Intelligent sensing array early warning system for full-life damage of mixed tower structure

Abstract

The invention discloses a mixed tower structure full-life damage intelligent sensing array early warning system, which relates to the field of mixed tower structure detection and comprises a multi-modal data collection module, an array topology optimization module, a self-adaptive signal processing module, a digital twin life prediction module, a grading early warning module and a visualization system. The multi-modal data collection module comprises a multi-modal sensor array, a self-powered module and a wireless transmission module. According to the invention, a full-scale sensing network is constructed, full-dimension damage perception from distributed monitoring to sudden damage capture and structural modal analysis is realized, wavelet transform and blind source separation are combined to eliminate environmental noise interference, a damage characteristic ultrasonic attenuation coefficient, an acoustic emission energy spectrum peak value, optical fiber strain gradient anomaly and vibration modal frequency deviation are extracted, and the detection accuracy is improved. And classification and positioning of damage types and intelligent diagnosis of severity levels are realized through a convolutional neural network and long and short memory neural network hybrid model, and a closed-loop processing flow from data acquisition to feature analysis is formed.

Description

Technical Field

[0001] The present invention relates to the field of detection of hybrid tower structures, and particularly to an intelligent perception array early warning system for the whole-life damage of hybrid tower structures. Background Art

[0002] The tower barrel of a wind turbine is a key structure for supporting the wind turbine unit. In recent years, concrete-steel hybrid tower barrels have gradually become popular due to their advantages such as low cost, easy availability of materials, and good anti-fatigue performance. When the wind turbine is operating, the hybrid tower structure continuously bears alternating stresses such as wind loads and mechanical vibrations, which are prone to cause fatigue damage. Microcracks inside the concrete may expand into macroscopic cracks under long-term loads, threatening the structural safety.

[0003] Disclosed patent: A health online monitoring system and method for the concrete tower barrel structure of a wind turbine (Publication No.: CN116517788A), including a data acquisition module, a data transmission module, a data processing module, and a user interface; the data acquisition module is used to acquire the monitoring data of each part of the concrete tower barrel, the data transmission module is used to transmit the monitoring data to the data processing module, the data processing module is used to process and analyze the monitoring data, so as to realize the evaluation of the local and overall health states of the concrete tower barrel structure of the wind turbine, and the user interface is used to visually display the evaluation results of the local and overall health states of the concrete tower barrel structure. While analyzing the local health states of each part of the concrete tower barrel structure through the monitoring results of multiple sensors, the present invention can also evaluate the overall health state of the concrete tower barrel.

[0004] The above patent has the following defects: The types of sensors are single and the coverage range is limited, unable to detect microcracks, delamination defects, and sudden damages, relying on traditional statistical methods, lacking deep learning algorithms, with low damage identification accuracy and insufficient real-time performance, and unable to achieve accurate evaluation of the remaining life. Summary of the Invention

[0005] The purpose of the present invention is to solve the defects existing in the prior art, and propose an intelligent perception array early warning system for the whole-life damage of hybrid tower structures.

[0006] To achieve the above purpose, the present invention adopts the following technical solution: An intelligent perception array early warning system for the whole-life damage of hybrid tower structures, including a multi-modal data collection module, an array topology optimization module, an adaptive signal processing module, a digital twin life prediction module, a hierarchical early warning module, and a visualization system. The multi-modal data collection module includes a multi-modal sensor array, a self-power supply module, and a wireless transmission module. The self-power supply module integrates a piezoelectric vibration power generation unit and a thermoelectric power generation module to provide long-term maintenance-free power for the sensors. The wireless transmission module uses a hybrid transmission of a wide area network and a wireless network. Low-frequency data is transmitted through the wide area network, and high-frequency data is transmitted back in real time through the wireless network.

[0007] As a further description of the above technical solution:

[0008] The multi-modal sensor array forms a monitoring network through a piezoelectric ceramic array, a distributed fiber optic sensor, an acoustic emission sensor, and a low-frequency vibration sensor. The piezoelectric ceramic array is embedded in the concrete section of the hybrid tower, emitting and receiving ultrasonic pulses, and its operating frequency can be adaptively switched to detect defects of different scales. The distributed fiber optic sensor is arranged along the axial and circumferential directions of the tower barrel to form a monitoring network, monitoring the strain distribution and temperature field changes. The acoustic emission sensor is arranged in a circular array at the flange connection to capture the transient elastic wave signals generated by material fracture. The low-frequency vibration sensor is arranged at the top of the tower and the foundation section to collect low-frequency vibration signals and identify the deviation of the structural modal frequency. An anemometer is installed at the top of the tower to collect wind speed.

[0009] As a further description of the above technical solution:

[0010] The array topology optimization module is based on the finite element model of the hybrid tower, dynamically densifies the sensor layout in the concrete-steel transition section and flange connection in the damage-sensitive area to form a dense grid, and adjusts the sensor weights in real time according to the signal-to-noise ratio and data redundancy to optimize the monitoring network coverage. The signal-to-noise ratio and data redundancy are core parameters set based on communication standards, engineering practices, and numerical simulations.

[0011] As a further description of the above technical solution:

[0012] The adaptive signal processing module eliminates environmental noise through wavelet transform and blind source separation, extracts the ultrasonic attenuation coefficient, the peak value of the acoustic emission energy spectrum, the anomaly of the fiber optic strain gradient, and the deviation of the vibration modal frequency, constructs a hybrid model containing a convolutional neural network and a long short-term memory neural network, extracts the spatial features of damage, and captures the temporal evolution law of damage.

[0013] As a further description of the above technical solution:

[0014] The digital twin life prediction module integrates a concrete damage model and a steel fatigue model, combines real-time data, periodically updates material parameters, simulates crack propagation using the finite element method, and evaluates the remaining life by combining probability calculation methods.

[0015] As a further description of the above technical solution:

[0016] The hierarchical early warning module is provided with a three-level early warning mechanism, and the early warning threshold is dynamically adjusted according to historical data and environmental parameters.

[0017] As a further description of the above technical solution:

[0018] The visualization system constructs a three-dimensional visualization interface based on GIS and BIM technology to display the damage location, expansion path and predicted life. The three-dimensional geometric model of the mixed tower is constructed through BIM technology. Through AR glasses, the damage location and maintenance instructions are superimposed, and voice interaction is supported to retrieve historical data.

[0019] As a further description of the above technical solution:

[0020] The whole process of the system is as follows:

[0021] S1. Dynamic topology optimization: Based on the mixed tower finite element model, the sensor density is dynamically increased in damage-sensitive areas, and the sensor weight is adjusted in real time according to the signal-to-noise ratio and data redundancy to optimize the monitoring network coverage.

[0022] S2. Data Collection

[0023] A. The piezoelectric ceramic array emits ultrasonic waves to detect concrete microcracks and delamination defects. The acoustic emission sensor captures transient elastic waves and locates sudden damage. The distributed optical fiber sensor monitors strain distribution and temperature field. The low-frequency vibration sensor collects vibration signals and identifies structural modal anomalies. The anemometer collects wind speed.

[0024] B. Data transmission: Low-frequency data is transmitted through a low-power wide area network, and high-frequency data is transmitted back in real time through a high-speed wireless network.

[0025] S3. Data preprocessing: Wavelet transform is used to eliminate wind load and temperature noise, and blind source separation is used to extract damage feature vectors. The feature vectors include ultrasonic attenuation coefficient, acoustic emission energy spectrum peak, optical fiber strain gradient anomaly and vibration mode frequency offset. A lightweight convolutional neural model is preliminarily run to screen potential damage signals.

[0026] S4, Cloud-based Intelligent Analysis

[0027] A. Input feature vector, convolutional neural network extracts crack morphology, spatial characteristics of delamination defects, long short memory neural network captures the time sequence of crack propagation, and outputs damage type, location and severity level.

[0028] B. The concrete damage model and steel fatigue model receive real-time damage data, build a high-fidelity structural twin, simulate the crack propagation path of the mixed tower structure, predict the remaining life, and provide a scientific basis for graded early warning.

[0029] S5. Gradual warning: According to the size of the crack, the three-level warning mechanism is dynamically triggered. The first-level warning is a local sound and light alarm, and a text message / email is pushed to the operation and maintenance platform. The second-level warning automatically generates a maintenance plan and pushes it to the mobile terminal. The third-level warning links the fan control system to reduce load or shut down urgently. Based on historical data and environmental parameters, the alarm logic is dynamically optimized.

[0030] S6, Visual feedback

[0031] A. Real - time display on a three - dimensional visualization interface: The damage location is marked with three - dimensional coordinates, the extended path is displayed with a red - highlighted trajectory, and the remaining life is displayed with a countdown;

[0032] B. AR - assisted maintenance: Superimpose damage location guidance, retrieve historical data through voice interaction, and generate standardized maintenance processes.

[0033] S7, Operation and maintenance response: Perform hierarchical operations according to the warning level. For a first - level response, manual inspection is carried out to confirm the damage; for a second - level response, bolt re - tightening or local reinforcement is performed; for a third - level response, a full - scale overhaul or replacement of damaged components is carried out after shutdown.

[0034] The present invention has the following beneficial effects:

[0035] 1. In the present invention, by integrating piezoelectric ceramic arrays, distributed fiber optic sensors, acoustic emission sensors, and low - frequency vibration sensors, a full - scale sensing network is constructed to achieve all - dimensional damage perception from micro - crack detection, strain / temperature distribution monitoring to sudden damage capture and structural modal analysis. Combining wavelet transform and blind source separation to eliminate environmental noise interference, extracting damage features such as ultrasonic attenuation coefficient, peak value of acoustic emission energy spectrum, abnormal fiber optic strain gradient, and vibration modal frequency shift, and realizing intelligent diagnosis of damage type classification, location, and severity level through a hybrid model of convolutional neural network and long short - term memory neural network, forming a closed - loop processing flow from data acquisition to feature analysis.

[0036] 2. In the present invention, integrating the concrete damage plasticity model and the steel fatigue model, periodically updating material parameters, simulating the crack propagation path by combining the extended finite element method, and evaluating the confidence interval of the remaining life through Monte Carlo probability simulation, quantifying the structural reliability under different environmental loads, triggering a third - level response based on the crack length, dynamically optimizing the warning threshold according to historical data and environmental parameters, and linking with the fan control system to perform load reduction or shutdown, realizing a closed - loop control from damage perception to operation and maintenance response.

[0037] 3. In the present invention, by integrating a vibration / thermal - difference power generation module to provide long - term maintenance - free power for sensors, combining a hybrid transmission scheme of wide - area network and high - speed wireless network, realizing long - distance transmission of low - frequency data and real - time transmission of high - frequency data, constructing a three - dimensional visualization model of the hybrid tower structure based on GIS and BIM technologies, dynamically displaying the damage location, crack propagation path, and remaining life prediction results, and superimposing high - precision maintenance guidance through AR glasses, supporting voice interaction to retrieve historical data and operation instructions, realizing a full - link closed - loop management from data acquisition, analysis to operation and maintenance response. Brief description of the drawings

[0038] Figure 1 It is the system architecture diagram of the present invention;

[0039] Figure 2 This is the system flow chart of the present invention. Specific implementation manners

[0040] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0041] Referring to Figure 1-2 , an embodiment provided by the present invention: an intelligent perception array early warning system for the full-life damage of a hybrid tower structure, including a multi-modal data collection module, an array topology optimization module, an adaptive signal processing module, a digital twin life prediction module, a hierarchical early warning module, and a visualization system. The multi-modal data collection module includes a multi-modal sensor array, a self-powered module, and a wireless transmission module. The self-powered module integrates a piezoelectric vibration power generation unit and a thermoelectric power generation module to provide long-term maintenance-free power for the sensors. The wireless transmission module uses a hybrid transmission of a wide area network and a wireless network. Low-frequency data is transmitted through the wide area network, and high-frequency data is real-time transmitted back through the wireless network.

[0042] The multi-modal sensor array forms a monitoring network through a piezoelectric ceramic array, a distributed optical fiber sensor, an acoustic emission sensor, and a low-frequency vibration sensor. The piezoelectric ceramic array is embedded in the concrete section of the hybrid tower to emit and receive ultrasonic pulses, and its working frequency can be adaptively switched to detect defects of different scales. The distributed optical fiber sensor is arranged along the axial and circumferential directions of the tower barrel to form a monitoring network to monitor the strain distribution and temperature field change. The acoustic emission sensor is arranged in a ring array at the flange connection to capture the transient elastic wave signal generated by material fracture. The low-frequency vibration sensor is arranged at the top of the tower and the foundation section to collect low-frequency vibration signals and identify the deviation of the structural modal frequency. A wind speed meter is installed at the top of the tower to collect wind speed. The array topology optimization module dynamically densifies the sensor arrangement in the concrete-steel transition section and the flange connection in the damage-sensitive area based on the finite element model of the hybrid tower to form a dense grid, and adjusts the sensor weights in real time according to the signal-to-noise ratio and data redundancy to optimize the monitoring network coverage. The signal-to-noise ratio and data redundancy are core parameters set based on communication standards, engineering practices, and numerical simulations.

[0043] The adaptive signal processing module eliminates environmental noise through wavelet transform and blind source separation, extracts ultrasonic attenuation coefficient, peak value of acoustic emission energy spectrum, abnormal fiber strain gradient, and vibration mode frequency shift, constructs a hybrid model containing convolutional neural network and long short-term memory neural network, extracts spatial features of damage, captures the time-series evolution law of damage. The core function of the adaptive signal processing module is to separate noise and extract damage features from the original signals collected by the multi-modal sensor array, and achieve accurate classification, location, and severity assessment of damage types through a deep learning model, providing reliable input for subsequent life prediction and early warning. This module realizes the full-chain intelligent processing from the original signal to damage decision-making, providing core technical support for the full-life health monitoring of the hybrid tower structure. The adaptive signal processing module receives the original data and environmental parameters collected by the multi-modal sensor array. The environmental parameters include wind speed and temperature. The wavelet transform for noise reduction uses the Daubechies5 wavelet basis to perform multi-scale decomposition on the original signal, and performs threshold processing through the hard threshold method to filter out low-frequency environmental noises such as wind load and temperature fluctuations, retaining high-frequency damage features and improving the signal-to-noise ratio. The blind source separation applies the FastICA algorithm to separate independent damage source signals from the mixed signals. Assuming that the damage signal and noise are statistically independent, the damage features are extracted by maximizing non-Gaussianity. The following key features are extracted from the noise-reduced signal. The ultrasonic attenuation coefficient reflects the energy loss caused by scattering or absorption when ultrasonic waves propagate in materials. The larger the coefficient, the denser the cracks or the more severe the delamination. The peak value of the acoustic emission energy spectrum characterizes the peak value of the elastic wave energy released by material fracture. The higher the peak value, the more intense the damage event. The abnormal fiber strain gradient indicates the strain change per unit length. The abnormal gradient marks local stress mutation. The vibration mode frequency shift reflects the change in the natural frequency of the structure. The larger the shift, the more severe the stiffness degradation. The lightweight convolutional neural network model is run to preliminarily screen potential damage signals, which can reduce the cloud computing load. The data is deeply analyzed in the cloud, and the feature vector is input into the hybrid model of convolutional neural network and long short-term memory neural network. The convolutional neural network extracts the spatial features of cracks, delamination, and loosening, and the long short-term memory neural network captures the time-series evolution law of crack propagation. The time-series evolution law refers to the dynamic trends and patterns of damage features changing over time, including damage propagation trends, material property degradation, the combined effects of environment and load, and structural dynamic response characteristics. The damage type, location, and severity level are output, and a structured damage report is output, including damage type, location, size, and propagation trend. The results are transmitted to the digital twin module and the hierarchical early warning module, and a three-level early warning mechanism is triggered according to the severity level. The severity level is divided based on the feature amplitude. The feature amplitude is a key parameter extracted from the multi-modal sensor data, used to quantify the severity of damage, including crack length, delamination area, strain gradient, and loosening degree. The crack length is jointly inverted from the ultrasonic signal attenuation coefficient and the peak value of the acoustic emission energy spectrum. Level 1 (mild): crack length ≤ 2mm,Level 2 (moderate): 2 mm < crack length < 5 mm, Level 3 (critical): crack length ≥ 5 mm, the delamination area and strain gradient are jointly inversed by the ultrasonic attenuation coefficient and the abnormal fiber optic strain gradient, Level 1 (mild): delamination area ≤ 0.1 m, 2 , strain gradient < 2 με / m, Level 2 (moderate): 0.1 m 2 < delamination area < 0.5 m 2 , 2 με / m ≤ strain gradient < 5 με / m, Level 3 (critical): delamination area ≥ 0.5 m 2 , strain gradient ≥ 5 με / m, the degree of looseness is jointly inversed by the vibration mode frequency shift and the peak value of the acoustic emission energy spectrum, Level 1 (mild): frequency shift ≤ 0.2 Hz, acoustic emission events ≤ 5 times / minute, Level 2 (moderate): 0.2 Hz < frequency shift < 0.5 Hz, 5 times / minute < acoustic emission events ≤ 10 times / minute, Level 3 (critical): frequency shift ≥ 0.5 Hz, acoustic emission events > 10 times / minute. According to historical data and environmental parameters, the characteristic threshold is dynamically updated to reduce the false alarm rate.

[0044] The digital twin life prediction module integrates a concrete damage model and a steel fatigue model, combines real-time data, periodically updates material parameters, uses the finite element method to simulate crack propagation, and combines probability calculation methods to evaluate the remaining life. The core function of the digital twin life prediction module is to construct a digital twin model of the hybrid tower. Through dynamic simulation and probability calculation, it simulates the crack propagation path of the hybrid tower structure, predicts the remaining life, and provides a scientific basis for hierarchical early warning. It integrates a concrete damage model, a steel fatigue model, and real-time monitoring data to construct a high-fidelity structural twin. The concrete damage model is used to describe the stiffness degradation and crack propagation behavior of concrete under alternating loads, and the steel fatigue model is used to quantify the fatigue damage accumulation of steel in long-term vibration. The damage parameters and vibration mode frequency offsets output by the adaptive signal processing module are input, and the parameters of the concrete damage model and the steel fatigue model are periodically updated. The parameters of the concrete damage model include the damage factor, equivalent plastic strain, elastic modulus, and temperature correction coefficient. The damage factor is calculated from the strain gradient anomaly monitored by the distributed optical fiber sensor and characterizes the degree of concrete stiffness degradation. The distributed optical fiber sensor monitoring data is received every 10 minutes for calculation to obtain a new damage factor. The equivalent plastic strain is obtained through vibration mode frequency offset and finite element inversion. The initial value of the elastic modulus is based on laboratory material test data and is dynamically adjusted according to the temperature field change. The temperature correction coefficient comes from the temperature monitoring data of the optical fiber sensor. When the detected temperature difference is 5 degrees Celsius or more,Update the temperature correction coefficient. The parameters of the steel fatigue model include the back stress tensor, isotropic hardening parameter, cyclic hardening coefficient, and fatigue damage threshold. The back stress tensor characterizes the kinematic hardening effect under cyclic loading and is calibrated by the structural modal frequency shift collected by the low-frequency vibration sensor. The isotropic hardening parameter is dynamically calibrated based on the crack propagation events captured by the acoustic emission sensor and the historical load spectrum, characterizing the overall strength degradation of the material. The cyclic hardening coefficient is initially fitted with the laboratory steel fatigue test data and subsequently dynamically corrected by the changes in the vibration modal parameters and abnormal fiber strain gradient monitored in real time. The fatigue damage threshold is jointly calibrated by the material property degradation database and the real-time monitoring data to ensure the synchronization of the twin body and the physical structure state. Receive real-time monitoring data every 10 minutes to trigger parameter optimization. When the crack propagation speed or the structural modal shift exceeds the preset threshold, update the parameters immediately. Integrate the vibration modal parameters, the peak value of the acoustic emission energy spectrum, and the fiber strain data, and adjust the parameters of the steel fatigue model through the backpropagation algorithm. Optimize through Monte Carlo simulation, generate 1000 groups of random load samples, combine the historical data and environmental parameters, and calculate the optimal parameter combination. The acoustic emission sensor captures the crack propagation events and dynamically calibrates the historical load spectrum. Simulate the crack propagation by the extended finite element method, divide the hybrid tower model, construct the finite element mesh, and mark the position, size, and direction of the crack with special functions. Initialize the crack position and size according to the real-time crack data, set the crack propagation direction and step size, calculate the stress intensity factor, and simulate the stress field at the crack tip by the extended finite element method. The crack tip is the starting point of the propagation. Input the load spectrum of the real-time monitored crack length, wind speed, and vibration amplitude to obtain the stress intensity factor. According to the stress intensity factor and the fracture toughness of the material, judge whether the crack reaches the propagation condition. If the stress intensity factor exceeds the threshold, the crack will extend along the direction of the maximum principal stress. Every time the extended finite element method simulates a step, the position of the crack tip is updated once, and the new crack path is re-marked on the finite element mesh. Monitor the crack length and load in real time through the sensor, dynamically adjust the stress intensity factor and the propagation direction, make the simulation closer to the real situation, do not need to re-divide the mesh, and save a large amount of calculation time. Through repeated loading of steel and concrete in the laboratory, measure the law of crack growth and obtain two key parameters C and m. Adjust the parameters every 10 minutes with the latest monitoring data to ensure that the model conforms to the actual situation. Crack propagation speed = C×(stress intensity factor at the crack tip)^ m , input the stress intensity factor and material parameters, directly calculate the crack propagation speed. Monte Carlo simulation simulates 1000 possible wind speed and temperature changes, calculates the crack propagation speed in each case, and finally gives the remaining life range, outputs the three-dimensional visual crack propagation path and the remaining life countdown display.

[0045] The hierarchical warning module is set with a three - level warning mechanism. The warning thresholds are dynamically adjusted according to historical data and environmental parameters. The core functions of the hierarchical warning module are to set a three - level response mechanism and dynamic threshold adjustment. According to the damage data and environmental parameters monitored in real time, it dynamically triggers warning responses at different levels, realizing the full - process safety control from mild damage reminder to critical failure emergency disposal. Based on the crack length, damage type and expansion trend, it divides the risk levels, combines historical data with environmental interference, and optimizes the warning trigger conditions. It triggers audible and visual alarms, pushes maintenance plans or controls the fan to reduce load / stop. Set a three - level warning mechanism. The first - level warning is mild damage. The trigger conditions are crack length ≤ 2mm, strain gradient anomaly < 2με / m or looseness frequency deviation ≤ 0.2Hz. The response actions are local audible and visual alarms, pushing text messages / emails to the operation and maintenance platform, and prompting manual inspection. The second - level warning is moderate damage. The trigger conditions are crack length 2 - 5mm, strain gradient 2 - 5με / m or looseness frequency deviation 0.2 - 0.5Hz. The response actions are automatically generating maintenance plans and pushing maintenance guidelines to mobile terminals. The third - level warning is critical failure. The trigger conditions are crack length ≥ 5mm, strain gradient ≥ 5με / m or looseness frequency deviation ≥ 0.5Hz. The response actions are to link the fan control system to reduce load or emergency stop, start the emergency plan, trigger the expected threshold self - adaptation adjustment. The adjustment rule is based on the damage expansion rate of a 30 - day sliding window. When the wind speed ≥ 15m / s, the crack expansion threshold floats up by 20%. When the temperature fluctuation > 10℃, the strain gradient threshold decreases by 15%. Input damage parameters: crack length, delamination area and looseness degree. Environmental parameter data: wind speed, temperature and vibration load. Collect vibration signals through low - frequency vibration sensors, capture the dynamic responses of the structure generated by external excitations such as wind loads and mechanical vibrations. After noise reduction and feature extraction, obtain the vibration modal frequency deviation. Based on the finite - element model of the hybrid tower, combined with real - time vibration data, inversely deduce the amplitude and distribution of external vibration loads through inverse problem solving. Through the frequency - domain analysis of vibration signals, extract the main load frequency components, combine with the temperature data of distributed fiber optic sensors, and correct the influence of thermal expansion effects on vibration loads. According to wind speed data, optimize the aerodynamic damping coefficient in the load model. The prediction results include remaining life and crack propagation path. The cloud compares the current damage parameters with the dynamic thresholds. If any warning - level condition is met, trigger the corresponding response. When multiple conditions are met simultaneously, execute the highest - level warning, output the result warning instruction, local audible and visual alarm signal, text message / email notification content and fan control instruction, and give maintenance plans, recommended bolt re - tightening torque values and grouting repair process parameters.

[0046] The visualization system constructs a three-dimensional visualization interface based on GIS and BIM technologies to display the damage location, propagation path, and predicted life. It constructs a three-dimensional geometric model of the hybrid tower through BIM technology, and through AR glasses, superimposes the damage location and maintenance guidelines, and supports voice interaction to retrieve historical data. The core function of the visualization system is to intuitively present multi-modal monitoring data, damage prediction results, and maintenance guidelines in the form of three-dimensional visualization and augmented reality, providing intelligent support for operation and maintenance personnel in spatial positioning, damage assessment, and maintenance operations. Its goals include dynamic damage display: real-time annotation of damage location, propagation path, and remaining life; AR-assisted maintenance: superimposing maintenance guidelines through AR glasses to improve maintenance efficiency and accuracy; historical data backtracking: retrieving historical damage records and analyzing the evolution trend. Based on BIM technology, a three-dimensional geometric model of the hybrid tower is constructed, including details such as concrete segments, steel components, and flange connections, and material properties are imported. The damage information, environmental data, and prediction results collected in real time are obtained through GIS for the geographical coordinates, terrain, and surrounding environment data of the hybrid tower. The BIM model is exported in IFC format to ensure compatibility with the GIS platform. The monitoring data is converted into a format supported by the spatial database and bound to the geographical coordinates. Using the three-dimensional scene construction tool of ArcGIS Engine, the BIM model is embedded in the geographical space coordinate system to achieve the precise matching of the hybrid tower structure and the actual geographical location. Through the coordinate conversion algorithm, the model is ensured to be consistent with the real geographical environment. The damage parameters in the BIM model are updated in real time using Revit API and synchronized with the GIS database. Through the spatial database engine, the dynamic association between the monitoring data and the three-dimensional model is realized. In the BIM model, the damage location is displayed through color coding and three-dimensional coordinate annotation. Combining with the GIS map, a heat map is generated to show the correlation between the damage distribution and the geographical environment. Based on the crack propagation prediction of the digital twin module, the crack evolution direction is displayed as a dynamic trajectory line in the three-dimensional interface, and the remaining life prediction value is superimposed on the key parts of the model, supporting click to view the detailed confidence interval. The interactive function realizes view switching and data drilling, supporting various viewing modes such as top view, side view, and sectional view. Clicking on the damage point can retrieve historical monitoring data. Through AR glasses, the damage location and maintenance guidelines are superimposed on the real scene. Using the LOD technology, the model accuracy is dynamically adjusted according to the viewing angle to reduce the GPU load. WebGL is used to accelerate the graphics rendering. High-frequency data is preprocessed through edge computing, and only the key features are transmitted to the cloud. The incremental update algorithm is used to only refresh the changed part of the three-dimensional model. Input real-time damage data, environmental data, and maintenance plans. The real-time damage data includes damage type, location, size, crack propagation path, and remaining life. The environmental data includes wind speed, temperature, and vibration load. The maintenance plan includes bolt retightening torque and grouting parameters. The cloud server fuses the damage data with the BIM model to generate a three-dimensional visualization scene, and pushes the rendering result to the local terminal in real time through the 5G NR-U network. According to the damage location, the superimposed coordinates of the AR virtual marker are calculated.Match the maintenance process library, generate operation instructions, output the result 3D visualization interface, including damage location annotation, crack extension path, remaining life countdown, superimposed display of environmental parameters, AR maintenance instructions, virtual arrow navigation, highlight box positioning, voice prompts for key operation steps, generate historical analysis reports, including crack extension rate statistics, damage event timeline, and material degradation trend chart.

[0047] The whole process of the system is as follows:

[0048] S1. Dynamic topology optimization: Based on the mixed tower finite element model, the sensor density is dynamically increased in damage-sensitive areas, and the sensor weight is adjusted in real time according to the signal-to-noise ratio and data redundancy to optimize the monitoring network coverage.

[0049] S2. Data Collection

[0050] A. The piezoelectric ceramic array emits ultrasonic waves to detect concrete microcracks and delamination defects. The acoustic emission sensor captures transient elastic waves and locates sudden damage. The distributed optical fiber sensor monitors strain distribution and temperature field. The low-frequency vibration sensor collects vibration signals and identifies structural modal anomalies. The anemometer collects wind speed.

[0051] B. Data transmission: Low-frequency data is transmitted through a low-power wide area network, and high-frequency data is transmitted back in real time through a high-speed wireless network.

[0052] S3. Data preprocessing: Wavelet transform is used to eliminate wind load and temperature noise, and blind source separation is used to extract damage feature vectors. The feature vectors include ultrasonic attenuation coefficient, acoustic emission energy spectrum peak, optical fiber strain gradient anomaly and vibration mode frequency offset. A lightweight convolutional neural model is preliminarily run to screen potential damage signals.

[0053] S4, Cloud-based Intelligent Analysis

[0054] A. Input feature vector, convolutional neural network extracts crack morphology, spatial characteristics of delamination defects, long short memory neural network captures the time sequence of crack propagation, and outputs damage type, location and severity level.

[0055] B. The concrete damage model and steel fatigue model receive real-time damage data, build a high-fidelity structural twin, simulate the crack propagation path of the mixed tower structure, predict the remaining life, and provide a scientific basis for graded early warning.

[0056] S5. Gradual warning: According to the size of the crack, the three-level warning mechanism is dynamically triggered. The first-level warning is a local sound and light alarm, and a text message / email is pushed to the operation and maintenance platform. The second-level warning automatically generates a maintenance plan and pushes it to the mobile terminal. The third-level warning links the fan control system to reduce load or shut down urgently. Based on historical data and environmental parameters, the alarm logic is dynamically optimized.

[0057] S6, Visual feedback

[0058] A. Real-time display on the 3D visualization interface: The damage location is marked with 3D coordinates, the extended path is displayed with a red-highlighted trajectory, and the remaining life is displayed with a countdown;

[0059] B. AR-assisted maintenance: Overlay the damage location guidance, retrieve historical data through voice interaction, and generate a standardized maintenance process.

[0060] S7, Operation and maintenance response: Perform hierarchical operations according to the warning level. For the first-level response, manual inspection is carried out to confirm the damage. For the second-level response, bolt retightening or local reinforcement is carried out. For the third-level response, a comprehensive overhaul or replacement of damaged components is carried out after shutdown.

[0061] Finally, it should be noted that the above are only the preferred embodiments of the present invention and are not used to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. Intelligent perception array early warning system for the full-life damage of the hybrid tower structure, characterized in that: It includes a multi-modal data collection module, an array topology optimization module, an adaptive signal processing module, a digital twin life prediction module, a hierarchical warning module and a visualization system. The multi-modal data collection module includes a multi-modal sensor array, a self-powered module and a wireless transmission module. The self-powered module integrates a piezoelectric vibration power generation unit and a thermoelectric power generation module to provide long-term maintenance-free power for the sensors. The wireless transmission module uses a hybrid transmission of wide area network and wireless network. Low-frequency data is transmitted through the wide area network, and high-frequency data is real-time transmitted back through the wireless network.

2. The intelligent perception array early warning system for the full-life damage of the hybrid tower structure according to claim 1, wherein: The multi-modal sensor array forms a monitoring network through a piezoelectric ceramic array, a distributed optical fiber sensor, an acoustic emission sensor and a low-frequency vibration sensor. The piezoelectric ceramic array is embedded in the concrete section of the hybrid tower to emit and receive ultrasonic pulses, and its working frequency can be adaptively switched to detect defects of different scales. The distributed optical fiber sensor is arranged along the axial and circumferential directions of the tower barrel to form a monitoring network to monitor the strain distribution and temperature field change. The acoustic emission sensor is arranged in a ring array at the flange connection to capture the transient elastic wave signal generated by material fracture. The low-frequency vibration sensor is arranged at the top and the foundation section to collect low-frequency vibration signals and identify the structural modal frequency shift. A anemometer is installed at the top of the tower to collect wind speed.

3. The intelligent perception array early warning system for full-life damage of the hybrid tower structure according to claim 1, characterized in that: The array topology optimization module is based on the finite element model of the hybrid tower, dynamically densifies the sensor layout in the concrete-steel transition section and the flange connection in the damage-sensitive area to form a dense grid, and adjusts the sensor weights in real time according to the signal-to-noise ratio and data redundancy to optimize the monitoring network coverage. The signal-to-noise ratio and data redundancy are core parameters set based on communication standards, engineering practices and numerical simulations.

4. The intelligent perception array early warning system for the full-life damage of the hybrid tower structure according to claim 1, wherein: The adaptive signal processing module eliminates environmental noise through wavelet transform and blind source separation, extracts ultrasonic attenuation coefficient, peak value of acoustic emission energy spectrum, abnormal fiber strain gradient and vibration modal frequency shift, constructs a hybrid model containing convolutional neural network and long short-term memory neural network, extracts the spatial features of damage, and captures the time-series evolution law of damage.

5. The intelligent perception array early warning system for full-life damage of the hybrid tower structure according to claim 1, characterized in that: The digital twin life prediction module integrates a concrete damage model and a steel fatigue model, combines real-time data, periodically updates material parameters, simulates crack propagation using the finite element method, and evaluates the remaining life in combination with probability calculation methods.

6. The intelligent perception array warning system for full-life damage of the hybrid tower structure according to claim 1, characterized in that: The hierarchical warning module is provided with a three-level warning mechanism, and the warning threshold is dynamically adjusted according to historical data and environmental parameters.

7. The intelligent perception array early warning system for the full-life damage of the hybrid tower structure according to claim 1, wherein: The visualization system constructs a three-dimensional visualization interface based on GIS and BIM technologies to display the damage location, propagation path and predicted life. A three-dimensional geometric model of the hybrid tower is constructed through BIM technology. Through AR glasses, the damage location and repair guidelines are superimposed, and voice interaction is supported to retrieve historical data.

8. The intelligent perception array early warning system for the full-life damage of the hybrid tower structure according to claim 1, wherein: The full working process of the system is as follows: S1. Dynamic topology optimization: Based on the finite element model of the hybrid tower, dynamically densify the sensor density in the damage-sensitive area, and adjust the sensor weights in real time according to the signal-to-noise ratio and data redundancy to optimize the monitoring network coverage; S2. Data collection A. The piezoelectric ceramic array emits ultrasonic waves to detect concrete microcracks and delamination defects. The acoustic emission sensor captures transient elastic waves and locates sudden damage. The distributed optical fiber sensor monitors strain distribution and temperature field. The low-frequency vibration sensor collects vibration signals and identifies structural modal anomalies. The anemometer collects wind speed. B. Data transmission: Low-frequency data is transmitted through low-power wide area networks, and high-frequency data is transmitted back in real time through high-speed wireless networks; S3. Data preprocessing: Wavelet transform is used to eliminate wind load and temperature noise, and blind source separation is used to extract damage feature vectors. The feature vectors include ultrasonic attenuation coefficient, acoustic emission energy spectrum peak, fiber strain gradient anomaly, and vibration mode frequency offset. A lightweight convolutional neural model is preliminarily run to screen potential damage signals. S4, Cloud-based Intelligent Analysis A. Input feature vector, convolutional neural network extracts crack morphology and spatial characteristics of delamination defects, long short memory neural network captures the time sequence of crack extension, and outputs damage type, location and severity level; B. The concrete damage model and steel fatigue model receive real-time damage data, build a high-fidelity structural twin, simulate the crack propagation path of the mixed tower structure, predict the remaining life, and provide a scientific basis for graded warning; S5, graded warning: According to the size of the crack, the three-level warning mechanism is dynamically triggered. The first-level warning is a local sound and light alarm, and SMS / email is pushed to the operation and maintenance platform. The second-level warning automatically generates a maintenance plan and pushes it to the mobile terminal. The third-level warning links the fan control system to reduce load or emergency shutdown. Based on historical data and environmental parameters, the alarm logic is dynamically optimized; S6. Visual feedback A. Real-time display of 3D visualization interface: 3D coordinates mark the damage location, red highlighted trajectory shows the expansion path and countdown shows the remaining life; B. AR-assisted maintenance: superimpose damage location guidance, call up historical data through voice interaction, and generate standardized maintenance processes; S7. Operation and maintenance response: Perform graded operations according to the warning level. The first-level response is manual inspection to confirm damage, the second-level response is bolt retightening or local reinforcement, and the third-level response is comprehensive inspection or replacement of damaged parts after shutdown.

Images