A method and system for calculating the structural strength of a glass steel column

Abstract

This invention discloses a method for calculating the structural strength of a fiberglass tower, comprising the following steps: acquiring real-time sensor data and static design parameters of the fiberglass tower; constructing a structural digital twin model synchronously mapped to the physical entity of the fiberglass tower based on the real-time sensor data and static design parameters; injecting the real-time sensor data as boundary conditions into the digital twin model; driving the digital twin model to perform structural strength evolution analysis under multiple working conditions; and calculating the dynamic safety factor of the fiberglass tower based on the structural failure boundary and real-time sensor data. This invention also discloses a structural strength calculation system, comprising: a data acquisition module, a digital twin model construction module, a digital twin engine module, a safety assessment module, and a human-computer interaction module. This invention achieves real-time and accurate assessment of the structural strength of fiberglass towers by constructing a digital twin model, solving the technical problem that traditional calculation methods cannot dynamically reflect the structural performance degradation process.

Description

Technical Field

[0001] This invention relates to the field of digital twin and structural simulation technology; and in particular to a method and system for calculating the structural strength of a fiberglass tower. Background Technology

[0002] Fiberglass reinforced plastic (FRP) towers are widely used in waste gas treatment, liquid storage, and chemical reaction processes in chemical, environmental protection, and marine engineering fields due to their excellent corrosion resistance and lightweight, high-strength characteristics. However, with the increasing size of industrial plants and the growing complexity of service environments, FRP towers face safety hazards such as structural strength degradation, cumulative localized damage, and sudden failures during long-term service.

[0003] Currently, strength assessment for engineering structures such as FRP towers mainly relies on the following methods: (1) Traditional theoretical calculation method: Based on the theory of composite material mechanics, the stress distribution of the tower under a specific load is calculated by analytical formula. This method has high calculation efficiency, but it cannot take into account the actual damage state and boundary condition changes of the tower, and the calculation accuracy is limited.

[0004] (2) Finite element simulation analysis method: Static or dynamic analysis is performed by establishing a three-dimensional finite element model of the tower and applying load boundary conditions. This method can obtain a relatively accurate stress distribution, but it is an offline analysis and cannot reflect the evolution of the tower's structural performance in real time during service.

[0005] (3) Periodic non-destructive testing method: The tower is periodically inspected using ultrasonic waves, infrared thermal imaging, etc., to assess its structural health status. This method can detect existing damage, but it is an intermittent inspection and cannot achieve continuous monitoring and early warning.

[0006] In recent years, digital twin technology has provided new approaches for structural strength assessment. For example, patent CN117973156A, which describes a digital twin simulation method for a strength testing system oriented towards real-world testing conditions, establishes a high-precision finite element model by incorporating manufacturing deviations from real test specimens. Based on the finite element simulation and real test results, a data fusion base model is established, enabling full-field monitoring of the test specimen's strength response. This method efficiently reduces the simulation dataset through clustering algorithms, decreasing the amount of finite element data and effectively improving computational efficiency. However, this method primarily focuses on strength assessment during the testing phase and has not yet addressed the long-term dynamic tracking of structural performance under actual service conditions.

[0007] Furthermore, patent CN119962124A, entitled "A Structural Strength Performance Evaluation Method Based on Multi-Level Virtual-Real Fusion," constructs a displacement field twin evaluation model and a strain field twin evaluation model. Through a synchronous strain-displacement twin evaluation model, it achieves real-time analysis and anomaly warning of structural strength parameters. This method solves the technical problem that existing strength tests cannot evaluate the strain and displacement parameters of the tested component in real time. However, this method is mainly aimed at metal structural components, and its applicability to anisotropic composite materials such as fiberglass has not been fully verified. Moreover, it does not address the impact of material performance degradation over service time on structural strength.

[0008] In summary, the common technical problems of existing methods are: they cannot achieve real-time dynamic assessment of the structural strength of FRP towers, they are difficult to reflect the impact of material performance degradation and cumulative damage evolution on structural strength, and when anomalies are detected, irreversible damage or even failure has often already occurred.

[0009] Therefore, there is an urgent need for a method and system that can realize real-time calculation, dynamic evaluation and early warning of the structural strength of FRP towers. Summary of the Invention

[0010] The purpose of this invention is to overcome the defects in the prior art and provide a method for calculating the structural strength of fiberglass towers, so as to solve the technical problem that the structural strength of fiberglass towers cannot be dynamically evaluated in real time in the prior art.

[0011] To achieve the above objectives, the technical solution of this invention is to design a method for calculating the structural strength of a fiberglass tower, comprising the following steps: Step S1: Obtain real-time sensing data and static design parameters of the fiberglass tower. The real-time sensing data includes stress-strain data, vibration frequency data, and corrosion potential data. The static design parameters include geometric dimension parameters, layup parameters, and material property parameters. Step S2: Based on the real-time sensing data and static design parameters, construct a structural digital twin model that is synchronously mapped to the physical entity of the FRP tower. Step S3: Inject the real-time sensing data as boundary conditions into the digital twin model, and combine it with historical load data to generate a multi-condition simulation scenario including normal and extreme conditions. Step S4: Drive the digital twin model to perform structural strength evolution analysis under multiple working conditions, construct the load-stress mapping relationship based on the analysis results, and predict the structural failure boundary of the FRP tower. Step S5: Calculate the dynamic safety factor of the fiberglass tower based on the structural failure boundary and real-time sensor data, and feed back early warning information on the weak structural areas to the physical entity through the digital twin model.

[0012] A further technical solution is that the specific steps for constructing the structural digital twin model in step S2 include: S21. Establish a three-dimensional solid model based on the geometric parameters of the fiberglass tower; S22. Perform finite element mesh generation on the three-dimensional solid model and define an anisotropic composite material constitutive model based on material property parameters; S23. Bind real-time sensing data as boundary constraints to the corresponding nodes of the 3D solid model to form a real-time data-driven dynamic digital twin.

[0013] A further technical solution is that, in step S2, when constructing the structural digital twin model, a physical field simulation model is constructed using finite element theory, while a reduced-order proxy model is constructed using deep neural networks, forming a hybrid driving architecture of physical information neural networks.

[0014] A further technical solution is to use a generative adversarial network or variational autoencoder to augment the extreme operating condition data when predicting the structural failure boundary in step S4, thereby expanding the failure mode data under small sample conditions and improving the accuracy of structural failure boundary prediction.

[0015] A further technical solution is that, in step S4, when performing structural strength evolution analysis, a neural network proxy model is used to replace part of the physical field simulation calculation. The neural network proxy model is trained based on historical simulation data and is used to respond to structural strength prediction requests in real time.

[0016] A further technical solution is that the specific steps for constructing the load-stress mapping relationship in step S4 include: S41. Collect load data and corresponding stress response data under multiple working conditions to construct a training dataset; S42. Based on the training dataset, a load-stress mapping proxy model is established using support vector regression or random forest algorithm; S43. Embed the load-stress mapping proxy model into the digital twin model to achieve rapid prediction of structural stress under arbitrary load input.

[0017] A further technical solution is that, after step S5, the following is also included: S6. The calculated dynamic safety factor and early warning information of structurally weak areas are fed back to the physical entity of the FRP tower or the remote operation and maintenance platform through the data interface to realize closed-loop management of the tower's structural health.

[0018] The real-time sensing data in step S1 is collected by a wireless sensor network deployed in key parts of the fiberglass tower, and after data cleaning and feature extraction by edge computing nodes, it is uploaded to the cloud digital twin platform.

[0019] A further technical solution is that the structural strength calculation method for fiberglass towers also includes a model adaptive update step: when the deviation between real-time sensing data and the simulation results of the digital twin model exceeds a preset threshold, a model parameter calibration mechanism is triggered, and the digital twin model is incrementally learned and iteratively optimized using the latest sensing data.

[0020] The specific steps for calculating the dynamic safety factor in step S5 include: S51. Extract the maximum stress value σ_max of the FRP tower under the current working condition based on real-time sensor data; S52. Determine the remaining strength S_remaining of the FRP tower under the current damage state based on the structural failure boundary. S53. Calculate the dynamic safety factor SF=S_remaining / (γ×σ_max), where γ is the material partial factor; S54. When the dynamic safety factor SF is less than a preset threshold, a warning signal is triggered and the location information of the structurally weak area is output.

[0021] The present invention also provides a technical solution: a structural strength calculation system for fiberglass towers, comprising: The data acquisition module is used to acquire real-time sensor data and static design parameters of the fiberglass tower. A digital twin model construction module, connected to the data acquisition module, is used to construct a structural digital twin model based on the real-time sensing data and static design parameters; The digital twin engine module, connected to the digital twin model construction module, includes a physical field simulation unit and an artificial intelligence analysis unit, used to drive the digital twin model to perform multi-condition structural strength evolution analysis; A safety assessment module, connected to the digital twin engine module, is used to calculate the dynamic safety factor and predict the structural failure boundary; The human-computer interaction module is used to visually display the status of the twin model and output safety warning information.

[0022] A further technical solution is that the digital twin engine module also includes: The physics simulation unit is used for high-fidelity structural simulation of FRP towers based on the finite element method. Reduced-order model unit, used to build simulation proxy models based on deep neural networks to accelerate structural response prediction; The data augmentation unit is used to augment extreme condition data using generative adversarial networks.

[0023] The data acquisition module includes: Wireless sensor networks are deployed in key parts of FRP towers to collect data on stress and strain, vibration frequency, and corrosion potential. Edge computing nodes are used to perform data cleaning, feature extraction, and anomaly detection on the collected real-time sensor data. The data upload interface is used to upload the processed data to the cloud-based digital twin platform.

[0024] The structural strength calculation system also includes: The model adaptive update module, connected to the digital twin engine module, is used to trigger model parameter calibration and incremental learning update when the deviation between real-time sensing data and simulation results exceeds a preset threshold.

[0025] The human-computer interaction module includes: The 3D visualization unit is used to display the structural stress distribution cloud map of the digital twin model of the FRP tower in real time in a 3D graphical manner; The early warning push unit is used to push early warning information and the location of structurally weak areas to the remote operation and maintenance platform when the dynamic safety factor is lower than a preset threshold. The report generation unit is used to generate a structural strength assessment report for FRP towers, including a safety factor curve, failure probability prediction, and maintenance recommendations.

[0026] The advantages and beneficial effects of this invention are as follows: By constructing a digital twin model and embedding a reduced-order proxy model, real-time calculation and dynamic evaluation of the structural strength of FRP towers were achieved, overcoming the shortcomings of traditional finite element analysis, which is time-consuming and unable to respond in real time.

[0027] By injecting real-time sensor data as boundary conditions into the digital twin model, the simulation analysis can reflect the actual damage state and boundary condition changes of the tower, thus improving the accuracy of structural strength assessment.

[0028] By analyzing the structural strength evolution and predicting the failure boundary, the degradation trend of FRP tower structure performance was predicted, providing a basis for decision-making in preventive maintenance.

[0029] Through twin data feedback and model adaptive updates, a closed-loop management mechanism of "perception-analysis-early warning-optimization" has been formed, continuously improving the accuracy and reliability of the digital twin model.

[0030] This invention belongs to the field of emerging software and new information technology services, and involves emerging technologies such as digital twins, artificial intelligence agent models, and edge computing, which have good prospects for industrial application. Attached Figure Description

[0031] Figure 1This is a flowchart illustrating an embodiment of the structural strength calculation method for a fiberglass tower according to the present invention. Figure 2 This is the structural strength calculation system for the fiberglass tower of Embodiment 2 of the present invention; Figure 3 yes Figure 2 Structural block diagram of the human-computer interaction module; Figure 4 yes Figure 3 The three-dimensional stress distribution cloud map output by the human-computer interaction module; Figure 5 This is a three-dimensional schematic diagram of the fiberglass tower body in Embodiment 3 of the present invention; Figure 6 yes Figure 5 The main view; Figure 7 yes Figure 6 A magnified view of a portion of the middle slider; Figure 8 This is a schematic diagram of Embodiment 4 of the present invention; Figure 9 yes Figure 8 A magnified view of the upper part; Figure 10 yes Figure 9 A schematic diagram of the decomposition process; Figure 11 yes Figure 9 Exploded view of the unidirectional drive mechanism.

[0032] In the diagram: 1. Data acquisition module; 2. Digital twin model construction module; 3. Digital twin engine module; 4. Security assessment module; 5. Human-computer interaction module; 6. Model adaptive update module; 14. Edge computing node; 51. 3D visualization unit; 52. Early warning push unit; 53. Report generation unit; 100. Fiberglass tower; 101. Track; 102. Slider; 103. Wind resistance vane; 1033. Windward limiting block; 1034. Downwind limiting block; 401. First flange; 402. Second flange; 403. Elastic gasket; 404. Wedge sealing ring; 405. One-way drive mechanism; 406. Limiting ring; 407. Displacement sensor; 4054. External threaded sleeve; 4055. Internal threaded ring; 4056. Inertial weight. Detailed Implementation

[0033] The specific embodiments of the present invention will be further described below with reference to the accompanying drawings and examples. The following examples are only used to more clearly illustrate the technical solutions of the present invention and should not be construed as limiting the scope of protection of the present invention.

[0034] Example 1: As Figure 1As shown, this invention provides a method for calculating the structural strength of a fiberglass tower, comprising the following steps: Step S101: Obtain real-time sensor data and static design parameters of the fiberglass tower.

[0035] Taking a fiberglass waste gas absorption tower with a diameter of 4m and a height of 25m in a chemical plant as an example, fiber optic strain sensors, acceleration sensors and corrosion potential sensors are deployed in key parts such as the middle, lower and bottom skirts of the tower body, with the sampling frequency set to 100Hz.

[0036] The collected real-time sensing data includes: stress and strain data such as circumferential strain ε_c and axial strain ε_a; vibration frequency data such as first natural frequency f_1 and second natural frequency f_2; and corrosion potential data, namely corrosion potential E_corr (unit: mV, relative to a saturated calomel reference electrode), used to assess the degree of corrosion activity of the inner wall of the tower. The measured corrosion potential is converted into cumulative corrosion thinning using a pre-established empirical model of corrosion potential-corrosion rate (calibrated by accelerated corrosion tests in the laboratory).

[0037] Static design parameters are extracted from the design drawings, including: geometric dimensions such as tower inner diameter D_i=4000mm, wall thickness t=25mm, height H=25000mm; ply parameters: [±45° / 0° / 90°]s symmetrical ply, 16 layers in total; Material properties include fiber modulus E_11 = 33 GPa, transverse modulus E_22 = 11.5 GPa, shear modulus G_12 = 4.5 GPa, and Poisson's ratio ν_12 = 0.35. Edge computing node 14 has a built-in data preprocessing workflow, which is executed sequentially: (1) Anomaly detection: Based on the 3σ principle, outliers that deviate from the mean by more than 3 times the standard deviation are removed; (2) Median filtering: The window length is set to 5 to remove impulse noise; (3) Feature extraction: Calculate the root mean square value (RMS), peak value and frequency domain dominant frequency in the time domain.

[0038] The processed data is uploaded to the cloud-based digital twin platform via a 5G module.

[0039] Step S102: Construct a structural digital twin model that is synchronously mapped to the physical entity of the fiberglass tower.

[0040] Building a digital twin model includes the following sub-steps: Sub-step S1021: Based on the geometric parameters of the FRP tower, establish a 3D solid model in ANSYS or COMSOL Multiphysics software. Considering the axisymmetric characteristics of the tower, a 1 / 4 scale model is established to reduce computational load. Symmetrical boundary conditions (zero normal displacement constraint, tangential freedom) are applied to the three symmetry planes. All 50 working cases are solved on the 1 / 4 scale model. After solving, the stress distribution of the full model is restored through symmetry mapping.

[0041] Sub-step S1022: Perform finite element mesh generation on the 3D solid model. The cylindrical section uses 8-node hexahedral elements (SOLID185), with the mesh size controlled within 20mm, and localized mesh refinement in the opening and connection areas. Define an anisotropic composite material constitutive model based on material property parameters, inputting the aforementioned parameters E_11, E_22, G_12, ν_12, etc., and define a failure criterion. The Tsai-Wu criterion is adopted, and its expression is: F_i.σ_i+F_ij.σ_i.σ_j≤1 Where F_i and F_ij are the intensity tensor components, and σ_i and σ_j are the principal axial stress components. For the plane stress state, the specific expression is: F_1.σ_1+F_2.σ_2+F_11.σ_1 2 +F_22.σ_2 2 +F_66.τ_12 2 +2F_12.σ_1.σ_2≤1 In the formula, σ_1 is the fiber direction stress, σ_2 is the transverse stress, and τ_12 is the in-plane shear stress; the strength tensor components F_1, F_2, F_11, F_22, F_66, and F_12 are determined by the basic strength parameters (fiber direction tensile strength X_t, fiber direction compressive strength X_c, transverse tensile strength Y_t, transverse compressive strength Y_c, and in-plane shear strength S) through standard formulas.

[0042] Sub-step S1023: Bind the real-time sensing data as boundary constraints to the corresponding nodes of the 3D solid model. Specifically, apply the measured strain values ​​at the strain sensor measurement points as displacement boundary conditions to the corresponding nodes of the model, and use the vibration frequency data to calibrate the elastic constants of the model, forming a real-time data-driven dynamic digital twin.

[0043] Simultaneously, a reduced-order surrogate model is constructed using a Physical Information Neural Network (PINN). The inputs to PINN are load parameters and position coordinates, and the output is stress components. The loss function comprises two parts: data fitting loss and physical constraint loss. After training, the computational speed of the PINN surrogate model is 3-4 orders of magnitude faster than traditional finite element analysis. Data fitting loss refers to the deviation from the finite element simulation results or measured data; physical constraint loss refers to the residuals of the equilibrium equations and constitutive relations.

[0044] Step S103: Inject real-time sensor data as boundary conditions into the digital twin model to generate a multi-condition simulation scenario.

[0045] The measured strain data at the current moment is applied as the displacement boundary condition to the digital twin model. Simultaneously, load spectrum records for the past 12 months are extracted from the historical load database, including: operating pressure, wind load, and seismic load. The operating pressure is 0.2-0.5 MPa; the wind load is the basic wind pressure of 0.35 kN / m². 2 The seismic load fortification intensity is 7 degrees.

[0046] Based on the above data, 50 simulation scenarios were automatically generated using a working condition generation algorithm, covering: 20 normal operating conditions: combinations of operating pressure, wind load, and temperature load; 20 design reference conditions: combinations of design pressure, ultimate wind load, and seismic load; 10 extreme operating conditions: accidental impact, sudden temperature change, and extreme state after corrosion thinning.

[0047] Step S104: Drive the digital twin model to perform structural strength evolution analysis and predict the failure boundary.

[0048] The finite element method was used to solve 50 working conditions sequentially, and the maximum principal stress σ_1 and the maximum von Mises stress σ_v under each working condition were extracted.

[0049] Based on the calculation results of 50 finite element simulation cases, 200 supplementary sample points were first generated in the load parameter space using Latin hypercube sampling. The stress response was then rapidly predicted using a PINN surrogate model, resulting in 250 input-output data pairs. On this basis, a load-stress mapping relationship was constructed: using load parameters (pressure P, wind load W, and seismic acceleration a) as input and the maximum stress σ_max as output, a surrogate model was established using a random forest algorithm. The random forest contains 100 decision trees, employing bootstrap sampling and random feature selection. After training, the model was tested on the R... 2 It reaches 0.96 or higher.

[0050] Failure Boundary Prediction: Using 250 sets of data as the training set, a variational autoencoder (VAE) was employed to augment the extreme load conditions. The VAE encoder compressed the original load parameters into a 2D latent variable z, and the decoder reconstructed the load parameters from z. By performing Gaussian interpolation sampling in the latent space, 5000 virtual extreme load condition samples were generated and fed into a digital twin model to calculate the corresponding stress response. The load-stress surface was plotted to determine the critical load combination where the stress exceeds the material strength, i.e., the structural failure boundary.

[0051] Step S105: Calculate the dynamic safety factor and output the early warning information.

[0052] Based on the real-time sensor data at the current moment, the measured value of the maximum stress in the tower body is extracted as σ_max_sensor = 85.3MPa.

[0053] In its original, uncorroded state, the tensile strength of the fiberglass blade root, S_initial, is 500 MPa. Based on the structural failure boundary, the residual strength, S_remaining, under the current damage state is determined. Using the measured corrosion potential E_corr = -380 mV, and employing Faraday's law and an empirical model of the electrochemical parameters and corrosion rate of fiberglass in corrosive environments (this model was pre-calibrated through accelerated corrosion tests in the laboratory, establishing the relationship between corrosion potential, corrosion current density, and annual corrosion depth), the equivalent corrosion thinning (annual corrosion depth × service time) over an 8-year service life is calculated to be 1.2 mm. Interpolation yields the residual strength S_remaining = 312 MPa.

[0054] Taking the material partial factor γ = 2.205 (referring to the GL2010 standard), calculate the dynamic safety factor: SF=S_remaining / (γ_Ma×σ_max)=312 / (2.205×85.3)=1.66 Where SF is the dynamic safety factor, S_remaining is the remaining strength under the current damage state (value is 312MPa), γ_Ma is the material partial factor (value is 2.205), and σ_max is the measured maximum stress (value is 85.3MPa).

[0055] The preset safety threshold is set to 1.5. Since SF=1.66>1.5, the current structural condition is safe, but it is close to the threshold. The system outputs a warning message: It is recommended to arrange a shutdown for maintenance within 6 months, focusing on the circumferential weld seam in the middle of the cylinder, 8m-10m from the bottom.

[0056] Step S106: Closed-loop management and model update.

[0057] The warning information is pushed to the factory's remote operation and maintenance platform through the data interface, and simultaneously sent to the equipment manager via a mobile APP.

[0058] During subsequent operation, when the absolute value of the relative deviation between the measured stress values ​​and the model prediction values ​​at key monitoring points (mid-section of the cylinder and skirt connection) continuously exceeds 10% and lasts for more than 24 hours, an adaptive model update is triggered. The relative deviation is defined as |σ_measured-σ_predicted| / σ_measured×100%. The adaptive model update adopts a transfer learning strategy: using the weights of the original PINN model as initial values, the weights of the output layer are fine-tuned using the sensor data from the most recent 30 days (after anomaly detection and filtering), while the network layers of the physical constraint part remain frozen. The fine-tuning cycle is once every 30 days, with each iteration lasting 50 epochs, enabling the model to reflect the gradual degradation trend of the tower structure performance.

[0059] Example 2: Structural Strength Calculation System for Fiberglass Towers See Figure 2 This embodiment provides a structural strength calculation system for fiberglass towers, including a data acquisition module 1, a digital twin model construction module 2, a digital twin engine module 3, a safety assessment module 4, a human-computer interaction module 5, and a model adaptive update module 6.

[0060] Data acquisition module 1 is deployed in key parts of the FRP tower 100, including fiber optic strain sensors, acceleration sensors, corrosion potential sensors, and edge computing nodes 14. Fiber optic strain sensors are located in the middle of the tower body, for example, at 1 / 2H from the bottom, with 4-8 points evenly distributed circumferentially. They are also located above the skirt connection. Each measuring point is equipped with a biaxial fiber optic strain sensor (one sensitive grating along the circumferential direction and one sensitive grating along the axial direction), with 4 measuring points evenly distributed circumferentially (spaced 90° apart) to simultaneously monitor circumferential and axial strain. They are also generally located at the edges of openings. The fiber optic sensors are adhered to the outer surface of the tower body. Before adhesion, the surface needs to be sanded and cleaned, and a special epoxy resin adhesive is used for bonding. A protective coating such as silicone or polyurethane is applied to prevent UV aging and mechanical damage. Accelerometers are set at the center of the top head of the tower, with two horizontal bidirectional (X, Y) sensors and three points evenly distributed around the circumference in the middle of the cylinder. They are also generally set at the bottom skirt of the tower 1m from the ground. Corrosion potential sensors are generally set in the liquid accumulation area at the bottom of the tower, such as the inner wall 200~500mm from the bottom of the tower, usually 1~2. They are also set in the normal working liquid level range of 500mm above and below the gas-liquid interface and inside the inlet and outlet of the pipe. The sensors communicate with the edge computing node 14 wirelessly. The edge computing node 14 has a built-in data preprocessing process, which is executed in sequence as follows: (1) Anomaly detection: Based on the 3σ principle, outliers with a deviation from the mean of more than 3 times the standard deviation are removed; (2) Median filtering: The window length is set to 5 to remove impulse noise; (3) Feature extraction: The root mean square value (RMS), peak value and frequency domain main frequency are calculated. The processed data is uploaded to the cloud digital twin platform through the 5G module.

[0061] The digital twin model building module 2 is connected to the data acquisition module 1. Based on the received sensor data and static design parameters, it builds a structural digital twin model on a cloud server. This module calls the APIs of ANSYS or COMSOL for finite element modeling, and simultaneously calls the TensorFlow or PyTorch framework to build the PINN proxy model.

[0062] The digital twin engine module 3 is connected to the digital twin model building module 2. The digital twin engine module 3 includes: Physics simulation unit: High-fidelity structural simulation based on the finite element method; Reduced-order model unit: Based on deep neural networks, a simulation proxy model is constructed to accelerate structural response prediction; Data augmentation unit: Uses generative adversarial networks to augment extreme condition data; The safety assessment module 4 is connected to the digital twin engine module 3 to perform dynamic safety factor calculation and failure boundary prediction functions.

[0063] The human-computer interaction module 5 is used to visually display the status of the twin model and output safety warning information. For example... Figure 3 As shown, the human-computer interaction module 5 includes a 3D visualization unit 51, an early warning push unit 52, and a report generation unit 53. The 3D visualization unit 51 displays the structural stress distribution cloud map of the digital twin model of the fiberglass tower in real time using 3D graphics (see...). Figure 4 The function of the early warning push unit 52 is to push early warning information to the remote operation and maintenance platform when the dynamic safety factor is lower than the preset threshold; the report generation unit 53 is used to generate a structural strength assessment report for the FRP tower, including the safety factor change curve, failure probability prediction and maintenance suggestions.

[0064] The model adaptive update module 6 is connected to the digital twin engine module 3. When the deviation between the real-time sensing data and the simulation results exceeds a preset threshold, it triggers model parameter calibration and incremental learning update.

[0065] Technical effect verification Using the method and system of this embodiment, a 4m diameter, 25m high fiberglass reinforced plastic (FRP) exhaust gas absorption tower in a chemical plant was subjected to online monitoring and dynamic evaluation for a period of 6 months. The results showed that: (1) The PINN proxy model has a single stress prediction time of <0.1 seconds, which meets the real-time monitoring requirements, while traditional finite element analysis takes about 30 minutes per analysis.

[0066] (2) The average relative error between the stress value predicted by the digital twin model and the measured value is 5.8%, which is better than the 15.2% of the traditional theoretical calculation method.

[0067] (3) The system successfully issued an early warning of a structural stress over-limit event caused by an abnormal increase in wind load, which gave the operators time to adjust the process parameters in a timely manner and avoided potential equipment damage.

[0068] (4) As a typical application of emerging software and new information technology services, this system realizes intelligent management of the structural health of industrial equipment and has good promotion value.

[0069] Example 3: The difference from Example 1 is that a mechanical structure device is provided to be used in conjunction with the aforementioned structural strength calculation system to dynamically offset the additional bending moment caused by wind load and reduce the circumferential stress difference between the windward and leeward sides of the FRP tower.

[0070] See Figures 5 to 7 (For ease of illustration, Figure 5(The circular track is not shown). The mechanical structure device in this embodiment is specifically a wind-driven adaptive counterweight stress adjustment device, which includes a circular track 101 fixedly installed on the outer wall of the fiberglass tower 100, a sliding counterweight slider 102 slidably disposed on the track 101, and wind resistance vanes 103 hinged to the front and rear sides of the slider 102 along the track movement direction; a reset elastic element connected between the slider 102 and the windward end of the track 101 and limit buffers disposed at the lowest point (windward side) and the highest point (leeward side) of the track 101 are also provided to absorb the impact energy when the slider 102 reaches the two ends of the track.

[0071] The track encircles the fiberglass tower body circumferentially. The track is an inclined circular track with an angle between its surface and the horizontal plane, being lowest on the windward side and highest on the leeward side; that is, the lowest point of the track is on the windward side of the tower body, and the highest point is on the leeward side. (This embodiment requires a relatively fixed prevailing wind direction in the area where the tower is located, such as coastal areas or canyon passes; during installation, the prevailing wind direction must be determined based on local meteorological data, aligning the highest point of the track with the leeward side and the lowest point with the windward side. If the actual wind direction deviates from the preset direction, for example, within 30°, this device can still generate some reverse bending moment, but the effect will be reduced.) The slider 102 can move freely along the track. The slider is made of high-density materials such as cast iron or lead alloy, and its mass is determined based on the tower body's design wind load bending moment calculation, typically 1% to 5% of the tower body's self-weight. Wind resistance blades 103 are hinged to the front and rear sides of the slider 102 along the track's direction of movement. Each wind resistance vane 103 includes a vane body in the shape of a rectangular flat plate, a hinge shaft, a windward limiting block 1033 and a downwind limiting block 1034 fixed on a slider 102; the vane body has a front (windward side) and a back (leeward side) with opposite sides, and a hinge edge and a free edge connecting the front and back sides, and the thickness of the vane body is t; the hinge shaft is arranged along the hinge edge of the vane body and is used to hinge the vane body to the slider 102; the center line of the hinge shaft is offset from the front of the vane body and is biased towards the back side. Specifically, the distance δ1 from the hinge shaft to the front is 0.6t~0.8t, and the distance δ2 from the back is 0.2t~0.4t, that is, the hinge shaft is arranged close to the back side; the downwind limiting block 1034 is located on the back side of the vane body and is used to limit the minimum closing angle of the vane. A windward limiting block 1033 is fixed to the upper surface of the slider 102 near the hinge axis, and a limiting surface is provided at the root of the wing 1031. When the wing rotates upward to a set angle (45°~90°), the limiting surface contacts the limiting block 1033, limiting the maximum opening angle of the wing. When the wing is closed, the wing body lies flat on the upper surface of the slider 102, and the upper surface itself serves as the closing support surface, so no additional limiting structure is needed; if a downwind limiting block 1034 is provided on the upper surface, the downwind limiting block 1034 serves as the limiting structure when the wind is blowing in the direction of the wind.

[0072] Its working principle is as follows: When the wind blows towards the front of the winglet body: the wind pressure acts directly on the front. Because the hinge shaft is located near the back, the wind pressure generates an opening torque on the hinge shaft, causing the winglet body to rotate around the hinge shaft until it contacts the windward limiting block 1033, reaching the open state. When the wind blows towards the back of the winglet body: the wind pressure acts directly on the back. Because the hinge shaft is near the back, the lever arm of the wind pressure on the hinge shaft is extremely small, and the resulting opening torque is also extremely small. When the winglet is in the closed state, the wind pressure on the back is mainly borne by the slider housing, and the winglet body is pressed against the downwind limiting block 1034, maintaining the closed state. It should be noted that the "windward opening and downwind closing" characteristic of this winglet mechanism refers to the local relationship between the winglet itself and the wind direction. Regardless of how the actual wind direction changes, this micro-mechanism can work adaptively. However, the effectiveness of the entire device still depends on the matching of the fixed tilt direction of the track with the prevailing wind direction. This embodiment is suitable for application scenarios where the prevailing wind direction is relatively fixed (such as monsoon areas or coastal areas where a certain direction prevails all year round; since the wind direction is basically fixed, it is better to use a single wing only on the front side, because the rear wing is always closed, so the rear wing can be omitted).

[0073] The reset elastic element is a tension spring or a rubber elastic rope, used to pull the slider back to its initial position on the windward side when there is no wind. The limiting buffer includes a first limiting buffer and a second limiting buffer. The first limiting buffer is located at the lowest point of track 101 (i.e., the windward side), and the second limiting buffer is located at the highest point of track 101 (i.e., the leeward side). Each limiting buffer includes: a buffer seat fixed to the end of the track, an elastic body disposed on the buffer seat, and a buffer head disposed on the end of the elastic body facing the slider. When the slider 102 moves to the end of track 101, the slider 102 contacts the buffer head and compresses the elastic body, absorbing impact energy.

[0074] The working principle is as follows: Large fiberglass towers (especially at openings, nozzles, skirt connections, and diameter transition sections) are prone to generating localized stress peaks far exceeding design values ​​due to geometric discontinuities and the anisotropic nature of the material, leading to localized buckling or creep failure. For example, under wind loads, the fiberglass tower body undergoes bending deformation. The windward side bears circumferential compressive stress (pointing towards the tower center), while the leeward side bears circumferential tensile stress (away from the tower center). This circumferential stress difference causes elliptical deformation of the tower's horizontal cross-section: the windward sides bulge outwards, while the central region of the leeward side contracts inwards. When the circumferential stress exceeds the material's critical value, it may trigger elliptical instability or localized buckling of the tower body. This embodiment adopts a reverse approach, changing the conventional method of combating wind loads to "guiding" the energy of the wind load to improve the structural stress state. Utilizing the wind load itself, through a passive mechanical device, the originally harmful wind pressure is transformed into a means of mitigating stress differences. The specific principle is as follows: In windless or lightly windy conditions, the slider 102, under the tension of the reset elastic element 104, stops at the lowest point on the windward side of the circular track 101. At this time, the mass of the slider is concentrated on the windward side of the tower, but due to the symmetrical arrangement (multiple devices can be evenly arranged along the circumference, or the small mass of the slider itself) and the balance of the reset elastic element, the additional static load bending moment generated on the tower is close to zero.

[0075] Under strong wind conditions, i.e., when wind load is applied, the wind blows from the windward side to the leeward side. The drag vanes 103 on the slider 102 open under wind pressure, increasing the slider's drag coefficient. The wind force propels the slider to slide along the annular track 101 towards the leeward side. Because the track is spatially inclined (lower on the windward side, higher on the leeward side), the slider slides from the windward side to the leeward side under the action of the wind force. When the slider reaches the leeward side, its line of action of gravity coincides with the direction of the wind load bending moment but in the opposite direction. At this time, the additional bending moment generated by gravity is completely opposite to the wind load bending moment, and the cancellation effect reaches its maximum. At this time, the slider is located on the leeward side, and its point of action of gravity is deviated from the center line of the tower, generating an additional bending moment that causes the tower to bend towards the windward side. The direction of this additional bending moment is opposite to the wind load bending moment (causing the tower to bend towards the leeward side). After the two are superimposed, the net bending moment borne by the tower is reduced, thereby reducing the circumferential compressive stress on the windward side and the circumferential tensile stress on the leeward side.

[0076] When the wind direction remains constant but the wind speed changes, as the wind speed increases, the wind's ability to overcome spring force and gravity increases, allowing the slider to slide closer to the highest point on the leeward side, generating a larger reverse bending moment. When the wind speed decreases, the slider retracts under the influence of gravity and spring force, reducing the reverse bending moment. This achieves wind speed self-adaptation.

[0077] When the wind stops, and the wind speed decreases to a level that cannot overcome the tension of the reset elastic element 104, the slider automatically returns to its initial position on the windward side under the action of the elastic element, thus avoiding the adverse effects of long-term static load on the tower body.

[0078] The key parameters are shown in the table below:

[0079] Where R is the radius of the track's projection (which is a circle) in the horizontal plane, equal to the outer radius of the tower; Δh is the height difference of the track along the circumference: the height difference from the lowest point (windward side) to the highest point (leeward side); Δx is the small displacement required for the slider to start from the lowest point; This embodiment employs a purely passive drive system, requiring no external energy source. It utilizes wind power to drive the slider's movement, resulting in energy savings and high reliability. The slider position automatically changes with wind direction and speed, tracking wind load changes in real time without manual intervention, achieving adaptive adjustment. A reset mechanism ensures the slider returns to its initial position when there is no wind, avoiding the adverse effects of prolonged static load on the tower. Deep integration with the structural strength calculation system forms a closed-loop intelligent structural health management system encompassing perception, calculation, and execution. The main components are a ring track, slider, spring, and vanes, making it easy to manufacture and install. It is suitable for both new tower construction and retrofitting of existing towers, featuring a simple structure and low cost.

[0080] Example 4: The difference from Example 1 is that, as shown in Example 4... Figures 8 to 11 As shown, the flange connection of the FRP tower is equipped with a fatigue self-tightening flange sealing structure. Taking the pipe flange connection as an example: the first flange 401 is fixedly connected to the upper end of the vertical pipe at the top of the FRP tower 100, and the second flange 402 is fixedly connected to the lower end of the external pipe. The first flange 401 and the second flange 402 are connected by bolts, and the aforementioned fatigue self-tightening flange sealing structure is provided between them. Specifically, it includes an elastic gasket 403 disposed between the sealing surfaces of the first flange 401 and the second flange 402, a wedge-shaped sealing ring 404 disposed on the outside of the elastic gasket 403, a one-way drive mechanism 405 connected between the first flange 401 and the second flange 402, a limiting retaining ring 406 for limiting the maximum self-tightening stroke of the wedge-shaped sealing ring 404, and a displacement sensor 407 for monitoring the cumulative displacement of the wedge-shaped sealing ring 404 and outputting it to the structural strength calculation system. A retaining ring 406 is fixedly installed on the inner side of the flange (the side closest to the center of the tower body), located at the end of the inward movement path of the wedge-shaped sealing ring 404; a displacement sensor 407 is installed on the retaining ring 406 (either embedded or mounted). The retaining ring 406 is fixedly installed on the innermost side of the flange (the side closest to the center of the tower body), and its upper surface facing the wedge-shaped sealing ring 404 is higher than the upper surface of the elastic gasket 403. The inner wall of the inner ring of the wedge-shaped sealing ring 404 (the side facing the center of the tower body) is provided with a limiting protrusion. When the wedge-shaped sealing ring 404 moves inward to a set stroke, the limiting protrusion contacts the retaining ring 406, achieving the limitation of the maximum self-tightening stroke. The elastic gasket 403 is located between the retaining ring 406 and the wedge-shaped sealing ring 404. Its height is lower than the retaining surface of the retaining ring 406, ensuring that the wedge-shaped sealing ring 404 is not blocked by the elastic gasket 403 during movement. Only when the wedge-shaped sealing ring 404 moves into place will the retaining protrusion contact the retaining ring 406.

[0081] The first flange 401 and the second flange 402 have opposing sealing surfaces; the wedge-shaped sealing ring 404 has a wedge-shaped cross section, and the wedge-shaped surface forms an oblique angle contact with the flange sealing surface; the unidirectional drive mechanism 405 uses the relative displacement generated by the alternating load to drive the wedge-shaped sealing ring 404 to move unidirectionally inward, gradually pressing the elastic gasket 403.

[0082] The unidirectional drive mechanism 405 includes 2 to 4 drive units evenly distributed along the circumference of the flange. These drive units are located on the outer edge of the flange. Each drive unit includes an external threaded sleeve 4054 fixed to the second flange 402, an internal threaded ring 4055 fixedly connected to and meshing with the wedge-shaped sealing ring 404, and an inertial counterweight 4056 slidably mounted on the external threaded sleeve 4054. The inertial counterweight 4056 is cylindrical, made of high-density materials such as lead, tungsten alloy, or cast iron, and has a guide hole in its center, allowing it to slidably mount on the external threaded sleeve 4054. The diameter D of the inertial counterweight 4056 is 30mm to 80mm, its height H is 20mm to 60mm, and its mass is determined by calculation based on the required driving force.

[0083] When the alternating load causes vibration, the inertial weight 4056 moves up and down, driving the internal threaded ring 4055 to rotate in one direction through the one-way clutch, causing the wedge-shaped sealing ring 404 to move inward along the thread.

[0084] The working principle is as follows: During installation, the wedge-shaped sealing ring 404 is in its outermost position, the elastic gasket 403 is compressed to the initial sealing pressure, and the bolts are tightened according to the design preload. When the tower vibrates under wind load or pressure fluctuations, a small relative displacement occurs between the upper and lower flanges. The unidirectional drive mechanism 405 converts each displacement into a unidirectional step movement of the wedge-shaped sealing ring 404. Every time the wedge-shaped sealing ring 404 moves inward a small distance, its wedge-shaped slope squeezes the elastic gasket 403, further compressing the gasket and increasing the sealing pressure. As service time increases, fiberglass creep causes a decrease in bolt preload, and the sealing pressure tends to decrease. However, the self-tightening flange sealing structure continues to work, constantly compressing the elastic gasket 403, automatically compensating for the loss of preload, and keeping the sealing pressure within the required range. When the wedge-shaped sealing ring 404 moves to the position of the limit ring 406, the self-tightening stroke ends. The displacement sensor 407 is a laser displacement sensor or a magnetostrictive displacement sensor. Its probe faces the end face of the wedge-shaped sealing ring 404 and is used to measure the cumulative displacement of the wedge-shaped sealing ring 404 in real time. The displacement sensor 407 sends the cumulative displacement data to the structural strength calculation system, and the system issues a maintenance warning to prompt the replacement of the sealing component.

[0085] This embodiment employs a reverse thinking approach, using fatigue to combat creep. It utilizes the energy of alternating loads to drive self-tightening, automatically compensating for creep relaxation and fundamentally solving the sealing failure problem. It requires no external energy source or control, offering high reliability and suitability for unattended chemical towers. The self-tightening stroke is designed to match the tower's design life, significantly reducing the frequency of manual bolt tightening. Real-time monitoring of the self-tightening stroke and sealing status enables predictive maintenance and allows for deep integration with structural strength calculation systems. All components are integrated at the flange connection, without increasing the tower's external dimensions, making it suitable for new construction and renovation projects, and facilitating technical upgrades.

[0086] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for calculating the structural strength of a fiberglass tower, characterized in that, Includes the following steps: Step S1: Obtain real-time sensing data and static design parameters of the fiberglass tower. The real-time sensing data includes stress-strain data, vibration frequency data, and corrosion potential data. The static design parameters include geometric dimension parameters, layup parameters, and material property parameters. Step S2: Based on the real-time sensing data and static design parameters, construct a structural digital twin model that is synchronously mapped to the physical entity of the FRP tower. Step S3: Inject the real-time sensing data as boundary conditions into the digital twin model, and combine it with historical load data to generate a multi-condition simulation scenario including normal and extreme conditions. Step S4: Drive the digital twin model to perform structural strength evolution analysis under multiple working conditions, construct the load-stress mapping relationship based on the analysis results, and predict the structural failure boundary of the FRP tower. Step S5: Calculate the dynamic safety factor of the fiberglass tower based on the structural failure boundary and real-time sensor data, and feed back early warning information on the weak structural areas to the physical entity through the digital twin model.

2. The method for calculating the structural strength of a fiberglass tower according to claim 1, characterized in that, The specific steps for constructing the structural digital twin model in step S2 include: S21. Establish a three-dimensional solid model based on the geometric parameters of the fiberglass tower; S22. Perform finite element mesh generation on the three-dimensional solid model and define an anisotropic composite material constitutive model based on material property parameters; S23. Bind real-time sensing data as boundary constraints to the corresponding nodes of the 3D solid model to form a real-time data-driven dynamic digital twin.

3. The method for calculating the structural strength of a fiberglass tower according to claim 2, characterized in that, In step S2, when constructing the structural digital twin model, the finite element theory is used to construct the physical field simulation model, and a reduced-order surrogate model is constructed using a deep neural network, forming a hybrid driving architecture of physical information neural network.

4. The method for calculating the structural strength of a fiberglass tower according to claim 3, characterized in that, In step S4, when predicting the structural failure boundary, a generative adversarial network or variational autoencoder is used to augment the extreme operating condition data and expand the failure mode data under small sample conditions to improve the accuracy of structural failure boundary prediction.

5. The method for calculating the structural strength of a fiberglass tower according to claim 1, characterized in that, In step S4, when performing structural strength evolution analysis, a neural network proxy model is used to replace part of the physical field simulation calculation. The neural network proxy model is trained based on historical simulation data and is used to respond to structural strength prediction requests in real time.

6. The method for calculating the structural strength of a fiberglass tower according to claim 1, characterized in that, The specific steps for constructing the load-stress mapping relationship in step S4 include: S41. Collect load data and corresponding stress response data under multiple working conditions to construct a training dataset; S42. Based on the training dataset, a load-stress mapping proxy model is established using support vector regression or random forest algorithm; S43. Embed the load-stress mapping proxy model into the digital twin model to achieve rapid prediction of structural stress under arbitrary load input.

7. The method for calculating the structural strength of a fiberglass tower according to claim 1, characterized in that, Following step S5, the following is also included: S6. The calculated dynamic safety factor and early warning information of structurally weak areas are fed back to the physical entity of the FRP tower or the remote operation and maintenance platform through the data interface to realize closed-loop management of the tower's structural health.

8. The method for calculating the structural strength of a fiberglass tower according to claim 1, characterized in that, It also includes a model adaptive update step: when the deviation between the real-time sensing data and the simulation results of the digital twin model exceeds a preset threshold, a model parameter calibration mechanism is triggered, and the digital twin model is incrementally learned and iteratively optimized using the latest sensing data.

9. A structural strength calculation system for fiberglass towers, used to implement the method described in any one of claims 1-8, characterized in that, include: The data acquisition module is used to acquire real-time sensor data and static design parameters of the fiberglass tower. A digital twin model construction module, connected to the data acquisition module, is used to construct a structural digital twin model based on the real-time sensing data and static design parameters; The digital twin engine module, connected to the digital twin model construction module, includes a physical field simulation unit and an artificial intelligence analysis unit, used to drive the digital twin model to perform multi-condition structural strength evolution analysis; A safety assessment module, connected to the digital twin engine module, is used to calculate the dynamic safety factor and predict the structural failure boundary; The human-computer interaction module is used to visually display the status of the twin model and output safety warning information.

10. The structural strength calculation system according to claim 9, characterized in that, The digital twin engine module also includes: The physics simulation unit is used for high-fidelity structural simulation of FRP towers based on the finite element method. Reduced-order model unit, used to build simulation proxy models based on deep neural networks to accelerate structural response prediction; The data augmentation unit is used to augment extreme condition data using generative adversarial networks.

Images