Uneven settlement safety evaluation method for power transmission tower

Abstract

This application discloses a method for safety assessment of uneven settlement of transmission towers, relating to the field of power engineering monitoring technology. The method includes: acquiring multiple SAR images covering the target transmission line corridor within a preset time period; determining multiple transmission towers and their tilt conditions based on the SAR images and a target detection model; determining the limit state values ​​corresponding to each tilt condition based on the tilt conditions, a surrogate model, and a finite element model. The limit state values ​​are calculated based on the allowable stress and maximum component stress of the transmission tower, and the finite element model is established based on the mechanical performance state of the transmission tower components. Based on the limit state values, a safety assessment of the transmission tower's settlement is performed to obtain the assessment results. This application obtains the limit state values ​​of the transmission tower through a surrogate model and a finite element model, correlating the tilt conditions of the transmission tower with structural mechanics, thereby improving the accuracy of the transmission tower safety assessment.

Description

Technical Field

[0001] This application relates to the field of power engineering monitoring technology, and in particular to a method for safety assessment of uneven settlement of transmission towers. Background Technology

[0002] Transmission lines are a crucial infrastructure of the power system, and transmission towers, as key load-bearing components of overhead transmission lines, directly impact the stable operation of the power grid. In geologically sensitive areas such as mining subsidence areas, soft soil foundations, expansive soil areas, and loess collapsing areas, uneven ground settlement is one of the main causes of differential deformation in transmission tower foundations. Uneven settlement induces additional stress in the transmission tower components, particularly manifesting as overall tilting caused by differential settlement on both sides of the tower foundation. This alters the structural stress state, and in severe cases, can lead to instability of the main materials, joint failure, or even the collapse of the entire tower, causing widespread power outages.

[0003] Current methods for assessing the safety of transmission towers only output the tilt amount or line-of-sight deformation of a single tower. However, due to differences in tower structure and tilt conditions, the same tilt amount can have varying impacts on tower safety. Therefore, current methods that only output tilt amounts have low accuracy in assessing transmission tower safety. Summary of the Invention

[0004] The main objective of this application is to provide a method for assessing the safety of transmission towers due to uneven settlement, aiming to solve the technical problem of low accuracy in assessing the safety of transmission towers.

[0005] To achieve the above objectives, this application proposes a method for safety assessment of uneven settlement of transmission towers, the method comprising: Acquire multiple SAR images covering the target route corridor within a preset time period; Based on the SAR image and the preset target detection model, the tilt conditions of multiple transmission towers in the SAR image and each of the transmission towers are determined. Based on the tilting conditions, the preset proxy model, and the preset finite element model, the limit state values ​​corresponding to each tilting condition are determined. The limit state values ​​are calculated based on the allowable stress value and the maximum component stress of the transmission tower, and the finite element model is established based on the mechanical performance state of the transmission tower components. Based on the aforementioned limit state values, a safety assessment of the settlement of the transmission tower is conducted, and the assessment results are obtained.

[0006] In one embodiment, the step of determining the limit state value corresponding to each tilting condition based on the tilting condition, a preset proxy model, and a preset finite element model includes: Each of the tilting conditions is input into the surrogate model to obtain the predicted limit state value corresponding to each tilting condition; Based on the predicted limit state value, the structural failure probability of each of the transmission towers is calculated; From the structural failure probabilities, a preset number of maximum failure probabilities are determined, and the target tilting condition corresponding to each of the maximum failure probabilities is determined. Each of the target tilting conditions is input into the finite element model to obtain the limit state value corresponding to each target tilting condition.

[0007] In one embodiment, before the step of determining the limit state value corresponding to each tilting condition based on the tilting condition, the preset proxy model, and the preset finite element model, the method further includes: Based on each of the aforementioned tilting conditions, calculate the mean and standard deviation of the tilting conditions; Based on the mean, the standard deviation, and the finite element model, an initial training sample and an inclined working condition sample pool are generated. The initial surrogate model is trained based on the initial training samples to obtain the preset model to be trained. Based on the inclined working condition sample pool, the initial proxy model is trained, and it is determined whether the output result of the initial proxy model meets the preset convergence condition. If not satisfied, then the extreme tilting condition is determined from the tilting condition sample pool. Based on the finite element model and the extreme tilting condition, the initial training sample is adjusted, and the process of training the preset model to be trained based on the initial training sample to obtain the initial surrogate model is repeated until the output result of the initial surrogate model satisfies the convergence condition, thus obtaining the surrogate model.

[0008] In one embodiment, the step of training the initial surrogate model based on the inclined working condition sample pool and determining whether the output result of the initial surrogate model meets the preset convergence condition includes: Each tilted working condition sample in the tilted working condition sample pool is input into the initial proxy model to obtain the sample mean and sample variance corresponding to each tilted working condition sample. The sample mean and sample variance are determined based on the difference between the tilted working condition sample and each initial sample in the initial training sample. Based on the sample mean, the sample variance, and the preset learning function, calculate the sample confidence value corresponding to each of the tilted working condition samples; Determine the minimum sample confidence value from all the sample confidence values, and determine whether the minimum sample confidence value is higher than a preset sample confidence threshold; If the value is higher, then the output of the initial proxy model is determined to satisfy the convergence condition.

[0009] In one embodiment, the step of determining the extreme tilting condition from the tilting condition sample pool and adjusting the initial training samples based on the finite element model and the extreme tilting condition includes: Determine the tilt condition sample corresponding to the minimum sample confidence value, and use the tilt condition sample as the extreme tilt condition. The inclined working condition sample is input into the finite element model to obtain the sample limit state value corresponding to the inclined working condition sample; Based on the tilted working condition sample and the sample limit state value, a sample to be supplemented is generated; The sample to be supplemented is added to the initial training sample.

[0010] In one embodiment, the step of generating initial training samples and a sample pool for inclined load conditions based on the mean, the standard deviation, and the finite element model includes: Based on the mean and the standard deviation, a preset first number of initial tilt conditions are generated; The initial tilting conditions are input into the finite element model to obtain the corresponding initial limit state values ​​for each initial tilting condition. The initial training samples are generated based on the initial limit state values ​​and the initial tilt condition. Based on the mean and the standard deviation, a preset second number of the inclined working condition sample pool is generated, wherein the first number is lower than the second number.

[0011] In one embodiment, the step of conducting a safety assessment of the settlement of the transmission tower based on the ultimate limit state value and obtaining the assessment result includes: Obtain the total number of tilting conditions evaluated for each of the aforementioned transmission towers; Based on the extreme state value, determine the number of risky tilting conditions that make the extreme state value less than a preset safety threshold. Calculate the number of risk tilting conditions and the total number of tilting conditions to obtain the structural failure probability corresponding to each transmission tower. Based on the structural failure probability and the preset failure probability threshold, the risk level of each transmission tower is determined.

[0012] In one embodiment, the failure probability threshold includes a first failure probability threshold, a second failure probability threshold, and a third failure probability threshold, wherein the first failure probability threshold is less than the second failure probability threshold, the second failure probability threshold is less than the third failure probability threshold, and the step of determining the risk level of each transmission tower based on the structural failure probability and the preset failure probability threshold includes: If the probability of structural failure is less than the first failure probability threshold, the risk level is determined to be green, and no intervention is required. If the probability of structural failure is greater than the first failure probability threshold but less than the second failure probability threshold, then the risk level is determined to be yellow, and the transmission tower is inspected based on a preset inspection cycle. If the probability of structural failure is greater than the second failure probability threshold but less than the third failure probability threshold, then the risk level is determined to be orange. If the probability of structural failure is greater than the third failure probability threshold, then the risk level is determined to be red.

[0013] In one embodiment, after the step of determining the risk level of each transmission tower based on the structural failure probability and a preset failure probability threshold, the method further includes: If the risk level is orange or red, then obtain the number of downstream towers and the number of affected users of the corresponding transmission tower; Obtain the operation and maintenance strategy of the target line corridor, and determine the operation and maintenance weight coefficient based on the operation and maintenance strategy; Based on the maintenance weight coefficient, the number of downstream towers, and the number of affected users, calculate the line impact range index; Based on the line impact range index, the transmission towers corresponding to the orange and red levels are sorted to obtain a priority list of critical towers.

[0014] In one embodiment, each of the SAR images is a SAR image of the same target transmission line corridor at different times. The step of determining multiple transmission towers in the SAR images and the tilt condition of each transmission tower based on the SAR images and a preset target detection model includes: The SAR image is input into the target detection model to obtain the set of tower location coordinates for each transmission tower; Based on the set of tower coordinates, the time baseline and spatial baseline between the SAR images of the same transmission tower are determined. The time baseline includes the difference in imaging time between the SAR images, and the spatial baseline includes the geometric distance of the corresponding satellite orbit in the vertical line-of-sight direction when each SAR image is imaged. Based on the time baseline and the spatial baseline, each SAR image of the same transmission tower is paired to obtain an interferometric network pair, wherein the time baseline of each SAR image in the interferometric network pair is less than a preset time baseline threshold, and the spatial baseline is less than a preset spatial baseline threshold. Based on the aforementioned interference network pairs, the time series of surface deformation and settlement rate map of each transmission tower location are obtained; Based on the time series of surface deformation at the tower location and the settlement rate map, the differential settlement value of each transmission tower is determined; Based on the differential settlement value, the same first vertical displacement is selected for the foundation nodes on the same side of the transmission tower, and the same second vertical displacement is selected for the remaining foundation nodes on the same side that do not take the first vertical displacement. Based on the first vertical displacement and the second vertical displacement, the differential settlement tilt in the direction of the target line corridor and the transverse direction are determined respectively, and the tilt condition is obtained.

[0015] One or more technical solutions proposed in this application have at least the following technical effects: This application acquires multiple SAR images covering the target transmission line corridor within a preset time period. Based on the SAR images and a preset target detection model, it determines multiple transmission towers in the SAR images and the tilt conditions of each transmission tower. Based on the tilt conditions, a preset proxy model, and a preset finite element model, it determines the limit state values ​​corresponding to each tilt condition. The limit state values ​​are calculated based on the allowable stress value and maximum component stress of the transmission tower. The finite element model is established based on the mechanical performance state of the transmission tower components. Based on the limit state values, a safety assessment of the settlement of the transmission tower is performed, and the assessment results are obtained.

[0016] Compared to current methods that only output the tilt or line-of-sight deformation of a single tower, leading to low accuracy in transmission tower safety assessments, this application uses a surrogate model and a finite element model to obtain the limit state values ​​of the transmission tower. This correlates the tilt condition of the transmission tower with structural mechanics, improving the accuracy of the safety assessment. Specifically, the limit state values ​​are calculated based on the allowable stress and maximum component stress of the transmission tower, and the finite element model is established based on the mechanical performance state of the transmission tower components. Therefore, this application establishes a correlation between the tilt condition and the structural mechanics of the transmission tower, allowing the stress of the transmission tower to be considered during safety assessments, thereby improving the accuracy of the safety assessment. Attached Figure Description

[0017] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0018] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a flowchart illustrating an embodiment of the method for safety assessment of uneven settlement of transmission towers in this application. Figure 2 This is a schematic diagram of the overall scenario process provided in Embodiment 1 of the method for safety assessment of uneven settlement of transmission towers in this application; Figure 3 This is a schematic diagram of a deep learning target detection network identifying transmission towers, provided in Embodiment 1 of the method for safety assessment of uneven settlement of transmission towers in this application. Figure 4 This is a schematic diagram of the time-series processing for extracting the tower location settlement curve, provided in Example 1 of the method for safety assessment of uneven settlement of transmission towers in this application. Figure 5 This is a flowchart illustrating Embodiment 2 of the method for safety assessment of uneven settlement of transmission towers in this application. Figure 6 This is a schematic diagram of the active learning iterative process of the surrogate model provided in Embodiment 2 of the method for safety assessment of uneven settlement of transmission towers in this application; Figure 7 The finite element model and schematic diagram of the paired equal settlement tilting method provided in Example 2 of the method for safety assessment of uneven settlement of transmission towers in this application; Figure 8 This is a flowchart illustrating Embodiment 3 of the method for safety assessment of uneven settlement of transmission towers in this application. Figure 9 This is a schematic diagram illustrating the data acquisition consent process involved in the safety assessment method for uneven settlement of transmission towers in this application.

[0020] The purpose, features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0021] It should be understood that the specific embodiments described herein are merely illustrative of the technical solutions of this application and are not intended to limit this application.

[0022] To better understand the technical solution of this application, a detailed description will be provided below in conjunction with the accompanying drawings and specific implementation methods.

[0023] It should be noted that the executing entity in this embodiment can be a computing service device with data processing, network communication, and program execution functions, such as a tablet computer, personal computer, or mobile phone, or an electronic device capable of performing the above functions, such as a transmission tower uneven settlement safety assessment device. The following description uses a transmission tower uneven settlement safety assessment device as an example to illustrate this embodiment and the subsequent embodiments.

[0024] Transmission lines are a crucial infrastructure of the power system, and transmission towers, as key load-bearing components of overhead transmission lines, directly impact the stable operation of the power grid. In geologically sensitive areas such as mining subsidence areas, soft soil foundations, expansive soil areas, and loess collapsing areas, uneven ground settlement is one of the main causes of differential deformation in transmission tower foundations. Uneven settlement induces additional stress in the transmission tower components, particularly manifesting as overall tilting caused by differential settlement on both sides of the tower foundation. This alters the structural stress state, and in severe cases, can lead to instability of the main materials, joint failure, or even the collapse of the entire tower, causing widespread power outages.

[0025] Current methods for assessing the safety of transmission towers only output the tilt amount or line-of-sight deformation of a single tower. However, due to differences in tower structure and tilt conditions, the same tilt amount can have varying impacts on tower safety. Therefore, current methods that only output tilt amounts have low accuracy in assessing transmission tower safety.

[0026] Furthermore, traditional manual inspection methods are inefficient and have limited coverage, making it difficult to conduct regular and comprehensive inspections of the numerous transmission towers within a large transmission line corridor. Moreover, existing methods are mostly static, single-item assessments; when a large number of risky towers exist simultaneously in an area, existing methods cannot prioritize their handling, hindering the rational allocation of maintenance resources.

[0027] Based on this, embodiments of this application provide a method for safety assessment of uneven settlement of transmission towers, referring to... Figure 1 , Figure 1 This is a flowchart illustrating the first embodiment of the method for safety assessment of uneven settlement of transmission towers in this application.

[0028] In this embodiment, the method for assessing the safety of uneven settlement of transmission towers includes steps S10 to S40: Step S10: Acquire multiple SAR images covering the target route corridor within a preset time period; It should be noted that the preset time period refers to a continuous time interval pre-set according to the assessment requirements, used to define the start and end time range of the acquired SAR (Synthetic Aperture Radar) images. The target line corridor refers to a specific overhead transmission line for which a safety assessment of uneven settlement of transmission towers is required, and the strip-shaped area within a certain width on both sides. Multiple SAR images refer to the number of two or more SAR images acquired within the same preset time period but on different imaging dates. SAR images are geospatial data of surface radar backscatter intensity or phase information acquired through active microwave remote sensing. The overall process of this embodiment, starting from acquiring multiple SAR images covering the target line corridor within the preset time period, can be referred to... Figure 2 .

[0029] It is understood that this embodiment acquires multiple SAR images covering the target route corridor within a preset time period. These images contain the phase and intensity information of surface radar echoes in the same area at different time points, thus providing the necessary multi-temporal remote sensing data source with temporal comparability for subsequent use of short baseline set temporal interferometry. This makes it possible to extract millimeter-level differential subsidence from these images, achieving high-frequency, low-cost deformation monitoring data accumulation for a large-scale route corridor that is difficult to cover by traditional manual measurements.

[0030] Step S20: Based on the SAR image and the preset target detection model, determine the multiple transmission towers in the SAR image and the tilt condition of each transmission tower; It should be noted that the pre-set target detection model refers to a deep learning network model that has been pre-trained and can automatically identify and locate specific targets in images, such as a single-stage or two-stage target detector based on a feature pyramid network. A transmission tower refers to a tower-shaped steel structure or lattice structure used to support conductors, ground wires, and auxiliary hardware in an overhead transmission line. The tilting condition refers to the structural stress state of a transmission tower when the tower tilts as a whole due to uneven foundation settlement, specifically characterized by the differential settlement on both sides of the tower base.

[0031] Understandably, this embodiment employs a pre-defined target detection model to process SAR images. This model can automatically extract and locate the visual features of transmission towers in the images, avoiding the enormous workload of manually labeling each tower. This enables efficient and automated identification of numerous transmission towers within a large-scale transmission corridor. Simultaneously, the temporal InSAR (Interferometric Synthetic Aperture Radar) method is used to extract differential settlement for each identified tower location to determine the tilt condition. The remote sensing deformation monitoring results are directly converted into the boundary condition input required for structural analysis, thereby establishing a quantitative correlation between surface settlement and tower stress state. This provides accurate and personalized operating condition data for subsequent finite element-based structural reliability assessment.

[0032] In one feasible implementation, each of the SAR images is a SAR image of the same target transmission line corridor at different times. The specific implementation of determining multiple transmission towers in the SAR images and the tilt condition of each transmission tower based on the SAR images and a preset target detection model can also be: The SAR image is input into the target detection model to obtain the set of tower location coordinates for each transmission tower; Based on the set of tower coordinates, the time baseline and spatial baseline between the SAR images of the same transmission tower are determined. The time baseline includes the difference in imaging time between the SAR images, and the spatial baseline includes the geometric distance of the corresponding satellite orbit in the vertical line-of-sight direction when each SAR image is imaged. Based on the time baseline and the spatial baseline, each SAR image of the same transmission tower is paired to obtain an interferometric network pair, wherein the time baseline of each SAR image in the interferometric network pair is less than a preset time baseline threshold, and the spatial baseline is less than a preset spatial baseline threshold. Based on the aforementioned interference network pairs, the time series of surface deformation and settlement rate map of each transmission tower location are obtained; Based on the time series of surface deformation at the tower location and the settlement rate map, the differential settlement value of each transmission tower is determined; Based on the differential settlement value, the same first vertical displacement is selected for the foundation nodes on the same side of the transmission tower, and the same second vertical displacement is selected for the remaining foundation nodes on the same side that do not take the first vertical displacement. Based on the first vertical displacement and the second vertical displacement, the differential settlement tilt in the direction of the target line corridor and the transverse direction are determined respectively, and the tilt condition is obtained.

[0033] It should be noted that the tower location coordinate set refers to the aggregated dataset of geographical coordinates of all transmission towers identified by the target detection model, with each coordinate corresponding to the location of a transmission tower. The temporal baseline refers to the difference in imaging dates between two SAR images, used to measure the time interval between the two images. The spatial baseline refers to the geometric distance between the satellite orbits corresponding to the two SAR images at the time of imaging, in the direction perpendicular to the radar line of sight, reflecting the spatial difference between the two imaging locations. The interferometric network pair refers to the interferogram combination formed by pairing multiple SAR images based on the temporal and spatial baseline selection criteria, where each pair satisfies both a temporal baseline and a spatial baseline less than a preset threshold.

[0034] The time series of surface deformation at a transmission tower site refers to the sequence data formed by arranging the cumulative deformation of the surface at a specific transmission tower site at different times relative to the initial time within a preset time period. The settlement rate map is a spatial distribution map reflecting the amount of surface settlement change per unit time; the linear settlement rate of each point can be obtained through time-series InSAR inversion. The differential settlement value refers to the difference in surface settlement between different foundation locations of the same transmission tower, used to quantify the degree of uneven foundation settlement.

[0035] The first vertical displacement refers to the same settlement displacement value applied to two foot nodes on the same side of the transmission tower. The second vertical displacement refers to the same settlement displacement value applied to two foot nodes on the other side of the transmission tower where the first vertical displacement was not applied. The difference between this value and the first vertical displacement characterizes the degree of tilt of the transmission tower. The differential settlement tilt refers to the overall tilt angle or displacement difference of the transmission tower caused by the difference between the first and second vertical displacements, calculated separately along the line direction and the transverse line direction of the target transmission line corridor.

[0036] It should also be noted that before the SAR image in this embodiment is input into the target detection model, it needs to undergo radiometric calibration, geocoding, and speckle noise filtering preprocessing. The target detection model in this embodiment uses a feature pyramid network combined with a rotating bounding box detection head to output the detection results of each transmission tower in the form of a rotating bounding box. After non-maximum suppression, the detection results are used to obtain the latitude and longitude coordinates of each tower location through geographic coordinate back projection, forming a set of tower location coordinates containing multiple transmission towers. The deep learning target detection network used in this embodiment to identify transmission towers can be referred to... Figure 3 The time-series processing for extracting the tower site settlement curve can be referenced. Figure 4 .

[0037] This implementation sets a time baseline threshold of 48 days and a spatial baseline threshold of 100m. An interferometric pair network is generated, and through steps such as removing the flattening effect, atmospheric phase estimation and removal, and time-series inversion, the time series of surface deformation and settlement rate maps for each tower location are obtained. Tower locations with settlement rates exceeding 5mm / year are marked with high-priority calculation indicators and given priority allocation of finite element analysis resources in subsequent steps.

[0038] Understandably, this implementation first automatically extracts the tower location coordinate set through a target detection model, avoiding manual annotation, thereby achieving rapid and fully automatic positioning of transmission towers within a large-scale transmission corridor. Then, based on dual thresholds of time and spatial baselines, SAR images are paired to generate interferometric network pairs. This filtering strategy eliminates image pairs with poor correlation, effectively suppressing decoherent noise and improving the accuracy and reliability of deformation extraction. By acquiring the time series of surface deformation and settlement rate maps at the tower locations and extracting differential settlement values, the InSAR monitoring results are quantified into specific displacement values. Finally, this embodiment transforms the differential settlement values ​​into a tilting modeling method where the same displacement is taken at the same side of the foundation and the two sides are paired with equal displacements. This modeling method not only conforms to the physical law of uneven settlement leading to overall tower tilting but also significantly simplifies the complexity of applying boundary conditions in finite element analysis. This achieves an efficient and reasonable connection from remote sensing monitoring to structural mechanics modeling, ensuring the accuracy and computational efficiency of subsequent failure probability calculations.

[0039] Furthermore, this implementation method replaces manual inspection and tower location labeling with deep learning target detection, enabling the identification of all tower locations covering a corridor of hundreds of kilometers within hours, significantly improving coverage efficiency. Employing time-series InSAR technology, differential settlement monitoring achieves millimeter-level accuracy, offering higher frequency and lower cost compared to traditional manual leveling measurements.

[0040] Step S30: Based on the tilting conditions, the preset proxy model, and the preset finite element model, determine the limit state values ​​corresponding to each tilting condition. The limit state values ​​are calculated based on the allowable stress value and the maximum component stress of the transmission tower, and the finite element model is established based on the mechanical performance state of the transmission tower components. It should be noted that the pre-defined surrogate model refers to a mathematical model pre-established using machine learning methods to approximate the calculation of limit state function values ​​using a high-fidelity finite element model. This model can predict the structural response under given input conditions with lower computational cost. The pre-defined finite element model refers to a numerical calculation model pre-established based on the Nonlinear Integrated Design and Analysis (NIDA) method to simulate the mechanical behavior of transmission tower structures. This model considers material nonlinearity and geometric nonlinearity and can calculate the stress response of each component under given boundary conditions.

[0041] The limit state value refers to the numerical value calculated using the limit state function, used to determine whether the transmission tower has experienced structural failure under a given settlement condition. Specifically, in this embodiment, a value less than 0 is considered a failure state. The allowable stress value refers to the maximum stress that the transmission tower component material is allowed to withstand, usually determined based on the material's yield strength or design specifications, and serves as a benchmark for structural safety assessment. The maximum component stress refers to the maximum stress value appearing in all components of the transmission tower under a specific settlement condition, calculated using a finite element model. The component mechanical performance state refers to the physical and mechanical characteristic parameters affecting the stress behavior of each component of the transmission tower, such as the material constitutive relationship (e.g., elastoplasticity), cross-sectional properties, and connection methods.

[0042] It should also be noted that the formula for the limit state function used to calculate the limit state value in this embodiment is: g(x) = σ_allow / σ_max(x) - 1 Where g(x) is the calculated limit state value, σ_allow is the allowable stress value, and σ_max(x) is the maximum component stress calculated by the finite element model under inclined conditions.

[0043] The finite element model established in this embodiment based on the mechanical performance state of the components is as follows: the model uses beam-column elements to discretize the main material (angle steel or round pipe) and diagonal bracing components of the transmission tower. The connection method between nodes is set to hinged or rigid connection according to the actual bolt connection form. The material constitutive model adopts a piecewise linear elastoplastic model, and the yield strength is taken as the standard value corresponding to Q355. Geometric nonlinearity is handled by updating the Lagrange formula, considering P-δ and P-Δ effects. Among them, the P-Δ effect refers to the second-order effect caused by the relative lateral displacement (inter-story displacement) at both ends of the overall structure or component, and the P-δ effect refers to the second-order effect caused by the bending deformation (deflection) of the component itself. In addition, in this embodiment, wind load and conductor tension under normal operating conditions are applied simultaneously during finite element calculation.

[0044] Understandably, this embodiment employs a pre-defined surrogate model to replace the direct invocation of a pre-defined finite element model for calculating limit state values. The surrogate model is trained using an active learning strategy with a small number of finite element calculation samples, achieving prediction accuracy close to that of the finite element model but at extremely low computational cost. This reduces the computational load required for extensive operational condition assessments by more than an order of magnitude while maintaining the accuracy of limit state value determination. Furthermore, the finite element model used in the training process of the surrogate model fully considers the mechanical performance state of the transmission tower components (including material nonlinearity and geometric nonlinearity), ensuring that the calculation of limit state values ​​accurately reflects the actual stress behavior of the transmission tower under large deformations. This achieves a balance between computational efficiency and structural response accuracy.

[0045] In one feasible implementation, the specific implementation of determining the limit state values ​​corresponding to each tilted working condition based on the tilted working condition, the preset proxy model, and the preset finite element model can also be: Each of the tilting conditions is input into the surrogate model to obtain the predicted limit state value corresponding to each tilting condition; Based on the predicted limit state value, the structural failure probability of each of the transmission towers is calculated; From the structural failure probabilities, a preset number of maximum failure probabilities are determined, and the target tilting condition corresponding to each of the maximum failure probabilities is determined. Each of the target tilting conditions is input into the finite element model to obtain the limit state value corresponding to each target tilting condition.

[0046] It should be noted that the predicted limit state value refers to the limit state function value directly calculated and output by the surrogate model after inputting the tilting condition into the preset model. This value is an approximate estimate of the calculation result from the finite element model. The highest failure probability refers to a preset number of probability values ​​selected from multiple transmission tower structural failure probabilities, sorted from largest to smallest. These probability values ​​correspond to the tower locations with the highest risk. The target tilting condition refers to the transmission tower tilting condition corresponding to the highest failure probability. This type of condition requires further accurate verification using a high-fidelity finite element model.

[0047] Understandably, this implementation uses a surrogate model with extremely low computational cost to make preliminary predictions and calculate the structural failure probability for all tilting conditions. Then, it selects only a predetermined number of target tilting conditions corresponding to the highest failure probability to call a high-fidelity finite element model for precise verification. Conditions that do not reach the highest failure probability threshold directly use the prediction results from the surrogate model. This approach retains the ability to accurately calculate high-risk tower locations while keeping the total number of finite element model calls within a predetermined limit. Compared to calling the finite element model for all conditions, this significantly reduces computational resource consumption and total computation time. Furthermore, since the selection criterion is the ranking of structural failure probabilities predicted by the surrogate model, this method can adaptively concentrate computational resources on truly high-risk tower locations, thus achieving an optimal balance between assessment accuracy and computational efficiency.

[0048] Step S40: Based on the extreme state value, conduct a safety assessment of the settlement of the transmission tower and obtain the assessment result.

[0049] It should be noted that safety assessment refers to the process of classifying and judging the structural safety status of transmission towers under current settlement conditions based on calculated limit state values ​​and according to preset judgment criteria. Assessment results refer to the conclusive information output by the safety assessment process, including the safety level of the transmission tower, its failure risk level, and a priority disposal list generated based on grid topology connectivity.

[0050] It should also be noted that the system in this embodiment automatically triggers a full-process update of the transmission tower uneven settlement safety assessment after the next phase of imagery is downloaded. The system can automatically trigger a full-process update with the arrival of new satellite imagery, achieving periodic dynamic refreshing of the regional safety level map without manual intervention.

[0051] In summary, this embodiment acquires multiple SAR images covering the target transmission line corridor within a preset time period. Based on the SAR images and a preset target detection model, it determines multiple transmission towers in the SAR images and the tilt conditions of each transmission tower. Based on the tilt conditions, a preset proxy model, and a preset finite element model, it determines the limit state values ​​corresponding to each tilt condition. The limit state values ​​are calculated based on the allowable stress value and maximum component stress of the transmission tower, and the finite element model is established based on the mechanical performance state of the transmission tower components. Based on the limit state values, a safety assessment of the settlement of the transmission tower is performed, and the assessment results are obtained.

[0052] Compared to current methods that only output the tilt or line-of-sight deformation of a single tower, leading to low accuracy in transmission tower safety assessments, this embodiment obtains the limit state values ​​of the transmission tower through a surrogate model and a finite element model. This correlates the tower's tilt condition with structural mechanics, improving the accuracy of safety assessments. Specifically, the limit state values ​​are calculated based on the tower's allowable stress and maximum component stress, and the finite element model is established based on the structural mechanical properties of the tower's components. Therefore, this embodiment establishes a correlation between the tilt condition and the transmission tower's structural mechanics, allowing the tower's stress to be considered during safety assessments, thus improving the accuracy of these assessments.

[0053] Based on the first embodiment of this application, in the second embodiment of this application, the content that is the same as or similar to that in Embodiment 1 above can be referred to the above description, and will not be repeated hereafter. Based on this, please refer to... Figure 5 Before step S30, the method for assessing the safety of uneven settlement of transmission towers further includes steps S01 to S05: Step S01: Calculate the mean and standard deviation of each tilting condition based on the tilting conditions. It is understandable that this embodiment calculates the mean and standard deviation based on the actual tilt conditions of each transmission tower. These statistical parameters are directly derived from the distribution characteristics of real monitoring data in the target area, thus providing random variable probability distribution parameters that conform to the actual engineering situation when sampling probabilistic operating conditions. At the same time, the mean and standard deviation, as core statistics describing the distribution of random variables, can guide the generation of a more representative Monte Carlo sample pool, resulting in higher fitting accuracy of the surrogate model near the limit state surface.

[0054] Step S02: Based on the mean, the standard deviation, and the finite element model, generate initial training samples and a sample pool for inclined working conditions; It should be noted that the initial training samples refer to the first batch of input-output data pairs used to construct the initial surrogate model. The input is the differential settlement tilt amount under the tilting condition, and the output is the limit state value calculated by the finite element model under this condition. The tilting condition sample pool refers to a large collection of candidate tilting conditions generated based on the mean and standard deviation using probabilistic sampling methods (such as Monte Carlo simulation), which is used for subsequent active learning iterations and failure probability estimation of the surrogate model.

[0055] Understandably, this step uses the Latin hypercube sampling method to generate initial training samples based on the obtained mean and standard deviation. This method can achieve uniform stratified sampling of samples in the probability space, thereby covering the main variation range of the tilted working condition with a smaller sample size, so that the initial surrogate model can obtain a good global approximation ability in the early stage of training.

[0056] Furthermore, this embodiment generates a large-scale sample pool of tilted load conditions. This sample pool provides a sufficient source of candidate points for the active learning strategy in the subsequent adaptive Monte Carlo simulation method, enabling the surrogate model to actively select the points with the greatest prediction uncertainty from the sample pool for iterative updates. Thus, while ensuring the accuracy of failure probability estimation, the number of calls to the finite element model is controlled to the order of magnitude of the initial training sample number plus the number of newly added samples during active learning, avoiding the huge computational overhead caused by directly performing finite element calculations on the entire sample pool.

[0057] In one feasible implementation, the specific implementation of generating the initial training samples and the inclined load case sample pool based on the mean, the standard deviation, and the finite element model can also be: Based on the mean and the standard deviation, a preset first number of initial tilt conditions are generated; The initial tilting conditions are input into the finite element model to obtain the corresponding initial limit state values ​​for each initial tilting condition. The initial training samples are generated based on the initial limit state values ​​and the initial tilt condition. Based on the mean and the standard deviation, a preset second number of the inclined working condition sample pool is generated, wherein the first number is lower than the second number.

[0058] It should be noted that the initial tilting condition refers to the first batch of tilting condition sample points generated based on the mean and standard deviation, used to construct the initial training samples; its number is determined by a preset first quantity. The initial limit state value refers to the corresponding limit state function value obtained through high-fidelity calculation after inputting the initial tilting condition into the finite element model. The preset first quantity refers to the pre-set number of samples used to generate the initial tilting condition; this number is relatively small and used to control the size of the initial training set. The preset second quantity refers to the pre-set number of samples used to generate the tilting condition sample pool; this number is much larger than the preset first quantity, used to ensure the statistical accuracy of the failure probability estimation.

[0059] It is understandable that the preset first number of samples used to generate the initial training samples in this implementation is much smaller than the preset second number used to generate the inclined working condition sample pool. Since the initial training samples require calling the computationally expensive finite element model one by one, the inclined working condition sample pool only serves as a candidate set and does not call the finite element model temporarily. Therefore, while ensuring that the initial surrogate model has a basic fitting capability, the number of high-cost finite element calculations in the initial stage is strictly controlled within the preset first number. At the same time, the existence of a large-scale inclined working condition sample pool provides a sufficiently rich source of candidate points for subsequent active learning strategies, enabling the surrogate model to select a small number of sample points from the massive samples during the iteration process that are most helpful in improving the fitting accuracy of the limit state surface for finite element verification. This achieves a reasonable balance between the overall computational cost (number of finite element calls) and the accuracy of failure probability estimation.

[0060] Step S03: Train the preset model to be trained based on the initial training samples to obtain the initial surrogate model; It should be noted that the preset training model refers to a machine learning model whose architecture has been selected before training begins but has not yet been fitted to the data. The initial surrogate model refers to the model obtained after preliminary training of the preset training model using initial training samples. This model can predict the limit state value and its prediction variance for any input point in the inclined working condition sample pool, but has not yet undergone active learning iterative optimization.

[0061] Understandably, this embodiment trains the model using initial training samples obtained from finite element model calculations. These samples cover the main variation range of the inclined working condition and the output values ​​have high fidelity, enabling the initial surrogate model to reasonably approximate the true limit state function in terms of global trends. Furthermore, the initial surrogate model in this embodiment not only outputs predicted values ​​but also prediction variance (i.e., uncertainty measure), providing a key evaluation metric for the subsequent active learning strategy. This allows subsequent iterations to proactively select the most valuable sample points for finite element verification based on prediction uncertainty, thereby significantly reducing the total number of calls to the high-cost finite element model while maintaining the accuracy of the surrogate model.

[0062] Step S04: Based on the inclined working condition sample pool, train the initial proxy model and determine whether the output result of the initial proxy model meets the preset convergence condition. It should be noted that the tilting condition sample pool refers to a large collection of candidate tilting conditions generated using a probability sampling method based on the mean and standard deviation, used for subsequent active learning iteration and failure probability estimation of the surrogate model. The output result refers to the predicted limit state value and its corresponding prediction uncertainty metric obtained by the initial surrogate model after predicting each sample point in the tilting condition sample pool. The preset convergence condition refers to a pre-set criterion for judging whether the surrogate model is sufficiently accurate. Typically, it is based on the fact that the number of sample points near the limit state surface and whose classification uncertainty exceeds a threshold in the surrogate model's prediction results for all sample pool points is less than a preset value; in this embodiment, it is the minimum value of the U learning function, U≥2.

[0063] In one feasible implementation, the initial surrogate model is trained based on the inclined working condition sample pool. A further implementation method for determining whether the output of the initial surrogate model satisfies the preset convergence condition is: Each tilted working condition sample in the tilted working condition sample pool is input into the initial proxy model to obtain the sample mean and sample variance corresponding to each tilted working condition sample. The sample mean and sample variance are determined based on the difference between the tilted working condition sample and each initial sample in the initial training sample. Based on the sample mean, the sample variance, and the preset learning function, calculate the sample confidence value corresponding to each of the tilted working condition samples; Determine the minimum sample confidence value from all the sample confidence values, and determine whether the minimum sample confidence value is higher than a preset sample confidence threshold; If the value is higher, then the output of the initial proxy model is determined to satisfy the convergence condition.

[0064] It should be noted that the sample mean refers to the expected estimate of the predicted limit state value output by the surrogate model after inputting the inclined load sample. The sample variance refers to the uncertainty measure of the predicted limit state value output by the surrogate model after inputting the inclined load sample, reflecting the dispersion of the predicted value around its mean. The learning function refers to the evaluation function used to quantify the uncertainty of the surrogate model's prediction for a certain sample point, such as the U learning function; the smaller its value, the greater the classification uncertainty at that sample point.

[0065] The sample confidence value is a numerical value calculated through the learning function, used to measure the confidence level of the surrogate model in the prediction results of the current sample point. The sample confidence threshold is a pre-set threshold value used to judge whether the confidence level of the surrogate model's prediction is acceptable. When the minimum sample confidence value is higher than this threshold, the surrogate model's predictions for all sample points are considered sufficiently reliable.

[0066] It is understood that this implementation calculates the sample confidence value based on the sample mean and sample variance, and selects the minimum sample confidence value to compare with the preset sample confidence threshold. The minimum sample confidence value reflects the confidence level of the surrogate model in predicting the most uncertain point in the entire sample pool, thereby ensuring that convergence is only determined when the model reaches the preset confidence level in the most difficult region to judge. This avoids the risk of distorted failure probability estimation caused by premature termination of training due to ignoring local prediction errors.

[0067] Step S05: If not satisfied, determine the extreme tilting condition from the tilting condition sample pool, adjust the initial training sample based on the finite element model and the extreme tilting condition, return to the step of training the preset model to be trained based on the initial training sample to obtain the initial surrogate model, until the output result of the initial surrogate model meets the convergence condition, and obtain the surrogate model.

[0068] It should be noted that the extreme tilting condition refers to the tilting condition sample selected from the tilting condition sample pool during the active learning iteration process that maximizes the uncertainty of the surrogate model's prediction. It usually corresponds to the sample point with the minimum learning function value.

[0069] In one feasible implementation, the specific implementation of determining the extreme tilting condition from the tilting condition sample pool, and adjusting the initial training samples based on the finite element model and the extreme tilting condition, can also be: Determine the tilt condition sample corresponding to the minimum sample confidence value, and use the tilt condition sample as the extreme tilt condition. The inclined working condition sample is input into the finite element model to obtain the sample limit state value corresponding to the inclined working condition sample; Based on the tilted working condition sample and the sample limit state value, a sample to be supplemented is generated; The sample to be supplemented is added to the initial training sample.

[0070] It should be noted that the sample limit state value refers to the true limit state function value obtained by high-fidelity nonlinear finite element analysis after inputting the limit tilting condition into the finite element model. The supplementary samples refer to input-output data pairs consisting of the limit tilting condition and its corresponding sample limit state value, used to add to the initial training samples to update the training set. The active learning iteration process of the surrogate model in this implementation can be referred to... Figure 6 The horizontal axis represents the number of iterations, and the vertical axis represents the uncertainty of the surrogate model prediction. The finite element model and the paired equal settlement tilt application method can be found in [reference needed]. Figure 7 .

[0071] It is understood that in this implementation, the tilted working condition sample corresponding to the minimum sample confidence value is directly determined as the extreme tilted working condition. This sample point is the point where the current surrogate model is the least reliable in the entire sample pool. It is located near the extreme state surface or in a region with a large prediction variance. Therefore, using this point as a new sample can minimize the prediction uncertainty of the surrogate model in the key region.

[0072] Furthermore, this embodiment calculates the true sample limit state value by inputting the extreme tilting condition into the finite element model, ensuring high fidelity of the output of the newly added training samples and avoiding error accumulation caused by training the surrogate model with approximations. Adding the generated supplementary samples to the initial training samples allows the surrogate model to more accurately fit the function behavior near the limit state surface in subsequent iterations, thereby maximizing the accuracy of the surrogate model with minimal cost per finite element calculation and significantly accelerating the convergence process.

[0073] In one embodiment, the step of determining the limiting tilt condition from the tilt condition sample pool further includes: For each sample in the inclined working condition sample pool, calculate its U learning function value as an indicator of the prediction uncertainty of that sample. Based on the differential settlement values ​​actually obtained from the historical time-series InSAR monitoring of the transmission tower, the probability density function of the settlement at the tower location is fitted, and the probability density value of each sample in the tilting condition sample pool under the probability density function is calculated as an indicator of the probability of the condition occurring for that sample. Based on the prediction uncertainty index and the probability of the operating condition occurring for each sample, the comprehensive sampling score of that sample is calculated. The inclined working condition sample with the highest comprehensive sampling score is selected as the ultimate inclined working condition, which is used to call the finite element model to calculate its true limit state value.

[0074] It should be noted that the U-learning function is an evaluation function used in the surrogate model to quantify the uncertainty of the predicted classification. The smaller its value, the less reliable the surrogate model's judgment of the sign of the limit state function at that sample point. The probability density function is a statistical distribution function fitted based on the historical differential settlement data actually monitored for the transmission tower, used to describe the relative probability of different settlement values ​​occurring under natural conditions. The probability density value corresponding to each sample point in the tilt condition sample pool reflects the probability of this settlement pattern actually occurring in the actual engineering environment. The comprehensive sampling score is a composite index obtained by multiplying the prediction uncertainty index and the probability index of the occurrence of the working condition, used to balance the knowledge exploration needs of the surrogate model with the probability of actual engineering risks.

[0075] It is understandable that in this embodiment, when selecting extreme tilting conditions, the prediction uncertainty index is multiplied by the actual probability index as a comprehensive sampling score. In contrast, traditional active learning only samples based on the uncertainty index alone. This step introduces a probability density function fitted based on time-series InSAR measured data, so that even if the prediction uncertainty of a sample is very high, if the settlement mode is almost impossible to occur in the real environment, it will not be selected first, thus avoiding wasting valuable finite element computing resources on physically unrealistic extreme conditions.

[0076] Furthermore, the samples with the highest comprehensive sampling scores are usually those settlement patterns that are difficult for the surrogate model to judge and have a high probability of occurrence in actual monitoring. These samples are located near the limit state surface and belong to common engineering conditions. Prioritizing their learning can maximize the accuracy of failure probability estimation with the fewest finite element calls, making the training of the surrogate model more focused on the risk areas of actual engineering concern.

[0077] In one embodiment, the step of fitting a probability density function of the tower location settlement based on the differential settlement values ​​actually obtained from historical time-series InSAR monitoring of the transmission tower, and calculating the probability density value of each sample in the tilting condition sample pool under the probability density function, further includes: Extract all historical differential settlement observations of the transmission tower along the line direction and the transverse line direction within a preset time period to form historical settlement sequences in two directions; For each direction of the historical subsidence sequence, the probability density function is nonparametrically fitted using the kernel density estimation method, where the bandwidth of the kernel density estimation is determined by cross-validation. Based on the geological type of the area where the transmission tower is located, the tail amplification factor corresponding to the type is queried from the preset geological correction factor table. The probability density function obtained by kernel density estimation is multiplied by the tail amplification factor on the side with the larger absolute value of settlement to obtain the corrected probability density function. For each sample in the inclined working condition sample pool, the difference in settlement along the track direction and across the track direction is substituted into the corrected joint probability density function to calculate the probability index of the working condition occurrence for that sample.

[0078] It should be noted that kernel density estimation is a nonparametric statistical method that does not presuppose that the data follows a specific distribution. Instead, it estimates the probability density function based on the historical data points themselves by superimposing kernel functions. This method can capture complex distribution characteristics such as multimodal and skewed distributions that may exist in actual monitoring data. The geological correction coefficient table is a pre-constructed coefficient mapping table for different geological types (such as mined-out areas, soft soil areas, and loess collapsible areas). It is used to amplify the value of the settlement probability density function at the tail end (areas with large settlement) to reflect the differences in the probability of extreme settlement events under different geological conditions.

[0079] The tail amplification factor, a value greater than one determined based on the geological type, is multiplied on the side of the probability density function representing large settlement. This reasonably increases the probability of large settlement events occurring in actual geologically sensitive areas, avoiding underestimation of probability due to insufficient samples of extreme events in historical data. The joint probability density function is the probability density function that two random variables, one along the track direction and the other across the track direction, both obeyed. In this embodiment, it is assumed that the two directions are independent. The joint density is obtained by multiplying the two marginal probability densities, which is used to simultaneously consider the combined probability of differential settlement in both directions.

[0080] It is understandable that this embodiment uses kernel density estimation instead of a preset normal distribution to fit the probability density of historical settlement data. However, the actual data of uneven settlement of transmission towers often exhibit skewed or multi-peak characteristics. For example, settlement in goaf areas may be concentrated in two modes: small settlement and sudden large settlement. Kernel density estimation can capture these complex distribution patterns unbiasedly, thereby making the probability index of the occurrence of the working condition more realistically reflect the actual situation on site and avoiding the distortion of probability estimation caused by incorrect distribution assumptions.

[0081] Furthermore, since different geological types (mining subsidence areas, soft soil areas, etc.) have a significant impact on the probability of extreme settlement events, and extreme event samples are scarce in historical monitoring data, directly using kernel density estimation will underestimate the tail probability. By correcting through geological experience coefficients, this embodiment can reasonably increase the sampling weight of high-risk working conditions when data is limited, thereby significantly improving the targeting of active learning sampling to actual engineering risks.

[0082] In summary, this embodiment utilizes actual monitoring data to calculate statistical parameters and generate stratified initial training samples and a large-scale sample pool, enabling the initial surrogate model to reflect the true probability distribution characteristics of the tilting condition immediately after training. Subsequently, a convergence criterion based on sample mean, sample variance, and the learning function is employed. This criterion directly quantifies the global prediction confidence of the surrogate model near the limit state surface, thus avoiding the blindness of manually setting the number of iterations. When the convergence condition is not met, the limit tilting condition corresponding to the minimum sample confidence value is actively selected for finite element verification and the training set is updated. This strategy concentrates finite element calls on the most uncertain region of the current model, thereby achieving a significant improvement in the accuracy of the surrogate model with the fewest iterations.

[0083] Based on the first and second embodiments of this application, in the third embodiment of this application, the content that is the same as or similar to that in embodiments one and two above can be referred to the above description, and will not be repeated hereafter. Based on this, please refer to... Figure 8 Step S40, the method for assessing the safety of uneven settlement of transmission towers further includes steps S41 to S44: Step S41: Obtain the total number of tilt conditions evaluated for each of the transmission towers; It should be noted that the total number of tilting conditions refers to the sample size of all possible tilting conditions considered when conducting a structural safety assessment of each transmission tower.

[0084] Step S42: Based on the extreme state value, determine the number of risky tilting conditions that make the extreme state value less than a preset safety threshold. It should be noted that the preset safety threshold refers to a pre-set limit state value used to determine whether a tilting condition leads to structural failure. It is typically set to 0; structural failure is determined when the limit state value is less than this threshold. The number of risky tilting conditions refers to the total number of tilting conditions assessed that satisfy the condition that the limit state value is less than the preset safety threshold, i.e., the number of condition samples leading to structural failure.

[0085] It is understood that this embodiment performs binary classification judgment on the limit state value of each tilting condition based on a preset safety threshold, so that the number of risky tilting conditions obtained directly reflects the frequency of structural failure under the current probability sampling. This number, combined with the total number of tilting conditions obtained, can calculate the structural failure probability, thereby transforming the stress ratio result calculated by finite element method into a failure probability index with engineering decision-making value, and providing a quantitative basis for subsequent safety level classification.

[0086] Step S43: Calculate the number of risk tilting conditions and the total number of tilting conditions to obtain the structural failure probability corresponding to each transmission tower; It should be noted that the structural failure probability refers to the proportion of all possible tilting conditions that could lead to the failure of the transmission tower structure. Specifically, it is the ratio of the number of risky tilting conditions to the total number of tilting conditions, and is used to quantify the possibility of the transmission tower failing under the current settlement risk.

[0087] It is understandable that this embodiment uses the ratio of the number of risky tilting conditions to the total number of tilting conditions to define the structural failure probability, thereby ensuring the reliability and stability of the assessment results. Furthermore, this probability value integrates the extreme state judgment results of all possible tilting conditions, comprehensively reflecting the overall risk level of the transmission tower under uncertain settlement, and avoiding the shortcomings of overly conservative or overlooked risks that may result from relying solely on a single condition or the most unfavorable condition for evaluation.

[0088] Step S44: Based on the structural failure probability and the preset failure probability threshold, determine the risk level of each transmission tower.

[0089] It should be noted that the preset failure probability threshold refers to the pre-set failure probability boundary value used to classify the risk level of transmission towers. It usually includes multiple graded thresholds (such as 0.1%, 1%, 5%), which map the continuous structural failure probability to discrete risk levels.

[0090] In one feasible implementation, the failure probability threshold includes a first failure probability threshold, a second failure probability threshold, and a third failure probability threshold, wherein the first failure probability threshold is less than the second failure probability threshold, the second failure probability threshold is less than the third failure probability threshold, and the specific implementation of determining the risk level of each transmission tower based on the structural failure probability and the preset failure probability threshold can also be: If the probability of structural failure is less than the first failure probability threshold, the risk level is determined to be green, and no intervention is required. If the probability of structural failure is greater than the first failure probability threshold but less than the second failure probability threshold, then the risk level is determined to be yellow, and the transmission tower is inspected based on a preset inspection cycle. If the probability of structural failure is greater than the second failure probability threshold but less than the third failure probability threshold, then the risk level is determined to be orange. If the probability of structural failure is greater than the third failure probability threshold, then the risk level is determined to be red.

[0091] It should be noted that the first failure probability threshold is a preset minimum failure probability boundary value used to distinguish between green and yellow levels, and is typically set to 0.1%. The second failure probability threshold is a preset intermediate failure probability boundary value used to distinguish between yellow and orange levels, and is typically set to 1%. The third failure probability threshold is a preset maximum failure probability boundary value used to distinguish between orange and red levels, and is typically set to 5%.

[0092] Understandably, this implementation method uses three progressive failure probability thresholds (first, second, and third) to divide the structural failure probability into four risk levels (green, yellow, orange, and red). Each level corresponds to a differentiated handling strategy (no intervention required, inspection according to the inspection cycle, early warning, and danger). This ensures that the assessment results not only reflect the safety status of the structure but also directly guide the priority and urgency of subsequent operation and maintenance actions. This enables the reasonable allocation of limited operation and maintenance resources to high-risk tower locations and improves the efficiency of power grid safety management.

[0093] In one feasible implementation, the specific implementation method after determining the risk level of each transmission tower based on the structural failure probability and a preset failure probability threshold can also be: If the risk level is orange or red, then obtain the number of downstream towers and the number of affected users of the corresponding transmission tower; Obtain the operation and maintenance strategy of the target line corridor, and determine the operation and maintenance weight coefficient based on the operation and maintenance strategy; Based on the maintenance weight coefficient, the number of downstream towers, and the number of affected users, calculate the line impact range index; Based on the line impact range index, the transmission towers corresponding to the orange and red levels are sorted to obtain a priority list of critical towers.

[0094] It should be noted that the number of downstream towers refers to the number of subsequent towers that cannot obtain power from the power source, as determined by grid topology connectivity analysis after the failure of the current transmission tower. The number of affected users refers to the scale of electricity users affected by power outages after the failure of the current transmission tower, usually quantified by the number of households or equivalent load. The operation and maintenance strategy refers to the guiding rules formulated by the grid operation and maintenance management unit regarding resource allocation, inspection priority, emergency response, etc., which includes the setting of weighting coefficients for calculating the line impact range index.

[0095] The operation and maintenance weighting coefficient refers to the coefficient used to weight the calculation of the line impact range index, determined according to the operation and maintenance strategy. It typically includes the weight of the number of downstream towers (α) and the weight of the number of affected users (β), and satisfies α + β = 1. The line impact range index is a quantitative indicator that comprehensively assesses the impact of transmission tower failure on grid connectivity and power supply capacity. It is calculated by weighting the operation and maintenance weighting coefficient, the number of downstream towers, and the number of affected users. The critical tower priority disposal list refers to a priority list of transmission towers with risk levels of orange or red, sorted in descending order of the line impact range index, used to guide the allocation order of operation and maintenance resources.

[0096] It should also be noted that the specific calculation formula for the line influence range index in this embodiment is as follows: I = α × N_tower + β × N_user Where I is the line impact range index, α is the weight of the number of downstream towers and β is the weight of the number of affected users, N_tower is the number of downstream towers and N_user is the scale of affected power supply users.

[0097] The specific operation and maintenance strategies in this embodiment may include: grid reliability, prioritizing remote transmission corridors and emphasizing grid integrity; user power supply, prioritizing important load areas (hospitals, airports, government, etc.); balancing strategies, used for general transmission corridors; verification strategies, used for routine configurations that meet power system criteria; important power supply tasks, used for temporary adjustments during major events; and minimizing economic losses.

[0098] It is understandable that this implementation method only calculates the line impact range index for high-risk tower locations with a risk level of orange or red, avoiding unnecessary topology analysis calculations for all tower locations. This reduces computational overhead while ensuring the comprehensiveness of the decision-making process. At the same time, the calculation formula for the line impact range index includes both the number of downstream towers and the number of affected users, and adjusts their relative importance through an operation and maintenance weight coefficient. This allows the final ranking result to flexibly adapt to decision preferences under different scenarios based on the operation and maintenance strategy, thereby achieving a comprehensive evaluation of power grid system risks.

[0099] Furthermore, the final output of this embodiment presents the priority list of critical towers in descending order, transforming the abstract risk level into an action sequence that can directly guide on-site work. This enables maintenance personnel to prioritize the towers that have the greatest impact on the power grid under limited resources, significantly improving the practicality of the regional transmission tower uneven settlement safety assessment results.

[0100] In one embodiment, the step of sorting the transmission towers corresponding to the orange and red levels based on the line impact range index to obtain a priority list of critical towers further includes: The time-series InSAR settlement rate data of the transmission tower in the most recent period are obtained. The settlement rate change curve is fitted by the least squares method. The second derivative of the current settlement rate is calculated. If the second derivative is greater than zero, the settlement is determined to be in an accelerated state and the acceleration factor is marked as the first acceleration factor. Otherwise, the acceleration factor is marked as the second acceleration factor. Based on the tower type design documents of this tower location, its redundancy coefficient is extracted to determine the ratio of the ultimate bearing capacity of the tower under standard design conditions to the normal operating load. The dynamic priority index is obtained by multiplying the line impact range index by the acceleration factor and then dividing by the redundancy coefficient. The line impact range index is calculated by adding the weight of the number of downstream towers and the number of downstream towers to the weight of the number of affected users and the number of affected users. Sort all orange and red tower positions in descending order of dynamic priority index, and output a list of critical towers to be prioritized.

[0101] It should be noted that the second derivative of the settlement rate refers to the value obtained by taking the second derivative of the settlement rate change curve acquired by time-series InSAR, and is used to determine whether the settlement is accelerating, constant, or decelerating. The acceleration factor is a weighted coefficient assigned based on the settlement acceleration state, used to amplify the treatment priority of towers in a deteriorating trend. The redundancy coefficient reflects the dimensionless parameter of the transmission tower structure's bearing capacity margin under standard design conditions, defined as the ratio of ultimate bearing capacity to normal operating load; a higher value indicates a stronger reserve capacity against settlement deformation. The dynamic comprehensive treatment priority index is a comprehensive score obtained by integrating the settlement dynamic trend and the structure's own redundancy on the basis of the original line influence range index, used to more accurately rank the treatment order of critical towers.

[0102] Understandably, this embodiment introduces the second derivative of the settlement rate to determine whether the settlement is accelerating. Accelerated settlement means that the differential settlement may quickly exceed the design expectation in the short term, which makes the towers with accelerated development receive higher priority for treatment, avoiding the delay in intervention for towers with deteriorating trends that may be caused by sorting by static index.

[0103] Furthermore, this embodiment extracts the redundancy coefficient of each tower location as the denominator. This coefficient reflects the inherent ability of the structure to resist settlement under different tower types and design standards. This parameter is usually ignored in the safety assessment of transmission towers and is assumed to be the same for all towers. This embodiment couples it with dynamic settlement characteristics, so that tower locations with smaller bearing capacity margins are automatically given higher priority, achieving focused attention on vulnerable towers. Moreover, by using a dynamic comprehensive treatment priority index, the settlement trend, structural redundancy, and impact on the power grid system are integrated into a single ranking index, thereby significantly improving the predictive ability of the key tower ranking results for the actual risk evolution and enabling the forward-looking allocation of power grid operation and maintenance resources.

[0104] In summary, this embodiment first transforms the limit state value into a statistically convergent structural failure probability. This probability integrates the failure judgment results of all possible operating conditions in a large-scale inclined operating condition sample pool, thereby avoiding the one-sidedness of evaluation based on a single operating condition or the most unfavorable operating condition, and providing a scientific and continuous quantitative basis for subsequent risk classification. Three progressive failure probability thresholds are used to divide the continuous probability into four discrete risk levels, each level corresponding to a clear handling recommendation. This transforms the assessment result from an abstract probability value into a decision label that can directly guide operation and maintenance actions.

[0105] For high-risk tower locations classified as orange and red, an impact range index is further calculated. This index considers both the number of downstream towers and the number of affected users, and the relative importance of the two is flexibly adjusted through an operation and maintenance weight coefficient. This achieves a comprehensive evaluation of the impact on the power grid system. The final output list of priority disposal for key towers is arranged in descending order of the impact range index, which allows operation and maintenance resources to be prioritized for tower locations that have the greatest impact on the power grid's power supply capacity after failure. This significantly improves the engineering practicality of the safety assessment of uneven settlement of regional transmission towers.

[0106] It should be noted that the above examples are only for understanding this application and do not constitute a limitation on the method for assessing the safety of uneven settlement of transmission towers in this application. Any simple modifications based on this technical concept are within the scope of protection of this application.

[0107] All user-related data involved in this application was obtained with the user's permission or consent, as per [reference]. Figure 9 In other words, when this application is applied to a specific product or technology, user permission is required to acquire and process the relevant data, and the processing of the relevant data must comply with the relevant laws, regulations and regulatory standards of the relevant countries and regions.

[0108] The above description is only a part of the embodiments of this application and does not limit the scope of protection of this application. All equivalent structural transformations made under the technical concept of this application and using the contents of the specification and drawings of this application, or direct / indirect applications in other related technical fields, are included in the patent protection scope of this application.

Claims

1. A method for safety evaluation of uneven settlement of a power transmission tower, characterized in that, The method includes: Acquire multiple SAR images covering the target route corridor within a preset time period; Based on the SAR image and the preset target detection model, the tilt conditions of multiple transmission towers in the SAR image and each of the transmission towers are determined. Based on the tilting conditions, the preset proxy model, and the preset finite element model, the limit state values ​​corresponding to each tilting condition are determined. The limit state values ​​are calculated based on the allowable stress value and the maximum component stress of the transmission tower, and the finite element model is established based on the mechanical performance state of the transmission tower components. Based on the aforementioned limit state values, a safety assessment of the settlement of the transmission tower is conducted, and the assessment results are obtained. The step of determining the limit state value corresponding to each tilting condition based on the tilting condition, the preset proxy model, and the preset finite element model includes: Each of the tilting conditions is input into the surrogate model to obtain the predicted limit state value corresponding to each tilting condition; Based on the predicted limit state value, the structural failure probability of each transmission tower is calculated, wherein the structural failure probability is the ratio of the number of risky tilting conditions to the total number of tilting conditions, and the number of risky tilting conditions is the total number of conditions among all tilting conditions that satisfy the limit state value being less than a preset safety threshold. From the structural failure probabilities, a preset number of maximum failure probabilities are determined, and the target tilting condition corresponding to each of the maximum failure probabilities is determined. Each of the target tilting conditions is input into the finite element model to obtain the limit state value corresponding to each target tilting condition; The formula for calculating the limit state function of the limit state value is as follows: g(x) = σ_allow / σ_max(x) - 1 Where g(x) is the calculated limit state value, σ_allow is the allowable stress value, and σ_max(x) is the maximum component stress calculated by the finite element model under inclined conditions.

2. The method of claim 1, wherein, Before the step of determining the limit state value corresponding to each tilted working condition based on the tilted working condition, the preset proxy model, and the preset finite element model, the method further includes: Based on each of the aforementioned tilting conditions, calculate the mean and standard deviation of the tilting conditions; Based on the mean, the standard deviation, and the finite element model, an initial training sample and an inclined working condition sample pool are generated. The initial surrogate model is trained based on the initial training samples to obtain the preset model to be trained. Based on the inclined working condition sample pool, the initial proxy model is trained, and it is determined whether the output result of the initial proxy model meets the preset convergence condition. If not satisfied, the extreme tilting condition is determined from the tilting condition sample pool. Based on the finite element model and the extreme tilting condition, the initial training sample is adjusted, and the process of training the preset model to be trained based on the initial training sample to obtain the initial surrogate model is repeated until the output result of the initial surrogate model satisfies the convergence condition, thus obtaining the surrogate model.

3. The method of claim 2, wherein, The steps of training the initial proxy model based on the tilted working condition sample pool and determining whether the output of the initial proxy model meets the preset convergence conditions include: Each tilted working condition sample in the tilted working condition sample pool is input into the initial proxy model to obtain the sample mean and sample variance corresponding to each tilted working condition sample. The sample mean and sample variance are determined based on the difference between the tilted working condition sample and each initial sample in the initial training sample. Based on the sample mean, the sample variance, and the preset learning function, calculate the sample confidence value corresponding to each of the tilted working condition samples; Determine the minimum sample confidence value from all the sample confidence values, and determine whether the minimum sample confidence value is higher than a preset sample confidence threshold; If the value is higher, then the output of the initial proxy model is determined to satisfy the convergence condition.

4. The method of claim 3, wherein, The step of determining the extreme tilting condition from the tilting condition sample pool, and adjusting the initial training samples based on the finite element model and the extreme tilting condition, includes: Determine the tilt condition sample corresponding to the minimum sample confidence value, and use the tilt condition sample as the extreme tilt condition. The inclined working condition sample is input into the finite element model to obtain the sample limit state value corresponding to the inclined working condition sample; Based on the tilted working condition sample and the sample limit state value, a sample to be supplemented is generated; The sample to be supplemented is added to the initial training sample.

5. The method of claim 2, wherein, The step of generating the initial training samples and the inclined load case sample pool based on the mean, the standard deviation, and the finite element model includes: Based on the mean and the standard deviation, a preset first number of initial tilt conditions are generated; The initial tilting conditions are input into the finite element model to obtain the corresponding initial limit state values ​​for each initial tilting condition. The initial training samples are generated based on the initial limit state values ​​and the initial tilt condition. Based on the mean and the standard deviation, a preset second number of the inclined working condition sample pool is generated, wherein the first number is lower than the second number.

6. The method of claim 1, wherein, The steps for conducting a safety assessment of the settlement of the transmission tower based on the ultimate limit state value and obtaining the assessment result include: Obtain the total number of tilting conditions evaluated for each of the aforementioned transmission towers; Based on the extreme state value, determine the number of risky tilting conditions that make the extreme state value less than a preset safety threshold. Calculate the number of risk tilting conditions and the total number of tilting conditions to obtain the structural failure probability corresponding to each transmission tower. Based on the structural failure probability and the preset failure probability threshold, the risk level of each transmission tower is determined.

7. The method of claim 6, wherein, The failure probability threshold includes a first failure probability threshold, a second failure probability threshold, and a third failure probability threshold. The first failure probability threshold is less than the second failure probability threshold, and the second failure probability threshold is less than the third failure probability threshold. The step of determining the risk level of each transmission tower based on the structural failure probability and the preset failure probability threshold includes: If the probability of structural failure is less than the first failure probability threshold, the risk level is determined to be green, and no intervention is required. If the probability of structural failure is greater than the first failure probability threshold but less than the second failure probability threshold, then the risk level is determined to be yellow, and the transmission tower is inspected based on a preset inspection cycle. If the probability of structural failure is greater than the second failure probability threshold but less than the third failure probability threshold, then the risk level is determined to be orange. If the probability of structural failure is greater than the third failure probability threshold, then the risk level is determined to be red.

8. The method of claim 7, wherein, After the step of determining the risk level of each transmission tower based on the structural failure probability and a preset failure probability threshold, the method further includes: If the risk level is orange or red, then obtain the number of downstream towers and the number of affected users of the corresponding transmission tower; Obtain the operation and maintenance strategy of the target line corridor, and determine the operation and maintenance weight coefficient based on the operation and maintenance strategy; Based on the maintenance weight coefficient, the number of downstream towers, and the number of affected users, calculate the line impact range index; Based on the line impact range index, the transmission towers corresponding to the orange and red levels are sorted to obtain a priority list of critical towers.

9. The method of claim 1, wherein, Each of the SAR images is a SAR image of the same target transmission line corridor at different times. The step of determining multiple transmission towers in the SAR images and the tilt condition of each transmission tower based on the SAR images and a preset target detection model includes: The SAR image is input into the target detection model to obtain the set of tower location coordinates for each transmission tower; Based on the set of tower coordinates, the time baseline and spatial baseline between the SAR images of the same transmission tower are determined. The time baseline includes the difference in imaging time between the SAR images, and the spatial baseline includes the geometric distance of the corresponding satellite orbit in the vertical line-of-sight direction when each SAR image is imaged. Based on the time baseline and the spatial baseline, each SAR image of the same transmission tower is paired to obtain an interferometric network pair, wherein the time baseline of each SAR image in the interferometric network pair is less than a preset time baseline threshold, and the spatial baseline is less than a preset spatial baseline threshold. Based on the aforementioned interference network pairs, the time series of surface deformation and settlement rate map of each transmission tower location are obtained; Based on the time series of surface deformation at the tower location and the settlement rate map, the differential settlement value of each transmission tower is determined; Based on the differential settlement value, the same first vertical displacement is selected for the foundation nodes on the same side of the transmission tower, and the same second vertical displacement is selected for the remaining foundation nodes on the same side that do not take the first vertical displacement. Based on the first vertical displacement and the second vertical displacement, the differential settlement tilt in the direction of the target line corridor and the transverse direction are determined respectively, and the tilt condition is obtained.

Images