Substation steel structure health state evaluation method fusing multi-source heterogeneous monitoring data

Abstract

The present application relates to the technical field of structural state perception, and particularly relates to a substation steel structure health state evaluation method fusing multi-source heterogeneous monitoring data, which solves the technical problem that the prior art cannot accurately perceive the health state of components in a sensor-free blind area under sparse sensor layout conditions. The method comprises the following steps: constructing an initial topological directed graph; obtaining monitoring data, extracting environmental thrust features and stress damage features of edges of laid sensors, and performing neighborhood estimation on stress damage features of blind area edges; taking the stress damage features as a penalty term, amplifying and updating the initial impedance to obtain a reconstructed topological directed graph; counting a first frequency proportion of each edge being passed through by a lowest impedance path in the initial topological directed graph and a second frequency proportion in the reconstructed topological directed graph; and according to a positive relative change of the second frequency proportion compared with the first frequency proportion, fusing the environmental thrust features to determine a health risk score of each edge.

Description

Technical Field

[0001] This invention relates to the field of structural condition sensing technology, specifically to a method for assessing the health status of substation steel structures by integrating multi-source heterogeneous monitoring data. Background Technology

[0002] The combined steel frame within a substation is the core supporting structure for high-voltage transmission lines and electrical equipment. It typically consists of a small number of main load-bearing steel columns and a large network of cross braces. In long-term outdoor service environments, the frame not only bears horizontal wind pressure loads from high altitudes but is also highly susceptible to uneven ground settlement. This can cause localized main load-bearing steel columns to sink or yield, forcing a redistribution of wind pressure transmission paths within the structure. The surrounding unmonitored cross braces then become backup force transmission channels, absorbing additional transferred loads.

[0003] To monitor the health of such steel structures, existing technologies typically utilize the connection relationships between members to construct a topological model and deploy multi-source sensors such as strain gauges and hydrostatic levels on some of the main load-bearing members. By collecting and analyzing monitoring data, the safety status of the structure can be determined.

[0004] However, due to limitations in engineering wiring and renovation costs, sensors cannot be installed on a large number of cross braces and secondary beams, creating monitoring blind spots. When a local main load-bearing steel column sinks or yields, the wind pressure transmission path within the structure is forced to redistribute, and the originally unloaded blind spot members may bear a large amount of transferred load and be on the verge of overload. Existing technologies rely on isolated data from a limited number of measuring points and cannot reflect changes in the load transmission path within the structural network. When members in the monitoring blind spot bear additional transferred loads due to load redistribution, the members with installed sensors may not have reached the alarm threshold, leading to missed detection of local hazards. To extrapolate the global stress state, conventional methods require the establishment of a three-dimensional finite element simulation model. However, finite element calculations are highly dependent on absolute mechanical parameters such as the aging elastic modulus of components and the stiffness of node connections. These parameters are difficult to obtain accurately on-site in existing projects, resulting in insufficient model accuracy, implementation difficulties, and poor overall evaluation results. Summary of the Invention

[0005] To address the technical problem that existing technologies cannot accurately detect the health status of components in sensorless blind zones under sparse sensor deployment conditions, the present invention aims to provide a method for assessing the health status of substation steel structures by integrating multi-source heterogeneous monitoring data. The specific technical solution adopted is as follows: Firstly, a method for assessing the health status of substation steel structures by integrating multi-source heterogeneous monitoring data is provided. This method includes: constructing an initial topological directed graph of the substation steel structure by using member connection nodes as vertices, members as edges, and the cross-sectional area of ​​the members to determine the initial impedance of each edge; acquiring monitoring data from multi-source sensors, extracting environmental thrust characteristics and force damage characteristics of edges with deployed sensors, and performing neighborhood estimation of the force damage characteristics of adjacent blind zone edges based on the force damage characteristics corresponding to the edges with deployed sensors; using the force damage characteristics of each edge as a penalty term to amplify and update the initial impedance in the initial topological directed graph to obtain a reconstructed topological directed graph; statistically analyzing the first frequency proportion of each edge traversed by the lowest impedance path in the initial topological directed graph, and the second frequency proportion of each edge traversed by the lowest impedance path in the reconstructed topological directed graph; and determining the health risk score of each edge by integrating environmental thrust characteristics based on the positive relative change between the second frequency proportion and the first frequency proportion.

[0006] Based on the above technical solution, the substation steel structure health status assessment method provided by this invention, which integrates multi-source heterogeneous monitoring data, abstracts the substation steel structure into a topological directed graph where the initial impedance is determined by the cross-sectional area of ​​the members. It then uses the stress and damage characteristics of members with deployed sensors to estimate the neighborhood of members in sensorless blind zones, enabling the perception of sensorless blind zone components without relying on absolute mechanical parameters. Furthermore, the initial impedance is amplified and updated using stress and damage characteristics as a penalty term, reconstructing a directed graph reflecting changes in load transmission capacity after structural damage. By comparing the frequency of each member being traversed by the lowest impedance path before and after reconstruction, the degree of positive load transfer due to obstructed load transmission in the main structure is extracted. Finally, environmental thrust characteristics are quantified to obtain the health risk score for each member. This solution can effectively perceive and quantitatively assess the overload risk caused by the transfer of loads in members within monitoring blind zones using limited measurement point data, significantly improving the accuracy of the overall health status assessment of substation steel structures.

[0007] In conjunction with the first aspect mentioned above, in one possible implementation, the monitoring data from the multi-source sensors includes: ambient temperature, strain data of the members on the members where sensors have been deployed, and settlement data of the members; the stress damage characteristics include: net strain data caused by external loads, and settlement difference data at different locations in the substation steel structure; the method for extracting the stress damage characteristics of the edges where sensors have been deployed specifically includes: statistically analyzing the temperature difference between the ambient temperature at the time to be evaluated and the reference installation temperature when the sensors were initially installed, and combining this with the linear expansion constant of the member material to determine the thermal expansion strain data caused by the change in ambient temperature; subtracting the thermal expansion strain data from the member strain data to obtain the net strain data; statistically analyzing the settlement difference between the member settlement data at adjacent column base positions in the substation steel structure, and combining this with the maximum allowable settlement value corresponding to the span between adjacent column base members to determine the settlement difference data.

[0008] In conjunction with the first aspect above, in one possible implementation, the method for neighborhood estimation of the force damage characteristics of adjacent blind zone edges based on the force damage characteristics corresponding to the edges of the deployed sensors specifically includes: querying the adjacent edges of the deployed sensors that share vertices with the blind zone edges in the initial topological directed graph; selecting the net strain data of the adjacent edge with the largest absolute value, and attenuating it as the net strain data of the blind zone edge; selecting the settlement difference data of the adjacent edge with the largest absolute value, and attenuating it as the settlement difference data of the blind zone edge.

[0009] In conjunction with the first aspect mentioned above, in one possible implementation, the method of using the stress damage characteristics of each edge as a penalty term to amplify and update the initial impedance in the initial topological directed graph to obtain a reconstructed topological directed graph specifically includes: determining a strain penalty factor based on the net strain data of each edge, determining a settlement penalty factor based on the settlement difference data of each edge, multiplying the initial impedance by the strain penalty factor and the settlement penalty factor to obtain the updated real-time impedance; and replacing the initial impedance with the real-time impedance to obtain the reconstructed topological directed graph.

[0010] In conjunction with the first aspect above, in one possible implementation, the method further includes: performing maximum impedance cutoff protection on the real-time impedance according to a preset upper limit value; if the real-time impedance is greater than the upper limit value, then the real-time impedance is cut off to the upper limit value.

[0011] In conjunction with the first aspect mentioned above, in one possible implementation, the monitoring data from the multi-source sensors includes: ambient wind speed; the method for extracting the environmental thrust features specifically includes: acquiring the ambient wind speed sequence within a preset time window before the time to be evaluated, smoothing the ambient wind speed sequence to obtain the equivalent wind speed of the overall wind force intensity within the preset time window; and using the ratio of the equivalent wind speed to the preset maximum design wind speed as the environmental thrust feature.

[0012] In conjunction with the first aspect above, in one possible implementation, the method for constructing the initial topological directed graph specifically includes: obtaining the three-dimensional spatial coordinates of the member connection nodes and the member connection relationships; mapping the three-dimensional spatial coordinates of the member connection nodes to vertices of the initial topological directed graph and mapping the member connection relationships to edges of the initial topological directed graph; determining the direction from the higher vertex to the lower vertex as the edge direction of the initial topological directed graph based on the elevation data of each vertex obtained from the three-dimensional spatial coordinate mapping; obtaining the cross-sectional area of ​​the member corresponding to each edge; calculating the average cross-sectional area of ​​the members in the substation steel structure; and using the ratio of the average cross-sectional area to the cross-sectional area of ​​each edge as the initial impedance of each edge in the initial topological directed graph.

[0013] In conjunction with the first aspect mentioned above, in one possible implementation, the method for calculating the first frequency percentage of each edge traversed by the lowest impedance path in the initial topological directed graph and the second frequency percentage of each edge traversed by the lowest impedance path in the reconstructed topological directed graph specifically includes: based on the elevation data of each vertex, defining the vertices of the windward surface region with elevations in a preset top interval as the starting set, and defining the vertices of the column base region with elevations in a preset bottom interval as the ending set; searching for the lowest impedance path in both the initial and reconstructed topological directed graphs, using any vertex in the starting set as the path search starting point and any vertex in the ending set as the path search ending point; and calculating the frequency of each edge traversed by the lowest impedance path in both the initial and reconstructed topological directed graphs, and calculating the proportion of the frequency to the number of lowest impedance paths to obtain the first frequency percentage and the second frequency percentage.

[0014] In conjunction with the first aspect mentioned above, in one possible implementation, the method for determining the health risk score of each edge based on the positive relative change of the second frequency proportion compared to the first frequency proportion, combined with environmental thrust characteristics, specifically includes: subtracting the first frequency proportion from the second frequency proportion of each edge, retaining only the part with a positive difference as a positive increment, and using the ratio of the positive increment to the first frequency proportion as a positive relative change characteristic; multiplying the environmental thrust characteristic, the positive relative change characteristic, and the initial impedance to obtain the health risk score of each edge.

[0015] In conjunction with the first aspect mentioned above, in one possible implementation, after determining the health risk score for each edge, the method further includes: if the maximum value of the health risk score is greater than the preset upper limit of the normal path transfer, then the preset alarm threshold of the sensor is reduced according to the degree to which the maximum value of the health risk score exceeds the upper limit of the normal path transfer; otherwise, the preset alarm threshold is maintained, thus obtaining the adjusted alarm threshold; if the monitoring data of the sensors deployed on the pole is greater than the adjusted alarm threshold, then the corresponding pole is determined to have a structural safety hazard, and the three-dimensional spatial coordinates of the pole are output.

[0016] In a second aspect, a substation steel structure health status assessment device that integrates multi-source heterogeneous monitoring data is provided, including a processor and a memory. The memory stores a computer program, and when the computer program is executed by the processor, it can perform the actions described in the first aspect or any possible implementation of the first aspect.

[0017] The present invention has the following beneficial effects: By abstracting the substation steel structure into a topological directed graph where the initial impedance is determined by the cross-sectional area of ​​the members, and utilizing the stress and damage characteristics of members with deployed sensors to perform neighborhood estimation for members in sensorless blind zones, this method enables the sensing of sensorless blind zone components without relying on absolute mechanical parameters. Based on this, the initial impedance is amplified and updated using stress and damage characteristics as a penalty term, reconstructing a directed graph reflecting changes in load transmission capacity after structural damage. By comparing the frequency of each member being traversed by the lowest impedance path before and after reconstruction, the degree of positive load transfer due to obstruction of the main load transmission path is extracted. Finally, environmental thrust characteristics are fused to quantify the health risk score of each member. This scheme can effectively sense and quantitatively assess the overload risk caused by the transfer of loads to members in monitoring blind zones using limited measurement point data, significantly improving the accuracy of the overall health status assessment of substation steel structures. Attached Figure Description

[0018] Figure 1 A flowchart of a method for assessing the health status of substation steel structures by integrating multi-source heterogeneous monitoring data, provided in one embodiment of the present invention; Figure 2 This is a schematic diagram of the hardware structure of a substation steel structure health status assessment device that integrates multi-source heterogeneous monitoring data, provided as an embodiment of the present invention. Detailed Implementation

[0019] The following description, in conjunction with the accompanying drawings, details the specific scheme of the substation steel structure health status assessment method provided by the present invention, which integrates multi-source heterogeneous monitoring data.

[0020] Please see Figure 1 The diagram illustrates a flowchart of a method for assessing the health status of substation steel structures by integrating multi-source heterogeneous monitoring data, according to an embodiment of the present invention. This method includes: S1. Construct the initial topological directed graph of the substation steel structure by using the connection nodes of the members as vertices, the members as edges, and the cross-sectional area of ​​the members to determine the initial impedance of each edge.

[0021] As a typical spatial truss structure composed of members and nodes, the mechanical essence of a substation's combined steel frame is the gradual transfer of wind loads from the upper windward surface to the lower column foundations through the topological connections between the members. This force transmission characteristic naturally aligns with the network connectivity model composed of vertices and edges. Therefore, physical member connection nodes can be abstracted as graph vertices, and members as graph edges, thus establishing a directed topological graph of the structure for path search and connectivity analysis. In the undamaged initial design state, the cross-sectional area of ​​a member directly determines its axial stiffness and load-bearing capacity: members with larger cross-sectional areas have smoother force transmission paths and bear a higher proportion in load distribution; conversely, slender secondary braces bear a lower share of force transmission. Based on this, the ratio of the global average cross-sectional area to the cross-sectional area of ​​each member is defined as the initial impedance of the corresponding edge of that member. This can characterize the inherent resistance of each edge to load transmission in the initial undamaged state in the form of a dimensionless relative value. The slenderer the member, the greater the initial impedance, and the lower its participation priority in the initial connectivity path.

[0022] In some implementations, the three-dimensional spatial coordinates of the member connection nodes and the member connection relationships are obtained, which can be extracted from the design drawings in specific implementations. Then, the three-dimensional spatial coordinates of the member connection nodes are mapped to the vertices of the initial topological directed graph, and the member connection relationships (each solid steel beam connecting two nodes) are mapped to the edges of the initial topological directed graph, generating a structural topology association graph that completely corresponds to the physical framework of the substation.

[0023] Since the overturning resistance of tall structures under horizontal wind loads is mainly transmitted from top to bottom along the height direction to the column foundation, setting the edge direction to point from high node to low node ensures that the flow direction of the graph model is consistent with the actual mechanical transmission path. Based on the elevation data of each vertex obtained from the three-dimensional spatial coordinate mapping, the direction from the high vertex to the low vertex is determined as the edge direction of the initial topological directed graph. In particular, when the elevation values ​​of the two vertices at the two ends of an edge are equal, the edge connecting the two vertices is set as a bidirectional directed edge.

[0024] The cross-sectional area of ​​the member corresponding to each edge is obtained. In practice, this can also be extracted from the design drawings. The larger the cross-sectional area of ​​the member, the higher its axial stiffness, and the greater its share of force transmission under the undamaged state. In the model, this should be represented by a smaller resistance to path search. Based on this, the average cross-sectional area of ​​all members in the substation steel structure is calculated. The ratio of the average cross-sectional area to the cross-sectional area of ​​each edge is used as the initial impedance of each edge in the initial topological directed graph, quantifying the degree of resistance to load transmission exhibited by each edge under the undamaged state. The larger the cross-sectional area of ​​the main load-bearing member, the smaller its initial impedance, and the more preferentially it is selected in the path search. The smaller the cross-sectional area of ​​the slender member, the larger its initial impedance value, and the lower its priority in the initial connected path allocation. This is consistent with the objective law that slender secondary members bear less force transmission tasks under the design conditions. After completing the initial impedance assignment of all edges, the system constructs the initial topological directed graph for subsequent dynamic comparison.

[0025] S2. Acquire monitoring data from multiple sources of sensors, extract environmental thrust characteristics and force damage characteristics of the edges of deployed sensors, and perform neighborhood estimation of the force damage characteristics of adjacent blind zone edges based on the force damage characteristics corresponding to the edges of deployed sensors.

[0026] The deterioration of the health status of substation steel structures during service is the result of the combined effects of external environmental thrust and internal structural damage. Therefore, assessing their condition requires obtaining both types of information simultaneously. Environmental thrust characteristics can quantify the degree to which the current wind load approaches the structural design wind resistance limit, while stress damage characteristics can characterize the degree of degradation in the force transmission capacity of monitored members. However, due to limitations in engineering wiring and renovation costs, a large number of cross braces and secondary beams cannot be equipped with sensors, creating monitoring blind spots. But in spatial pole structures, adjacent members achieve internal force transmission and deformation coordination through shared nodes. Damage to one member will inevitably cause a correlated change in the stress state of surrounding members at the node. Therefore, the stress damage characteristics of monitored members already contain some information about the stress state of their neighboring blind-spot members. Based on this physical continuity, by using known damage characteristics to extrapolate along the topological adjacency relationship to the edge of the blind spot, it is possible to assign stress damage characteristics to sensorless members that are coordinated with the measured deformation of the surrounding structures, thereby constructing a complete stress state.

[0027] Among them, multi-source sensors refer to different types of sensing devices that can sense environmental effects and structural responses respectively. Single-type monitoring data cannot simultaneously reflect changes in external thrust and the damage state of the structure itself. It is necessary to integrate multi-dimensional information such as wind speed, temperature, strain, and settlement in order to fully capture the internal and external factors that cause the redistribution of structural force transmission paths.

[0028] In some implementations, the monitoring data includes ambient wind speed collected by wind speed sensors. Extracting environmental thrust characteristics specifically involves: obtaining the ambient wind speed sequence within a preset time window prior to the time of evaluation. The duration of the preset time window is set based on wind load engineering experience, for example, 3 minutes, to cover a complete gust cycle. The ambient wind speed sequence is smoothed to eliminate instantaneous spikes caused by high-frequency gusts, obtaining the equivalent wind speed representing the overall wind intensity within the preset time window. Smoothing can be achieved by calculating the root mean square (RMS) value of the wind speed sequence, as wind pressure is proportional to the square of wind speed, and the RMS value better reflects the sustained energy level of the wind load during that period. Subsequently, the ratio of the equivalent wind speed to the preset maximum design wind speed is used as the environmental thrust characteristic. The larger the ratio, the closer the current wind force is to the structure's wind resistance design limit, and the higher the weight of its impact on the structural health status. The maximum design wind speed is the design extreme wind speed determined based on the meteorological conditions of the substation's location and the national building structure load code, for example, set to 35 m / s.

[0029] In some implementations, the monitoring data also includes ambient temperature collected by temperature sensors, strain data of the members collected by strain gauges on the members with deployed sensors, and settlement data of the members collected by a hydrostatic level. The stress-damage characteristics include: net strain data caused by external loads, and settlement difference data at different locations in the substation steel structure. Extracting the stress-damage characteristics of the edges with deployed sensors specifically includes: The temperature difference between the ambient temperature at the time of evaluation and the reference installation temperature of the sensor during initial installation (i.e., the temperature at which it was initially installed, debugged, and zeroed, for example, 20℃) is calculated, combined with the linear expansion constant of the rod material (obtained from the material property handbook, for example, the linear expansion constant of steel is 12×10⁻⁶). -6 The temperature difference (°C) is multiplied by the linear expansion constant to obtain the thermal expansion strain data caused by the change in ambient temperature. The thermal expansion strain data is then subtracted from the strain data of the rod collected by the strain gauge to remove the temperature effect and obtain the net strain data that truly reflects the degree of tensile and compressive deformation caused by external load.

[0030] The settlement difference (absolute value) between adjacent column bases of the substation steel structure was statistically analyzed using static leveling instruments. The absolute settlement of a single column base includes the uniform settlement component of the entire substation area. Uniform settlement does not generate significant additional internal forces within the structure and has little impact on the force transmission path of the members. However, the settlement difference between adjacent column bases directly reflects the degree of uneven settlement of the foundation. Uneven settlement can cause forced bending or tensile deformation of the members connecting the two column bases, resulting in additional stress in the member and its surrounding area, thereby weakening its ability to continue bearing the design load. The larger the settlement difference, the more severe the degradation of the local bearing stiffness in that area, and the more it should be penalized and amplified in the impedance model. By combining the maximum allowable settlement value corresponding to the span between adjacent column base members (based on the maximum allowable settlement extreme value given for this span in the building code, for example, taking 1 / 500 of the span as the allowable value), the settlement difference is calculated by ratio to the maximum allowable settlement extreme value to obtain dimensionless settlement difference data. The larger the data value, the more serious the uneven settlement of adjacent column bases, and the higher the risk of associated deformation in the corresponding member area.

[0031] Furthermore, the adjacent edges of all deployed sensors sharing vertices with the blind zone edge in the initial topological directed graph are queried. Considering that structural safety assessment should follow the conservative principle of choosing the highest possible value, the load transfer effect caused by the most severely damaged adjacent member is also the most significant. Based on this, it can be ensured that the blind zone edge is estimated according to the most unfavorable working condition. Based on this, the net strain data of the adjacent edge with the largest absolute value is selected, and after attenuation, it is used as the net strain data of the blind zone edge. Similarly, the settlement difference data of the adjacent edge with the largest absolute value is selected, and after attenuation, it is used as the settlement difference data of the blind zone edge. The setting of the attenuation coefficient reflects the decreasing law of damage when it is transmitted across nodes. For example, based on engineering experience, it is calibrated to 0.5, indicating that the damage degree of the blind zone member is half of the adjacent maximum damage.

[0032] If there are no adjacent edges of the blind zone with any shared vertices and deployed sensors, then the net strain data and settlement difference data of the blind zone edge are both set to zero, meaning it is considered that it has not been significantly affected by surrounding damage.

[0033] S3. Using the force damage characteristics of each edge as a penalty term, the initial impedance in the initial topological directed graph is amplified and updated to obtain the reconstructed topological directed graph.

[0034] Once a member in the steel frame of a substation yields or its connected foundation experiences uneven settlement, its ability to continue bearing and transmitting loads within the structural system will substantially decline. Mechanically, this manifests as a reduction in its axial stiffness and load transmission reliability, rendering it unsuitable as a priority load transmission channel. In the model, the impedance value of an edge is a quantitative representation of the degree to which it hinders load transmission; higher impedance means a lower priority for that channel in path search. Amplifying and updating the initial impedance using the stress damage characteristics as a penalty term essentially reproduces this physical law in the data model. This allows the path search algorithm to spontaneously reduce its dependence on these high-impedance edges in subsequent calculations, transferring many connected paths originally borne by the damaged member to other bypass members with relatively lower impedance. Reconstructing the topological directed graph, while maintaining the original vertex and edge connections, replaces the initial impedance of all edges in the network with the real-time impedance values ​​updated after damage penalty, thus constructing a network model that dynamically reflects the current load transmission capacity distribution of the structure.

[0035] In some implementations, the strain penalty factor is determined based on the net strain data of each edge. Specifically, the strain penalty factor is determined by adjusting the ratio of the absolute value of the net strain data to the preset yield strain of the steel used in the member, after adjusting for preset amplification weights. This characteristic represents the degree to which the member approaches the material's yield limit due to tensile or compressive deformation. The closer the deformation is to yield, the larger the value of this factor, and the more significant the amplification effect on impedance. The preset amplification weight of the strain penalty factor can be calibrated based on the mechanical properties of the steel. In structural health monitoring, it is generally considered that when the strain reaches 50% of the yield strain, the component has entered a significant damage stage. The impedance amplification target multiple of this critical state is used as the setting basis. For example, if the yield strain of the steel is set to 2000 microstrains (με), and it is desired that the strain penalty factor can amplify the impedance by approximately 2 times when the member reaches 50% yield strain (i.e., 1000με), then the amplification weight can be set to 2.0. In this case, the strain penalty factor, i.e., the impedance multiple, = 1 + 2.0 × (1000 / 2000) = 2. During the on-site installation and commissioning phase, the amplification weights were fine-tuned by comparing the system response of the structure under known normal conditions.

[0036] The settlement penalty factor is determined based on the settlement difference data of each side. Specifically, it is determined by adjusting the settlement difference data with a preset amplification weight. This factor characterizes the degree of bearing capacity reduction in the area where the member is located due to uneven foundation settlement; the greater the settlement difference, the larger the factor value. The preset amplification weight of the settlement penalty factor can be calibrated according to the warning values ​​in the foundation design code. When the settlement difference reaches the allowable limit value in the code, it means that the foundation is in a critical safety state, and a significant penalty should be applied to the impedance of the relevant members. For example, if the maximum allowable settlement difference between adjacent column bases is set to 20mm, and it is desired that the settlement penalty factor can amplify the impedance by approximately 3 times when the measured settlement difference reaches this allowable limit value (i.e., the settlement difference data is 1.0), then the amplification weight can be set to 2.0. In this case, the settlement penalty factor, i.e., the impedance multiple, = 1 + 2.0 × 1.0 = 3. During the on-site installation and commissioning phase, the amplification weight is fine-tuned by comparing the system response of the structure under known normal conditions.

[0037] Multiplying the initial impedance by the strain penalty factor and the settlement penalty factor yields the updated real-time impedance. A higher value indicates a stronger obstruction to load transfer by the member at the current moment, resulting in a lower probability of it being traversed during path searching. Replacing the initial impedance with the real-time impedance reconstructs the initial topological directed graph, yielding the reconstructed topological directed graph.

[0038] Preferably, if the field sensor encounters accidental electromagnetic interference or probe displacement, it may cause abnormal extreme values ​​in the extracted net strain data or settlement difference data, leading to an extremely large calculated real-time impedance value. This would directly sever the connectivity of that edge in the subsequent path search algorithm, causing global path search failure or result distortion. Therefore, to ensure the computational stability of the algorithm, a maximum impedance truncation protection is implemented on the real-time impedance based on a preset upper limit value. If the real-time impedance exceeds the upper limit value, it is truncated to the upper limit value, and the truncated real-time impedance is used as the final edge weight in the reconstructed topological directed graph. The upper limit value can be set based on a multiple of the initial impedance, for example, ten times the initial impedance, to achieve a balance between preventing connectivity deadlock and preserving the damage amplification effect.

[0039] S4. Calculate the proportion of the first frequency path traversed by the lowest impedance path in the initial topology directed graph for each edge, and the proportion of the second frequency path traversed by the lowest impedance path in the reconstructed topology directed graph for each edge. Based on the positive relative change of the proportion of the second frequency path compared to the proportion of the first frequency path, and by integrating environmental thrust characteristics, determine the health risk score for each edge.

[0040] In a non-destructive design state, the load transmission share of each member in the substation steel structure is allocated according to its cross-sectional stiffness. This initial allocation pattern is fully encoded in the first frequency proportion determined by the initial impedance. However, when some main load-bearing members are damaged, leading to a decline in load transmission capacity, the path search algorithm spontaneously seeks alternative low-impedance paths, causing a redistribution of the reconstructed second frequency proportion relative to the first frequency proportion. The difference between these two proportions essentially captures the topological redistribution process of load transfer paths within the structure. The positive relative change specifically identifies those members that passively bear additional transfer paths, which are precisely the overload risk signals that need to be captured. Based on this, environmental thrust characteristics are integrated to ensure that the same amount of path transfer receives a higher risk score under high wind speed conditions, because the same degree of topological redistribution will lead to more severe actual stress consequences under stronger external loads. Therefore, this method can integrate information from three independent dimensions—topological connectivity changes, external environmental loads, and the geometric defense capabilities of the members themselves—into a dimensionless health risk score, enabling quantitative ranking of the relative overload risk of all members in the network without the need for direct measurement of strain in blind zone members.

[0041] In some implementations, based on the elevation data of each vertex, the vertices of the windward surface region with elevations in a preset top interval are designated as the starting set. The elevation interval can be proportionally defined according to the total height of the structure; for example, the top interval might be the elevation range of the first 10% of the total height. This set represents the main stress-bearing node group of the wind load input structure. The vertices of the column base region with elevations in a preset bottom interval are designated as the ending set; for example, the bottom interval might be the elevation range of the last 10% of the total height. This set represents the constraint node group that ultimately transfers the load to the foundation. In both the initial and reconstructed topological directed graphs, using any vertex in the starting set as the path search start point and any vertex in the ending set as the path search end point, a preset network path optimization algorithm searches for the connected path with the minimum total impedance between each pair of nodes, i.e., the lowest impedance path. The network path optimization algorithm can employ the classic Dijkstra algorithm or the Floyd-Warshall algorithm.

[0042] In both the initial and reconstructed directed topology graphs, the frequency of each edge traversed by the lowest impedance path is counted, and the proportion of each frequency to the total number of lowest impedance paths is also counted, yielding the first frequency proportion and the second frequency proportion. The first frequency proportion characterizes the baseline participation of each member in the overall force transmission network under the undamaged design state. Main load-bearing members with larger cross-sections typically have a higher first frequency proportion due to their lower initial impedance, while secondary braces with smaller cross-sections have low or even zero participation in the initial state. Because the impedance of some members is penalized and amplified by damage, the path search algorithm actively avoids these areas of increased impedance when searching for the lowest impedance path, causing a change in the distribution of the path set. The second frequency proportion reflects the actual connectivity participation of each member in the current damaged state of the structure.

[0043] In some implementations, the second frequency percentage of each edge is subtracted from the first frequency percentage, and only the positive difference is retained as a positive increment. This indicates that after structural damage, the edge has taken over additional connectivity paths transferred from other damaged areas, meaning the member passively assumes additional force transmission tasks, which is the core signal of the member's overload risk. The ratio of the positive increment to the first frequency percentage is used as a positive relative change feature. The larger the value, the more drastically the transfer path taken over by the edge after structural damage increases compared to its initial participation level, and the higher its passive overload risk. A non-positive increment indicates that the edge's participation level remains unchanged or decreases, usually because its own impedance is amplified by penalty and bypassed by the algorithm. This does not belong to the dangerous mode that requires warning and is therefore not included in the scoring calculation. To ensure that the calculation with the first frequency percentage as the denominator does not trigger a division-by-zero anomaly due to the initial zero value of idle members, a preset very small positive constant, such as 0.0001, is superimposed on the denominator when calculating the ratio of the positive increment to the first frequency percentage.

[0044] The health risk score for each edge is obtained by multiplying the environmental thrust characteristics, positive relative change characteristics, and initial impedance. A larger environmental thrust characteristic indicates a more intense external wind load and a stronger driving force for path shifting. A larger positive relative change characteristic indicates a more dramatic increase in the edge's connectivity importance within the network topology. A larger initial impedance indicates a thinner cross-section of the corresponding member, weaker inherent defense capabilities, and a greater susceptibility to damage under the same transferred load. These three factors, representing external thrust, network connectivity increment, and intrinsic geometric defense, collectively determine the overall risk of an edge. Multiplying these factors ensures that any deterioration in any of these dimensions is prominently reflected in the score. After calculating the scores for all edges in the network, they are sorted in descending order, and the edge corresponding to the highest score is extracted as the most vulnerable link in the structure at the current moment.

[0045] Furthermore, after determining the health risk score for each edge, dynamic early warning judgment is performed. A normal path transfer upper limit is preset. This upper limit is used to define the degree of path borrowing that the structure can exhibit under normal environmental fluctuations. It can be set by continuously calculating the envelope extreme value of the global maximum score under abnormally low wind speed conditions during the initial operation of the substation. For example, it can be set as 1.05 to 1.15 times the maximum score under historical normal operating conditions, or it can be directly calibrated by engineering experts as an empirical constant. Historical normal operating conditions refer to the continuous operating period during the initial operation of the substation where the structure is normal and the ambient wind speed is 50% lower than the design wind speed.

[0046] If the maximum value of the health risk score exceeds the preset upper limit of the normal path transfer, it indicates that a path redistribution exceeding the normal tolerance range has occurred within the structure, and the overall topology defense of the system has degraded. In this case, based on the degree to which the maximum health risk score exceeds the upper limit of the normal path transfer, the preset alarm threshold of the sensor is reduced. Specifically, the ratio of the upper limit of the normal path transfer to the global maximum score value is calculated, and this ratio is used as a deviation factor multiplied by the preset alarm threshold to obtain the dynamically tightened alarm threshold. The preset alarm threshold is imported from building structure design and monitoring specifications. For example, the strain safety cutoff ratio corresponding to the allowable stress of the member material is taken as the fixed alarm red line. The alarm threshold after deviation factor compression is lower than the original fixed red line, forming a more stringent internal software defense. Furthermore, a lower limit protection is applied to the dynamically tightened alarm threshold. The larger of the dynamically tightened alarm threshold and the preset minimum alarm threshold is taken as the final alarm threshold to avoid false alarms caused by excessively low thresholds. The preset minimum alarm threshold can be set based on the safety lower limit of the allowable strain of the member material or engineering experience, for example, 10% of the preset alarm threshold.

[0047] If the maximum value of the health risk score is less than or equal to the upper limit of the normal path transfer, it indicates that the current load transmission path distribution of the structure is within the normal range. The preset alarm threshold is maintained, and the adjusted alarm threshold is obtained.

[0048] During the monitoring process, if the monitoring data (strain data or settlement difference data of the pole) of the sensors installed on the pole is greater than the adjusted alarm threshold, it is determined that there is a structural safety hazard in the corresponding pole where the sensor is installed or in the nearby blind zone pole. The three-dimensional spatial coordinates of the pole where the sensor is installed are output to guide the on-site maintenance personnel to locate the surrounding area for investigation.

[0049] Furthermore, for blind-spot members without sensor installation, if the corresponding health risk score exceeds a preset risk score threshold, the blind-spot member is determined to have an overload safety hazard caused by load transfer. The three-dimensional spatial coordinates of the blind-spot member are then output to guide on-site maintenance personnel in quickly locating and inspecting specific members. The risk score threshold can be calibrated based on the envelope value of the structural design bearing capacity and historical normal operating condition scores, for example, by taking 1.2 times the maximum health risk score of all members under historical normal operating conditions.

[0050] Based on the above technical solution, by abstracting the substation steel structure into a topological directed graph where the initial impedance is determined by the cross-sectional area of ​​the members, and utilizing the stress and damage characteristics of members with deployed sensors to perform neighborhood estimation for members in sensorless blind zones, it is possible to sense components in sensorless blind zones without relying on absolute mechanical parameters. On this basis, the initial impedance is amplified and updated using stress and damage characteristics as a penalty term, reconstructing a directed graph reflecting changes in load transmission capacity after structural damage. Then, by comparing the frequency percentage of each member traversed by the lowest impedance path before and after reconstruction, the degree of positive load transfer due to obstruction of the main load transmission path is extracted. Finally, environmental thrust characteristics are fused to quantify the health risk score of each member. This solution can effectively sense and quantitatively assess the overload risk caused by the transfer of loads to members in monitoring blind zones using limited measurement point data, significantly improving the accuracy of the overall health status assessment of substation steel structures.

[0051] This invention also provides a hardware structure diagram of a substation steel structure health status assessment device that integrates multi-source heterogeneous monitoring data, see [link / reference]. Figure 2 The substation steel structure health status assessment device 200, which integrates multi-source heterogeneous monitoring data, includes a processor 201 and optionally a memory 202 connected to the processor 201.

[0052] In one possible implementation, see Figure 2 The substation steel structure health status assessment device 200, which integrates multi-source heterogeneous monitoring data, also includes a transceiver 203. The processor 201, memory 202, and transceiver 203 are connected via a bus. The transceiver 203 is used to communicate with other devices or communication networks. Optionally, the transceiver 203 may include a transmitter and a receiver. The device in the transceiver 203 that performs the receiving function can be considered as a receiver, which is used to execute the receiving steps in the embodiments of the present invention. The device in the transceiver 203 that performs the transmitting function can be considered as a transmitter, which is used to execute the transmitting steps in the embodiments of the present invention.

Claims

1. A method for assessing the health status of substation steel structures by integrating multi-source heterogeneous monitoring data, characterized in that, include: Using the member connection nodes as vertices, the members as edges, and the cross-sectional area of ​​the members as the initial impedance of each edge, an initial topological directed graph of the substation steel structure is constructed. Acquire monitoring data from multiple sources of sensors, extract environmental thrust characteristics and force damage characteristics of the edges of deployed sensors, and perform neighborhood estimation of the force damage characteristics of adjacent blind zone edges based on the force damage characteristics corresponding to the edges of deployed sensors. Using the force damage characteristics of each edge as a penalty term, the initial impedance in the initial topological directed graph is amplified and updated to obtain the reconstructed topological directed graph. The percentage of the first frequency of each edge being traversed by the lowest impedance path in the initial topology directed graph and the percentage of the second frequency of each edge being traversed by the lowest impedance path in the reconstructed topology directed graph are statistically analyzed. Based on the positive relative change of the second frequency percentage compared to the first frequency percentage, the environmental thrust characteristics are integrated to determine the health risk score of each edge.

2. The method for assessing the health status of substation steel structures by integrating multi-source heterogeneous monitoring data according to claim 1, characterized in that, The monitoring data from the multi-source sensors include: ambient temperature, strain data of the poles on which sensors have been deployed, and settlement data of the poles; the stress damage characteristics include: net strain data caused by external loads, and settlement difference data at different locations in the substation steel structure. Extract the force damage characteristics of the edges where sensors have been deployed, including: The temperature difference between the ambient temperature at the time of evaluation and the reference installation temperature when the sensor was initially installed is statistically analyzed. Combined with the linear expansion constant of the rod material, the thermal expansion strain data caused by the change in ambient temperature are determined. The net strain data is obtained by subtracting the thermal expansion strain data from the strain data of the rod. The settlement difference between the settlement data of adjacent column base positions of the substation steel structure is statistically analyzed, and combined with the maximum allowable settlement value corresponding to the span between adjacent column base members, the settlement difference data is determined.

3. The method for assessing the health status of substation steel structures by integrating multi-source heterogeneous monitoring data according to claim 2, characterized in that, Based on the force damage characteristics corresponding to the edges of the deployed sensors, the force damage characteristics of adjacent blind zone edges are estimated in the neighborhood, including: Query the adjacent edges of the deployed sensors that share vertices with the blind zone edges in the initial topological directed graph; The net strain data of the adjacent edge with the largest absolute value is selected, and after attenuation, it is used as the net strain data of the blind zone edge. The settlement difference data of the adjacent side with the largest absolute value is selected and attenuated to serve as the settlement difference data of the blind zone side.

4. The method for assessing the health status of substation steel structures by integrating multi-source heterogeneous monitoring data according to claim 3, characterized in that, Using the force-damage characteristics of each edge as a penalty term, the initial impedance in the initial topological directed graph is amplified and updated to obtain a reconstructed topological directed graph, including: The strain penalty factor is determined based on the net strain data of each edge, and the settlement penalty factor is determined based on the settlement difference data of each edge. The initial impedance is multiplied by the strain penalty factor and the settlement penalty factor to obtain the updated real-time impedance. The initial impedance is replaced by the real-time impedance to obtain the reconstructed topological directed graph.

5. The method for assessing the health status of substation steel structures by integrating multi-source heterogeneous monitoring data according to claim 4, characterized in that, The method further includes: The real-time impedance is protected by maximum impedance cutoff based on a preset upper limit value. If the real-time impedance is greater than the upper limit value, the real-time impedance is cut off to the upper limit value.

6. The method for assessing the health status of substation steel structures by integrating multi-source heterogeneous monitoring data according to claim 1, characterized in that, The monitoring data from the multi-source sensors includes: ambient wind speed; Extracting environmental thrust characteristics, including: Obtain the environmental wind speed sequence within a preset time window before the time to be evaluated, and smooth the environmental wind speed sequence to obtain the equivalent wind speed of the overall wind intensity within the preset time window. The ratio of the equivalent wind speed to the preset maximum design wind speed is used as the environmental thrust characteristic.

7. The method for assessing the health status of substation steel structures by integrating multi-source heterogeneous monitoring data according to claim 1, characterized in that, Constructing the initial topological directed graph includes: Obtain the three-dimensional spatial coordinates of the link connection nodes and the link connection relationships, map the three-dimensional spatial coordinates of the link connection nodes to the vertices of the initial topological directed graph, and map the link connection relationships to the edges of the initial topological directed graph; Based on the elevation data of each vertex obtained by the three-dimensional spatial coordinate mapping, the direction from the higher vertex to the lower vertex is determined as the edge direction of the initial topological directed graph; Obtain the cross-sectional area of ​​the member corresponding to each edge, calculate the average cross-sectional area of ​​the members in the steel structure of the substation, and use the ratio of the average cross-sectional area to the cross-sectional area of ​​each edge as the initial impedance of each edge in the initial topological directed graph.

8. The method for assessing the health status of substation steel structures by integrating multi-source heterogeneous monitoring data according to claim 7, characterized in that, The percentage of the first frequency of each edge being traversed by the lowest impedance path in the initial topological directed graph, and the percentage of the second frequency of each edge being traversed by the lowest impedance path in the reconstructed topological directed graph, are calculated, including: Based on the elevation data of each vertex, the vertices of the windward area with elevation in the preset top interval are designated as the starting set, and the vertices of the column base area with elevation in the preset bottom interval are designated as the ending set. In both the initial topological directed graph and the reconstructed topological directed graph, the path with the lowest impedance is searched using any vertex in the starting point set as the starting point and any vertex in the ending point set as the ending point. In both the initial topological directed graph and the reconstructed topological directed graph, the frequency of each edge being traversed by the lowest impedance path is counted, and the proportion of the frequency to the number of lowest impedance paths is counted to obtain the first frequency proportion and the second frequency proportion.

9. The method for assessing the health status of substation steel structures by integrating multi-source heterogeneous monitoring data according to claim 8, characterized in that, Based on the positive relative change of the second frequency proportion compared to the first frequency proportion, and integrating the environmental thrust characteristics, a health risk score is determined for each edge, including: Subtract the first frequency percentage from the second frequency percentage of each edge, retain only the part with a positive difference as a positive increment, and take the ratio of the positive increment to the first frequency percentage as a positive relative change feature. The health risk score for each edge is obtained by multiplying the environmental thrust characteristics, the positive relative change characteristics, and the initial impedance.

10. The method for assessing the health status of substation steel structures by integrating multi-source heterogeneous monitoring data according to claim 9, characterized in that, After determining the health risk score for each edge, the following is also included: If the maximum value of the health risk score is greater than the preset upper limit of the normal path transfer, the preset alarm threshold of the sensor is reduced according to the degree to which the maximum value of the health risk score exceeds the upper limit of the normal path transfer; otherwise, the preset alarm threshold is maintained, and the adjusted alarm threshold is obtained. If the monitoring data from the sensors installed on the pole exceeds the adjusted alarm threshold, the corresponding pole is determined to have a structural safety hazard, and the three-dimensional spatial coordinates of the pole are output.

Images