures Based on BeiDou Fusion Positioning and Edge Intelligence

Abstract

This invention relates to the field of structural health monitoring technology and discloses a method for monitoring the structural health of transmission towers based on BeiDou fusion positioning and edge intelligence. The method includes acquiring position data output by a BeiDou positioning module and inertial measurement data output by a multimodal inertial measurement unit; performing multi-sensor data fusion processing on the position data and inertial measurement data at the edge to generate a state vector containing the position, velocity, and attitude of the transmission tower monitoring point; performing intelligent structural state analysis based on the state vector to calculate the tilt angle and displacement, and using an artificial intelligence model to evaluate and obtain a structural health index; and executing event-driven data reporting when any parameter exceeds a preset alarm threshold. This invention achieves high-precision dynamic perception of the transmission tower structural state by completing data analysis and decision-making at the monitoring front end, while significantly reducing system power consumption and improving the real-time performance of safety warnings.

Description

Technical Field

[0001] This invention relates to the field of structural health monitoring technology, specifically a method for monitoring the structural health of power transmission towers based on BeiDou fusion positioning and edge intelligence. Background Technology

[0002] As a critical load-bearing structure in power transmission networks, transmission towers are typically erected in the field and subjected to complex environmental factors such as wind loads, icing, uneven settlement, and geological activity over long periods. This can lead to excessive tilting, displacement, or abnormal vibration in their structures, which in severe cases can threaten the safe and stable operation of the power transmission system. Therefore, long-term and effective online monitoring of the structural health status of transmission towers is of great significance.

[0003] Existing technologies for monitoring power transmission tower structures have certain limitations. Some solutions rely solely on global navigation satellite systems (such as BeiDou) for displacement monitoring, but their low sampling frequency makes it difficult to capture the high-frequency dynamic characteristics of the structure, and their continuous operation consumes a lot of power. Other solutions use inertial measurement units (IMUs) to sense the vibration and tilt of the structure. Although they have fast dynamic response and low power consumption, due to the integral drift problem inherent in the sensors, they cannot accurately measure the long-term, slowly varying displacement of the structure. Without an accurate external reference, the calculation results will diverge over time, resulting in insufficient reliability.

[0004] To integrate the advantages of different sensors, some monitoring systems transmit all raw data collected from multiple sources to a remote monitoring center via wireless network for fusion analysis without processing. While this approach leverages the powerful computing resources of the cloud, it places high demands on wireless communication bandwidth and stability. Furthermore, continuous data transmission generates significant power consumption, making it difficult for battery-powered field monitoring nodes to operate autonomously for extended periods. In addition, the remote transmission and centralized processing of data introduce unavoidable delays, reducing the real-time nature of safety warnings and potentially missing the optimal window for emergency response. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a method for monitoring the structural health of transmission towers based on BeiDou fusion positioning and edge intelligence. This method solves the problems of insufficient monitoring accuracy, high system power consumption making long-term autonomous operation difficult, and the inability to perform real-time intelligent analysis and early warning at the monitoring front end in existing transmission tower structural monitoring methods.

[0006] The technical solution provided by this invention includes:

[0007] The system acquires position data output from the BeiDou positioning module and inertial measurement data output from the multimodal inertial measurement unit. To ensure precise alignment of the data in the time dimension, a unified timestamp is generated for the position data and the inertial measurement data using the second pulse signal output from the BeiDou positioning module, based on the same time reference.

[0008] Multi-sensor data fusion processing is performed on time-synchronized position data and inertial measurement data to generate a state vector containing the position, velocity, and attitude of the transmission tower monitoring point. The state vector... It can be represented as:

[0009] ;

[0010] in, It is a three-dimensional vector representing the position of the monitoring point in the preset global coordinate system; It is a three-dimensional vector representing the velocity of the monitoring point in the preset global coordinate system; Let be a quaternion representing the rotational attitude of the monitoring point from the carrier coordinate system to the preset global coordinate system. This fusion process specifically executes a fusion algorithm based on the extended Kalman filter, performing an optimal estimate of the state vector through a cyclic iteration of a prediction step and an update step. The prediction step utilizes the inertial measurement data to perform high-frequency state estimation using kinematic equations; the update step is triggered upon receiving the position data, using the position data as an observation to correct the predicted state. The state transition equation and observation equation for this process can be expressed as follows:

[0011] ;

[0012] ;

[0013] in, Based on Time-state pair Predicted value of the state at any given time; yes The optimal estimate of the state vector at time t; and These are the corrected angular velocity and specific force. It is a nonlinear state transition function; yes The observation vector at any given time, i.e., the three-dimensional position coordinates output by the BeiDou positioning module; It is an observation function used to transform the state vector Mapping to the observation space, in this case, means extracting the positional component from the state vector. ; It is the observation noise vector, and its covariance matrix is This indicates the accuracy of BeiDou positioning data.

[0014] Based on the state vector generated by the fusion processing, intelligent structural state analysis is performed. This analysis first calculates the tilt angle and displacement, characterizing the static deformation of the structure. Specifically, attitude quaternions and position vectors are extracted from the state vector. The pitch and roll angles of the monitoring points are calculated using the attitude quaternions, and the current position vector is subtracted from the pre-stored initial reference position to obtain a three-dimensional displacement vector. Next, structural vibration characteristic analysis is performed. A dynamic acceleration time series with the gravity component removed is obtained from the inertial measurement data, and a fast Fourier transform algorithm is applied to the dynamic acceleration time series to obtain the vibration spectrum. Then, one or more dominant frequencies and dominant amplitudes are identified from the vibration spectrum. Finally, a comprehensive health status assessment is performed. The calculated tilt angle, displacement, and identified dominant frequencies and amplitudes form a feature vector, which is input into a pre-trained artificial intelligence model. The model outputs a quantified structural health index.

[0015] The system performs event-driven data reporting and power consumption management. The real-time calculated tilt angle, displacement, and assessed structural health index are compared with preset alarm thresholds. When any parameter exceeds its corresponding preset alarm threshold, an abnormal event is identified, and an alarm data packet containing the alarm type, key data, and a timestamp is immediately reported. In the absence of abnormal events, a routine status report is triggered at a preset, relatively long time interval, reporting a heartbeat data packet containing the device's operating status. Simultaneously, fine-grained power management is implemented, supplying power to the BeiDou positioning module and low-power wireless communication module only when positioning tasks or data reporting are required; during non-operating periods, these two modules are placed in power-off or sleep mode.

[0016] In a preferred embodiment, the method further includes an initialization step at the beginning. While the system remains stationary for a preset time period, the output of the multimodal inertial measurement unit is continuously acquired, and the average value of the measurements is calculated as the zero-bias error vector of the gyroscope and the accelerometer, used for real-time compensation of subsequently acquired dynamic data.

[0017] This invention provides a method for monitoring the structural health of power transmission towers based on BeiDou fusion positioning and edge intelligence. It has the following beneficial effects:

[0018] 1. This invention performs multi-sensor data fusion processing based on the extended Kalman filter algorithm, deeply coupling the discrete, absolutely accurate position data provided by the BeiDou positioning module with the high-frequency, continuous inertial measurement data provided by the multimodal inertial measurement unit. This suppresses the time-varying error of inertial navigation and compensates for the low update frequency of BeiDou positioning, thereby outputting a continuous, smooth, and centimeter-level accurate dynamic estimation result of structural pose, realizing the accurate perception of minute deformations and vibrations of transmission towers.

[0019] 2. This invention completes the core data processing and intelligent structural state analysis tasks on the edge computing processing unit, reporting data only when necessary, thus avoiding continuous wireless transmission of large amounts of raw data. Combined with an event-driven data reporting mechanism and a power management strategy that provides on-demand, time-sharing power to high-power components such as the BeiDou positioning module and low-power wireless communication module, the system operates at extremely low power consumption most of the time, extending battery life and meeting the needs of long-term unattended monitoring in outdoor environments.

[0020] 3. This invention directly performs structural status assessment and abnormal event determination in the edge computing processing unit at the monitoring front end. When the calculated tilt angle, displacement, or assessed structural health index exceeds the preset alarm threshold, the system can immediately generate an alarm data packet and trigger the reporting process. Compared with the traditional mode of transmitting all data to a remote center for analysis, it eliminates the delay in data transmission and central server processing, shortens the time from the occurrence of an event to the issuance of an alarm, and improves the timeliness of safety warnings. Attached Figure Description

[0021] Figure 1 This is a schematic diagram of a transmission tower structure health monitoring system according to an embodiment of the present invention;

[0022] Figure 2 This is a flowchart of a method for monitoring the structural health of a power transmission tower according to an embodiment of the present invention.

[0023] Among them, 100 is the Beidou positioning module; 200 is the multimodal inertial measurement unit; 300 is the edge computing processing unit; 400 is the low-power wireless communication module; and 500 is the power management module. Detailed Implementation

[0024] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] See attached document Figure 1 This invention provides a power transmission tower structural health monitoring system based on BeiDou fusion positioning and edge intelligence. The system is deployed on the power transmission tower to acquire and process structural status data. The system includes at least one intelligent sensing node and a remote monitoring center.

[0026] The intelligent sensing node may specifically include: a Beidou positioning module 100, a multimodal inertial measurement unit 200, an edge computing processing unit 300, a low-power wireless communication module 400, and a power management module 500.

[0027] The BeiDou positioning module 100 receives signals from the BeiDou satellite navigation system and calculates the three-dimensional absolute coordinates of its location using real-time dynamic technology or precise point positioning technology. The BeiDou positioning module 100 then outputs the calculated location data to the edge computing processing unit 300.

[0028] The multimodal inertial measurement unit 200 integrates a three-axis accelerometer, a three-axis gyroscope, and a three-axis magnetometer. The three-axis accelerometer measures the three-axis acceleration in the carrier coordinate system; the three-axis gyroscope measures the three-axis angular velocity in the carrier coordinate system; and the three-axis magnetometer measures the geomagnetic field intensity vector. The multimodal inertial measurement unit 200 outputs its measured acceleration, angular velocity, and magnetic field data to the edge computing processing unit 300.

[0029] The edge computing processing unit 300 has its input terminals connected to the output terminals of the BeiDou positioning module 100 and the multimodal inertial measurement unit 200, respectively. The edge computing processing unit 300 receives and processes position data, acceleration data, and angular velocity data. The edge computing processing unit 300 executes preset algorithms, including: performing time synchronization and preprocessing on the received data; running a data fusion algorithm to generate fused structural pose data using position data, acceleration data, and angular velocity data as input; and running a structural state analysis algorithm to extract structural state features from the fused data and perform a health status assessment. The edge computing processing unit 300 outputs the analyzed key information (such as alarm information and feature data) to the low-power wireless communication module 400 and sends control commands to the power management module 500 according to a preset power management strategy.

[0030] A low-power wireless communication module 400 has its input connected to the output of an edge computing processing unit 300. This module is used to transmit received critical information to a remote monitoring center via low-power wide-area network technology.

[0031] The power management module 500 consists of a battery and a power management chip. The output of the power management module 500 is connected to the power inputs of the BeiDou positioning module 100, the multimodal inertial measurement unit 200, the edge computing processing unit 300, and the low-power wireless communication module 400, providing power to these modules. Simultaneously, the power management module 500 receives control commands from the edge computing processing unit 300 to control the power supply status (e.g., active or sleep) of each module.

[0032] In the system's workflow, the edge computing processing unit 300 performs data fusion operations, processing discrete BeiDou position measurements and continuous IMU measurements to estimate the state vector of the transmission tower monitoring point. The state vector It can be expressed by the following formula:

[0033] ;

[0034] in: It is a three-dimensional vector representing the position of the monitoring point in the preset global coordinate system; It is a three-dimensional vector representing the velocity of the monitoring point in the preset global coordinate system; It is a quaternion representing the rotational attitude of the monitoring point from the carrier coordinate system to the preset global coordinate system.

[0035] The edge computing processing unit 300 continuously updates the state vector. This enables dynamic tracking of the structural posture of transmission towers.

[0036] See attached document Figure 1 The BeiDou positioning module 100 is configured to acquire the global absolute position information of the transmission tower monitoring points. The output of the BeiDou positioning module 100 is connected to the input of the edge computing processing unit 300, providing it with high-precision three-dimensional coordinate data.

[0037] The BeiDou positioning module 100 specifically includes a multi-frequency, multi-mode satellite navigation receiver and an external high-gain antenna. The satellite navigation receiver is configured to receive and process signals from multiple frequency points of the BeiDou satellite navigation system, including B1I, B2I, B1C, and B2a, while also being compatible with receiving signals from other global navigation satellite systems such as the Global Positioning System (GPS) and GLONASS. The external high-gain antenna, such as a choke coil antenna, is installed in an unobstructed, open location on a power transmission tower to enhance the reception quality of satellite signals and suppress multipath effects.

[0038] To achieve centimeter-level or higher positioning accuracy, the BeiDou positioning module 100 is configured to operate in either real-time dynamic technology mode or precise point positioning mode. In real-time dynamic technology mode, the BeiDou positioning module 100 acts as a mobile station, receiving differential correction data broadcast from a geographically fixed reference station via its built-in radio or network communication unit. This differential correction data includes correction information for common error terms such as satellite orbital errors, satellite clock errors, ionospheric delays, and tropospheric delays. The BeiDou positioning module 100 uses this data to correct its local observations, thereby calculating a high-precision relative position.

[0039] The basic principle of positioning calculation is based on the measurement of satellite signal propagation time, and its pseudorange observation equation can be expressed as:

[0040] ;

[0041] in: It is the pseudorange measurement value between the satellite and the receiver; It is the three-dimensional coordinates of a satellite in space provided by satellite ephemeris; These are the three-dimensional coordinates of the receiver to be determined in the same coordinate system; It is the speed of light in a vacuum; It is the deviation between the receiver's clock and the standard time of the Global Navigation Satellite System, i.e., the receiver clock error; It is the deviation between the satellite's clock and the standard time of the Global Navigation Satellite System, i.e., satellite clock bias; This is a signal propagation delay error caused by the ionosphere; This is a signal propagation delay error caused by the troposphere; This represents the sum of other unmodeled errors, such as multipath effects and measurement noise.

[0042] By receiving and processing signals from at least four satellites, a system of equations can be established to solve for the coordinates to be determined. and receiver clock difference The BeiDou positioning module 100, through its multi-frequency receiving capability, can utilize the characteristic that different frequency signals have different propagation delays in the ionosphere to establish a linear combination to eliminate first-order ionospheric delay errors. This is one of the key steps in achieving high-precision positioning. In real-time dynamic technology mode, through differential calculations between the base station and the rover station, most common error terms (such as...) are eliminated. , , This can be eliminated or significantly weakened.

[0043] Finally, the Beidou positioning module 100 will periodically transmit the calculated data, which includes information such as three-dimensional coordinates, positioning quality factors, and timestamps, to the edge computing processing unit 300 through a serial communication interface in a preset data format.

[0044] ;

[0045] ;

[0046] in: It is the three-dimensional angular velocity measurement vector output by the gyroscope; It is the actual three-dimensional angular velocity vector of the carrier; It is the zero-bias error vector of the gyroscope; It is the measurement noise vector of the gyroscope; It is the three-dimensional specific force measurement vector output by the accelerometer; It is the actual three-dimensional force vector of the carrier; It is the zero bias error vector of the accelerometer; It is the measurement noise vector of the accelerometer.

[0047] These errors, especially the zero-bias error vector of the gyroscope... and the zero bias error vector of the accelerometer This is the main reason why the inertial navigation solution results diverge over time. Therefore, the edge computing processing unit 300 needs to perform correction and compensation algorithms on these raw data.

[0048] The multimodal inertial measurement unit 200 digitizes the measurement results of each internal sensor after passing through an internal analog-to-digital converter, and forms continuous data frames at a preset high sampling frequency (e.g., 100Hz or higher). The data frames are then transmitted in real time to the edge computing processing unit 300 through digital communication interfaces such as serial peripheral interfaces or inter-integrated circuit buses.

[0049] See attached document Figure 1 The edge computing processing unit 300 is the core processing component of the intelligent sensing node. Specifically, the edge computing processing unit 300 can be implemented by a low-power microcontroller or system-on-a-chip configured with memory and multiple communication interfaces. Its data input terminals are connected to the output terminals of the BeiDou positioning module 100 and the multimodal inertial measurement unit 200, respectively; its data output terminals are connected to the input terminals of the low-power wireless communication module 400; and its control output terminals are connected to the control input terminals of the power management module 500.

[0050] The edge computing processing unit 300 is configured to execute a series of instructions stored in internal non-volatile memory to perform data processing, analysis, and decision-making functions. These functions specifically include:

[0051] First, the edge computing processing unit 300 performs the reception and preprocessing of multi-source data. It receives position data from the BeiDou positioning module 100 and inertial measurement data from the multimodal inertial measurement unit 200 via a serial interface. It uses the second pulse signal from the BeiDou positioning module 100 as a high-precision time reference to assign a unified timestamp to all received data, ensuring precise alignment of the data in the time dimension. Simultaneously, it performs zero-bias correction on the received raw inertial measurement data; the correction process can be expressed by the following formula:

[0052] ;

[0053] ;

[0054] in: It is the three-dimensional angular velocity measurement vector output by the gyroscope; It is the zero-bias error vector of the gyroscope; It is the three-dimensional angular velocity vector after zero bias correction, used for subsequent calculations; It is the three-dimensional specific force measurement vector output by the accelerometer; It is the zero bias error vector of the accelerometer; It is a three-dimensional specific force vector after zero bias correction, used for subsequent calculations.

[0055] Secondly, the edge computing processing unit 300 performs multi-sensor data fusion based on the extended Kalman filter algorithm. This algorithm uses corrected inertial data (a three-dimensional angular velocity vector after zero-bias correction). and the three-dimensional specific force vector after zero bias correction As input, the state vector of the transmission tower monitoring point is obtained through kinematic equations. The algorithm performs high-frequency predictions. When valid location data is received from the BeiDou positioning module 100, it uses this location data as an observation to predict the state vector. The prediction and correction process is then used to make corrections. Through this iterative process, the edge computing processing unit 300 can continuously output a high-precision, high-update-rate, and time-continuous structural pose estimation result.

[0056] Next, the edge computing processing unit 300 performs intelligent structural state analysis based on the fused pose data. This analysis includes:

[0057] Small Deformation and Tilt Analysis: From the Fuded State Vector Extracting attitude quaternions and position vector Using attitude quaternions The pitch and roll angles of the transmission tower monitoring points were calculated; the position vectors were used to calculate the pitch and roll angles. The displacement in three-dimensional space is calculated by comparing it with the pre-stored initial reference position.

[0058] Vibration characteristic analysis: Acceleration data sequences over a period of time are extracted and transformed from the time domain to the frequency domain using the Fast Fourier Transform algorithm to obtain the vibration spectrum. One or more dominant frequencies and their corresponding amplitudes are identified from this spectrum; these parameters reflect the inherent vibration characteristics of the structure.

[0059] Then, the edge computing processing unit 300 performs a comprehensive assessment of the structural health status. It constructs a feature vector from multiple characteristic parameters obtained from the aforementioned analysis (e.g., maximum tilt angle, maximum displacement, dominant frequency, dominant amplitude, etc.) and inputs this vector into a pre-trained artificial intelligence model. The artificial intelligence model outputs a quantified structural health index.

[0060] Finally, the edge computing processing unit 300 performs event-driven decision-making and reporting control. It compares the real-time analyzed deformation, tilt angle, vibration characteristic parameters, and structural health index with multiple preset thresholds in memory. When any parameter exceeds its corresponding safety threshold, the edge computing processing unit 300 determines that an abnormal event has occurred and immediately constructs an alarm data packet containing the alarm type, key data, and timestamp, transmitting it to the low-power wireless communication module 400. When no abnormal event occurs, the edge computing processing unit 300 triggers routine data reporting only once at a relatively long preset time interval (e.g., one hour). Simultaneously, the edge computing processing unit 300 also sends control signals to the power management module 500 to put high-power components such as the Beidou positioning module 100 and the low-power wireless communication module 400 into sleep mode during non-working periods.

[0061] See attached document Figure 1 The low-power wireless communication module 400 is configured to wirelessly transmit information processed by the edge computing processing unit 300 to a geographically separated remote monitoring center. The data input terminal of the low-power wireless communication module 400 is connected to the data output terminal of the edge computing processing unit 300, and its operating state is controlled by the edge computing processing unit 300.

[0062] The low-power wireless communication module 400 can be implemented using a low-power wide-area network (LPWAN) technology. In one embodiment, the low-power wireless communication module 400 is a LoRa (long-range) transceiver that operates in the unlicensed ISM (Industrial, Scientific, and Medical) band, utilizing linear frequency modulation spread spectrum (LFM) to achieve low-power and long-distance data transmission. In another embodiment, the low-power wireless communication module 400 is a narrowband Internet of Things (IoT) or LTE-M modem that accesses the Internet through operator-licensed cellular network infrastructure, thereby communicating with a remote monitoring center.

[0063] The core operating feature of the low-power wireless communication module 400 lies in its event-driven communication mechanism, which is primarily controlled by the edge computing processing unit 300 to reduce the overall system power consumption. For the vast majority of the time, the low-power wireless communication module 400 is set to a deep sleep mode or a network power-saving mode by the edge computing processing unit 300. In this mode, the radio frequency circuitry and most of the internal logic circuitry of the low-power wireless communication module 400 are shut down, maintaining only minimal power consumption.

[0064] Only when the edge computing processing unit 300 determines that data needs to be reported, for example, when it detects that the structural state parameters of the transmission tower exceed a preset threshold and trigger an abnormal event, or when a preset periodic reporting time point is reached, will the edge computing processing unit 300 output a wake-up signal through a specific control signal line to activate the low-power wireless communication module 400 from the sleep mode to the working mode.

[0065] Upon being woken up, the low-power wireless communication module 400 receives the data packet to be sent from the edge computing processing unit 300 via a serial communication interface. Subsequently, the low-power wireless communication module 400 performs the necessary network attach or connection establishment process and transmits the data packet. After confirming that data transmission is complete or a transmission timeout has occurred, the low-power wireless communication module 400 immediately and automatically, or under the instruction of the edge computing processing unit 300, returns to its deep sleep or power-saving mode to await the next wake-up command.

[0066] The data packets generated by the edge computing processing unit 300 and transmitted to the low-power wireless communication module 400 have their content constructed according to the conditions that trigger the reporting. For abnormal event reporting, the data packet may include: a unique identifier of the sensor node, a code indicating the event type, a high-precision timestamp of the event occurrence, and key characteristic parameter values ​​related to the specific event (such as excessive displacement, tilt angle, or abnormal vibration frequency). For periodic reporting, the data packet may include: a unique identifier of the sensor node, a current timestamp, and one or more status codes indicating the device's own operating status (such as battery voltage).

[0067] See attached document Figure 1The power management module 500 is configured to provide stable and controllable power to the Beidou positioning module 100, multimodal inertial measurement unit 200, edge computing processing unit 300 and low-power wireless communication module 400 within the intelligent sensing node.

[0068] The power management module 500 specifically includes a power unit and a power management integrated circuit. In one embodiment, the power unit is one or more lithium thionyl chloride battery packs, which are characterized by low self-discharge rate and high energy density, making them suitable for long-term deployment. In another embodiment, the power unit may include a rechargeable battery (such as a lithium-ion battery) and an energy harvesting device (such as a solar panel) connected to the rechargeable battery.

[0069] A power management integrated circuit, such as a CN3722 chip or a similar circuit, is used to perform battery charge / discharge management, voltage conversion, and power path switching. This integrated circuit internally integrates multiple DC-DC converters or low-dropout linear regulators to convert the output voltage of the power supply unit into stable operating voltages at different levels required by each module. For example, it provides one 3.3V power supply to the edge computing processing unit 300 and the multimodal inertial measurement unit 200, and another independent power supply to the BeiDou positioning module 100 and the low-power wireless communication module 400.

[0070] The power management module 500, working in conjunction with the edge computing processing unit 300, achieves fine-grained control over system power consumption. Specifically, the power management module 500 includes multiple power switches, such as switching circuits composed of metal-oxide-semiconductor field-effect transistors. The control terminals (gates) of these power switches are connected to the general-purpose input / output pins of the edge computing processing unit 300. Components with high instantaneous power consumption, such as the BeiDou positioning module 100 and the low-power wireless communication module 400, have the aforementioned controlled power switches connected in series on their power input lines.

[0071] In the system's workflow, the edge computing processing unit 300 executes a time-sharing power supply strategy according to preset program logic. Only when a positioning task is required does the edge computing processing unit 300 output a control signal (e.g., a high-level signal) to the corresponding general-purpose input / output pin, turning on the power switch to power the BeiDou positioning module 100. After the positioning task is completed, it immediately outputs a reverse control signal (e.g., a low-level signal) to turn off the power switch, thereby cutting off the power to the BeiDou positioning module 100. Similarly, the power switch to power the low-power wireless communication module 400 is turned on only when data reporting is required. This on-demand, time-sharing power supply method reduces the static leakage current consumption of high-power modules to zero during non-working periods, thereby reducing the system's average power consumption. For the edge computing processing unit 300 and the multimodal inertial measurement unit 200, which require continuous operation, the power management module 500 provides uninterrupted power.

[0072] See attached document Figure 2 The monitoring method provided by the present invention includes system initialization and data acquisition as its initial steps.

[0073] When the intelligent sensing node is powered on, the edge computing processing unit 300 first executes an initialization program. This program includes a self-check of the functional status of each hardware module of the system to confirm that the BeiDou positioning module 100, the multimodal inertial measurement unit 200, and the low-power wireless communication module 400 are all in a normal and communicable state. A key step in the initialization program is to perform static zero-bias calibration on the multimodal inertial measurement unit 200. In this step, the system remains stationary for a preset time period, during which the edge computing processing unit 300 continuously collects data. The angular velocity and acceleration measurements are obtained from the multimodal inertial measurement unit 200. Subsequently, the edge computing processing unit 300 calculates the average of these measurements as the gyroscope's zero-bias error vector. and the zero bias error vector of the accelerometer The calculation process is as follows:

[0074] ;

[0075] ;

[0076] in: This is the total number of samples collected during static calibration; It is the first The subsampled three-dimensional angular velocity measurement vector; It is the first The three-dimensional force measurement vector of the subsampled sample.

[0077] The calculated zero bias error vector of the gyroscope and the zero bias error vector of the accelerometer The data is stored in the internal memory of the edge computing processing unit 300 for real-time compensation of subsequently acquired dynamic data. The initialization process also includes establishing a structural health baseline; that is, after the BeiDou positioning module 100 obtains a stable and high-precision positioning solution, the current position vector is recorded as the initial reference position. .

[0078] After initialization, the system enters a continuous data acquisition phase. The BeiDou positioning module 100 outputs data frames containing three-dimensional position coordinates and time information at a preset low frequency (e.g., 1Hz). Simultaneously, the multimodal inertial measurement unit 200 continuously outputs inertial measurement data containing three-dimensional angular velocity and three-dimensional acceleration at a preset high frequency (e.g., 100Hz).

[0079] To ensure the effectiveness of data from two different sources and at different frequencies in subsequent fusion processing, the edge computing processing unit 300 implements a precise time synchronization mechanism. This mechanism utilizes the second pulse signal output by the BeiDou positioning module 100. A dedicated input pin of the edge computing processing unit 300 is connected to the output of the second pulse signal. When this pin detects the rising edge of the second pulse signal, it indicates the precise start of a Coordinated Universal Time (UTC) second, triggering an interrupt. In the interrupt service routine, the edge computing processing unit 300 records the integer second timestamp of this moment and simultaneously reads or resets an internal high-resolution hardware counter. For each subsequent frame of inertial measurement data received, the edge computing processing unit 300 reads the current value of the hardware counter while receiving the data and combines it with the integer second timestamp recorded when the previous second pulse signal interrupted to generate a high-precision unified timestamp. Through this method, all acquired position data and inertial measurement data are assigned time information under the same time base, thereby achieving precise time alignment of heterogeneous data.

[0080] See attached document Figure 2 After data acquisition and time synchronization are completed, the method flow of this invention enters the edge-side multi-sensor data fusion processing stage. This stage is executed in the edge computing processing unit 300 and aims to fuse sensor data with different characteristics to generate a unified, continuous, and high-precision state estimate of the transmission tower monitoring point.

[0081] First, the data from the time-synchronized multimodal inertial measurement unit 200 is preprocessed. Specifically, the edge computing processing unit 300 reads the zero-bias error of the gyroscope determined during the initialization phase from its internal memory. and the zero bias error vector of the accelerometer and for the current moment The acquired raw inertial measurement data is compensated in real time to obtain the corrected angular velocity vector. and force vector .

[0082] Subsequently, a deep fusion algorithm based on extended Kalman filtering is executed. This algorithm iterates through a prediction step and an update step to process the state vector of the transmission tower monitoring points. Perform optimal estimation. State vector. The definition is consistent with the above, including position, velocity, and attitude.

[0083] The algorithm's prediction step utilizes frequently updated inertial measurement data for state deduction. This step is based on a kinematic model, whose state transition equation can be expressed as:

[0084] ;

[0085] in: yes The optimal estimate of the state vector at time t; Based on Time-state pair Predicted value of the state at any given time; It is a nonlinear state transition function, and its specific calculation process includes:

[0086] Posture update: ,in for The attitude quaternion estimate at time t. To represent quaternion multiplication, It is composed of angular velocity vector and time interval The calculated attitude change quaternion.

[0087] Speed ​​updates: ,in for The velocity vector estimate at time t. It is composed of attitude quaternions The calculated rotation matrix from the carrier coordinate system to the global coordinate system. It is the gravitational acceleration vector in the global coordinate system.

[0088] Location update: ,in for The estimated position vector at time t.

[0089] Simultaneously, in the prediction step, the error covariance matrix of the state estimation is also recursively derived:

[0090] ;

[0091] in: yes The posterior error covariance matrix at time t; yes The prior error covariance matrix at time t; It is a state transition function exist The Jacobian matrix of the state vector at that point; It is the covariance matrix of the process noise, which characterizes the accuracy of the motion model.

[0092] The algorithm's update step is triggered upon receiving valid location data from the BeiDou positioning module 100. This step uses this location data to correct the state obtained in the prediction step. This step is based on a measurement model, whose observation equation can be expressed as:

[0093] ;

[0094] in: yes The observation vector at any given time, i.e., the three-dimensional position coordinates output by the BeiDou positioning module 100; It is an observation function used to transform the state vector Mapping to the observation space, in this case, means extracting the positional component from the state vector. ; It is the observation noise vector, and its covariance matrix is This indicates the accuracy of BeiDou positioning data.

[0095] The specific calculation process for the update step is as follows:

[0096] Calculate Kalman gain : ,in It is the observation function exist The Jacobian matrix of the state vector.

[0097] Update state estimation : This is The posterior state estimate at time t is the optimal estimate at that time.

[0098] Update error covariance matrix : ,in It is an identity matrix.

[0099] Through continuous iteration of the above prediction and update steps, the edge computing processing unit 300 generates a temporally continuous, smooth, and accuracy-corrected sequence of state vectors for the transmission tower monitoring points. This sequence serves as the data foundation for subsequent intelligent analysis of structural states.

[0100] See attached document Figure 2 By using data fusion processing, continuous and high-precision state vector estimation can be obtained. Subsequently, the method flow of this invention enters the intelligent analysis and evaluation stage of the edge-side structural status. This stage is executed in the edge computing processing unit 300, which is used to extract key feature parameters from the fused data and to make a quantitative assessment of the structural health status of the transmission tower.

[0101] This stage first performs the calculation of small deformations and tilts. The edge computing processing unit 300 estimates the latest state vector. Extract attitude quaternions and position vector Using attitude quaternions To calculate the real-time tilt angle of the monitoring point, specifically including the pitch angle. and roll angle The calculation formula is as follows:

[0102] ;

[0103] ;

[0104] in: , , , These are attitude quaternions The four components; It is the pitch angle, representing the rotation angle about the Y-axis of the carrier coordinate system; It is the roll angle, which represents the rotation angle about the X-axis of the carrier coordinate system; It is a return point The bivariate arctangent function of polar coordinate angles.

[0105] At the same time, by using the current position vector Compared with the reference position vector stored during initialization By performing vector subtraction, the three-dimensional displacement vector of the monitoring point is obtained. .

[0106] This stage then proceeds to analyze the structural vibration characteristics. The edge computing processing unit 300 extracts a data segment from the preprocessed acceleration data stream that contains... A time window of consecutive sampling points. To obtain dynamic acceleration purely caused by external forces, the gravitational component needs to be removed from the acceleration in the global coordinate system. Dynamic acceleration vector The calculation formula is:

[0107] ;

[0108] in: It is derived from the current attitude quaternion The calculated rotation matrix from the carrier coordinate system to the global coordinate system; yes The specific force vector after time correction; This is the gravitational acceleration vector in the global coordinate system. Then, the Fast Fourier Transform (FFT) algorithm is applied to the dynamic acceleration time series along one of the axes (e.g., the vertical component) to transform it from the time domain to the frequency domain, obtaining the spectrum of the time series. The calculation of the Discrete Fourier Transform can be expressed as:

[0109] ;

[0110] in: It is the number of points of transformation, that is, the total number of samples within the time window; It is the first in the time series One dynamic acceleration sample value; It is an index in the frequency domain, ranging from arrive ; It corresponds to frequency The complex form of the spectral components, The sampling frequency; It is the imaginary unit. This is determined by the calculated spectral amplitude. By searching for peak values, the dominant frequencies of one or more structures and their corresponding amplitudes can be determined.

[0111] Finally, this stage performs a comprehensive structural health assessment based on an artificial intelligence model. The edge computing processing unit 300 combines the multiple feature parameters calculated in the preceding steps into a feature vector. This feature vector may include: maximum pitch angle, maximum roll angle, maximum displacement, dominant frequency and its offset, dominant amplitude, etc., within a specific evaluation period. The input feature vector is fed into an artificial intelligence model that has been pre-trained and deployed in an edge computing processing unit 300. This model processes the input feature vector and outputs a single scalar value. This value is defined as the structural health index. For example, this index... The value range is [0, 1], and a higher value indicates that the structural state is closer to the healthy baseline state. This quantified health index The calculated original characteristic parameters, such as tilt angle, displacement, and vibration frequency, will be compared with preset alarm thresholds to trigger subsequent decision-making and data reporting processes.

[0112] See attached document Figure 2 After completing the intelligent analysis and evaluation of the structural state, the method flow of this invention enters the event-driven data reporting and system management stage. This stage is executed in the edge computing processing unit 300, and its core is to make decisions based on the analysis results and control the data reporting behavior and the power consumption state of the system itself.

[0113] This stage begins with an abnormal event determination process. The edge computing processing unit 300 continuously compares the various structural state characteristic parameters calculated in real time in the previous stage with a set of preset alarm thresholds stored in its non-volatile memory. This determination logic can be expressed as follows: an abnormal event is determined to have occurred when any of the following conditions are met.

[0114] ;

[0115] in: It is the pitch angle calculated in real time. It is the roll angle calculated in real time; and It is the preset tilt angle alarm threshold; It is a three-dimensional displacement vector calculated in real time; It is the preset displacement modulus alarm threshold; It is a structural health index that is assessed in real time; It is the preset health index alarm threshold; Represents a logical OR operation.

[0116] Once an abnormal event is detected, the edge computing processing unit 300 immediately executes the alarm reporting process. It first constructs an alarm data packet containing: a unique device identifier for this smart sensor node, the exact timestamp of the event, an event code to distinguish the alarm type (e.g., to differentiate between tilt exceeding limits and low health index), and the current values ​​of one or more key characteristic parameters that triggered the alarm. Subsequently, the edge computing processing unit 300 sends a control signal to the power management module 500 via its input / output pins to turn on the power switch supplying power to the low-power wireless communication module 400. After confirming that the low-power wireless communication module 400 is powered on and ready, the edge computing processing unit 300 sends the alarm data packet to the low-power wireless communication module 400 via a serial interface, whereby the low-power wireless communication module 400 performs wireless transmission. Upon completion of the transmission task, the edge computing processing unit 300 immediately sends a control signal to cut off the power to the low-power wireless communication module 400, returning it to a zero-power state.

[0117] In the absence of any abnormal events, the edge computing processing unit 300 follows a periodic reporting and low-power management strategy. An internal low-power timer within the unit counts at preset, relatively long intervals (e.g., several hours or 24 hours). When the timer reaches the preset duration, a routine status reporting process is triggered. In this process, the edge computing processing unit 300 constructs a heartbeat data packet, which may contain: a device identifier, a current timestamp, the current battery voltage value measured by the power management module 500, and a system status code indicating that the system is operating normally. The subsequent steps of waking up the low-power wireless communication module 400, sending the data packet, and putting the low-power wireless communication module 400 into sleep mode are consistent with the alarm reporting process.

[0118] During the intervals between two reporting tasks, the edge computing processing unit 300 performs fine-grained power management. Except for a brief period of power supply to the BeiDou positioning module 100 when global position correction is needed, the BeiDou positioning module 100 and the low-power wireless communication module 400 are completely powered off for the majority of the time. The edge computing processing unit 300 itself and the continuously operating multimodal inertial measurement unit 200 operate in the lowest power consumption mode. This mechanism, which strictly binds data reporting behavior to structural state anomaly events and incorporates periodic heartbeats, achieves effective management of system power consumption.

Claims

1. A method for monitoring the structural health of power transmission towers based on BeiDou fusion positioning and edge intelligence, characterized in that, include: Acquire position data output by the BeiDou positioning module and inertial measurement data output by the multimodal inertial measurement unit; The location data and the inertial measurement data are subjected to multi-sensor data fusion processing. Specifically, the multi-sensor data fusion processing is performed by executing a fusion algorithm based on the extended Kalman filter algorithm. Through iterative prediction and update steps, the state vector is optimally estimated to generate a state vector containing the position, velocity and attitude of the transmission tower monitoring point. Based on the state vector, structural state intelligent analysis is performed, which includes: The tilt angle and displacement are calculated from the state vector. The dynamic acceleration time series with the gravity component removed is obtained from the inertial measurement data; The fast Fourier transform algorithm is applied to the dynamic acceleration time series to obtain the vibration spectrum, and one or more dominant frequencies are identified from the vibration spectrum. The calculated tilt angle, displacement, and the identified dominant frequency and dominant amplitude are combined to form a feature vector; The feature vector is input into a pre-trained artificial intelligence model, which then outputs a quantified structural health index. The system performs event-driven data reporting. When any parameter among the tilt angle, displacement, or structural health index exceeds a preset alarm threshold, an alarm data packet containing alarm type, key data, and timestamp is reported. It also includes a power management step: powering the BeiDou positioning module and the low-power wireless communication module only when positioning tasks or data reporting are required; During non-working periods, the BeiDou positioning module and the low-power wireless communication module are placed in power-off or sleep mode.

2. The method for monitoring the structural health of transmission towers based on BeiDou fusion positioning and edge intelligence according to claim 1, characterized in that, Before the multi-sensor data fusion processing, the method further includes: using the second pulse signal output by the Beidou positioning module to generate a unified timestamp for the location data and the inertial measurement data under the same time reference.

3. The method for monitoring the structural health of transmission towers based on BeiDou fusion positioning and edge intelligence according to claim 2, characterized in that, The prediction step uses the inertial measurement data to perform state deduction through kinematic equations; The update step is triggered when the location data is received, and the location data is used as an observation to correct the predicted state.

4. The method for monitoring the structural health of transmission towers based on BeiDou fusion positioning and edge intelligence according to claim 1, characterized in that, The calculation of the tilt angle and displacement specifically includes: Extract the attitude quaternion and position vector from the state vector; The pitch and roll angles of the monitoring points are calculated using the attitude quaternions. Subtract the current position vector from the pre-stored initial reference position to obtain the three-dimensional displacement vector.

5. The method for monitoring the structural health of transmission towers based on BeiDou fusion positioning and edge intelligence according to claim 1, characterized in that, The event-driven data reporting also includes: triggering a routine status report at a preset long time interval when no abnormal event occurs, and reporting a heartbeat data packet containing the device's working status.

6. The method for monitoring the structural health of transmission towers based on BeiDou fusion positioning and edge intelligence according to claim 1, characterized in that, The method also includes an initialization step at the beginning: while the system remains stationary for a preset time, the output of the multimodal inertial measurement unit is continuously acquired, and the average value of the measured values ​​is calculated as the zero bias error vector of the gyroscope and the zero bias error vector of the accelerometer, which is used to compensate for the dynamic data acquired in real time.

Images