A method and system for service state inversion and performance evaluation of in-service cable net structure

Abstract

This invention discloses a method and system for service status inversion and performance evaluation of in-service cable-net structures. The method acquires information based on structural inspection and health monitoring data, and determines a reasonable range for unknown parameters to be identified. It completes the global node coordinate evolution and, under given candidate parameters, combines theoretical initial values ​​with inverse iterative correction to obtain a structural equilibrium state consistent with the current geometry. Under sparse temperature measurement points, it reconstructs the global temperature field time history, applies temperature-dominated external excitations and other monitoring excitations to the structural state model for time-series analysis, and obtains the structural time-series response synchronized with monitoring. It constructs a "parameter-time-series response" dataset and trains a hybrid neural network model to identify unknown parameters. Finally, it substitutes the identification results back into the current state model of the in-service structure and verifies its consistency. After successful verification, it performs a unified loading comparison with the design state model, outputting performance indicators such as bearing capacity and stiffness, as well as degradation evaluation results.

Description

Technical Field

[0001] This invention belongs to the field of civil engineering spatial structure health monitoring and performance evaluation technology, specifically involving a method and system for inverting the service status and evaluating the performance of in-service cable net structures. Background Technology

[0002] Cable-net structures in service are typical prestressed flexible structural systems, and their mechanical state (geometry and stiffness level) is determined by the global cable forces. However, the following problems often exist in the engineering structural evaluation of existing cable-net structures in service: (1) Uncertain load: The roof dead load / additional load has undergone multiple changes, and it is difficult to obtain a reliable real load condition under normal structural use conditions.

[0003] (2) The absolute cable force is difficult to measure: The force measurement of in-service cables is often limited by factors such as sensor placement conditions, calibration conditions and prohibition of destructive operations, making it difficult to measure the true cable force.

[0004] (3) Incomplete design data and accumulated deviations: The structure has a long service life and design data is missing. After long-term service, the relaxation of anchors / cables causes unknown degradation of cable force, resulting in a deviation between the structure's "design state" and "current state" in terms of geometry and stiffness. As a result, the design state model can no longer represent the current service state of the structure.

[0005] In existing technologies, some methods utilize the "force-shape correspondence" or geometry-based inversion concepts to identify cable forces in cable net structures, but these methods often rely on high-fidelity initial models with known load parameters. Furthermore, identification methods based on dynamic response are often difficult to implement in engineering applications when modalities are unclear, parameter coupling is significant, and excitation is complex. To achieve state inversion and quantifiable performance evaluation of in-service cable net structures under conditions of unknown load parameters, difficult-to-measure cable forces, and incomplete data, a systematic method that is feasible for engineering implementation is urgently needed. Summary of the Invention

[0006] To address the technical problems existing in the prior art, this invention aims to provide a method for inverting the service status and evaluating the performance of in-service cable net structures. This method is implemented under conditions where the structural geometry is measurable but the cable forces are unknown and the actual loads are uncertain. Specifically, it performs global geometric evolution and structural state inversion on the cable net structure, based on only local detection / monitoring information. With a fixed current geometric shape, it obtains the structural state under different parameters by changing uncertain load parameters, and then acquires time-series response data based on a finite element model of the structural state, creating a dataset of different load parameters and time-series responses. A neural network model suitable for identifying time-series response characteristics is built, trained, and used to identify unknown load parameters. Based on the identified load conditions, a current structural state model is established, and its performance indicators, such as bearing capacity and stiffness, are quantitatively evaluated by comparing it with the structural design state.

[0007] To achieve the above objectives, the technical solution adopted by the present invention is as follows: In a first aspect, the present invention provides a method for inverting the service status and evaluating the performance of in-service cable net structures, comprising the following steps: S1, acquire structural inspection information and structural health monitoring information of the in-service cable net structure; the information includes at least: geometric measurement information of some key nodes, component specification information, structural timing response information and external excitation information; S2, Based on the structural detection information and engineering constraints, determine the unknown parameters to be identified and their reasonable value ranges, and establish an unknown parameter space; S3, under the condition of only having some key node geometric measurement information, obtain a global node coordinate set that can characterize the current geometric shape of the structure through coordinate evolution; S4. For any specified parameter in the unknown parameter space, combined with the strong correlation between the "geometry and prestress" of the cable net structure, the initial theoretical value is calculated based on the force density method and the cable force is corrected by reverse iteration to obtain the structural equilibrium state that is geometrically consistent with the global node coordinate set, thereby obtaining the structural service state model under the candidate parameter. S5, under the condition of sparse temperature measurement points, reconstruct the global temperature field time history of the structure, and apply the external excitation information to the service state model of the structure, and perform time series analysis to obtain the structure time series response information synchronized with the monitoring; S6, obtain samples by traversing the unknown parameter space. N Set the parameters, repeat steps S4 to S5, and construct the dataset of "unknown parameters - structural time response"; N It is a natural number; S7. Construct a hybrid neural network model for recognizing time-series response features, train it using the dataset, input the real monitored time-series response into the trained model, and output the unknown parameter recognition results of the in-service structure. S8, the identification results are used to establish the current state model of the in-service structure and the response consistency verification is performed; after the verification is passed, the in-service state model and the design state model are compared and analyzed under a unified loading system, and the performance indicators and their degradation evaluation results are output.

[0008] As a further technical solution, the structural health monitoring information also includes displacement time history and / or strain time history, and the external excitation information includes temperature-related time-series excitation information, wind and snow load-related time-series excitation information, and snow load-related time-series excitation information.

[0009] As a further technical solution, the coordinate evolution of step S3 includes: fitting the cable shape based on the known point geometric information to complete the cable node coordinates, and combining the damping write-back of the intersection of the main cable and the secondary cable and global smoothing to realize the global node coordinate evolution until the geometric error meets the preset threshold; or directly obtaining the global geometric information through technical means.

[0010] As a further technical solution, step S4 also includes a secondary cable force correction process, which includes: first obtaining the calculated cable force or initial strain as the initial state, and then performing secondary correction on the global cable force through reverse iteration until the geometric error between the model shape and the global node coordinate set obtained in step S3 is less than a specified threshold.

[0011] As a further technical solution, in step S5, reconstructing the global temperature field of the structure includes using a graph regularization method to discretize the cable net into a graph structure and solving it based on topological adjacency relationships, so that the reconstructed temperature field simultaneously satisfies the requirements of measurement point consistency, spatial smoothness, temporal continuity, and boundary correction, thereby suppressing the drift of areas without measurement points.

[0012] As a further technical solution, when constructing the dataset in step S6, the input time-series response is normalized and noise perturbation is added to enhance the robustness of the model. At the same time, a random mask can be added to simulate the detection of bad points or short-term missed detection.

[0013] As a further technical solution, the hybrid neural network model in step S7 includes at least a convolutional structure for extracting temporal local features, a recurrent neural network structure for modeling temporal dependencies, and an attention mechanism for highlighting key segments related to temperature excitation, and outputs regression results for identifying continuous parameters and / or classification results for identifying discrete parameters.

[0014] As a further technical solution, step S7 outputs statistical results through multiple forward calculations during the inference stage, which are used to characterize the reliability and risk boundary of the identification results under the condition of no truth value.

[0015] As a further technical solution, the comparative analysis in step S8 includes: comparing the design state model and the in-service state model under a unified loading regime to obtain the bearing capacity and stiffness degradation index, and combining the cable force comparison to analyze the cable force change characteristics.

[0016] Secondly, the present invention also provides a service status inversion and performance evaluation system for in-service cable net structures, comprising: The data acquisition module is configured to acquire structural inspection information and structural health monitoring information of the in-service cable net structure; the information includes at least: geometric measurement information of some key nodes, structural time-series response information, and external excitation information; The unknown parameter space establishment module is configured to determine the unknown parameters to be identified and their reasonable value ranges based on the structural detection information and engineering constraints, and to establish the unknown parameter space. The coordinate evolution module is configured to obtain a global set of node coordinates that can characterize the current geometric shape of the structure by means of coordinate evolution, given only partial key node geometric measurement information. The state inversion module is configured to take any specified parameter in the unknown parameter space, combine the strong correlation characteristics of "geometry-prestress" of the cable net structure, calculate the theoretical initial value based on the force density method and the reverse iterative cable force correction to obtain the structural equilibrium state that is geometrically consistent with the global node coordinate set, thereby obtaining the structural service state model under the candidate parameter. The temperature field reconstruction and time series analysis module is configured to reconstruct the global temperature field time history of the structure under the condition of sparse temperature measurement points, and apply external excitation information to the structure service state model to perform time series analysis to obtain the structure time series response information synchronized with monitoring. The dataset construction module is configured to traverse and sample within the unknown parameter space to obtain... N The parameters are set, and the state inversion module and temperature field reconstruction and time series analysis module are called to construct a dataset of "unknown parameters - structural time series response". The parameter identification module is configured to construct a hybrid neural network model for identifying time-series response features, train it using the dataset, input the real monitored time-series response into the trained model, and output the unknown parameter identification results of the in-service structure. The performance evaluation module is configured to use the identification results to build a current state model of the in-service structure and perform response consistency verification. After the verification is passed, the in-service state model and the design state model are compared and analyzed under a unified loading regime, and the performance indicators and their degradation evaluation results are output.

[0017] The working principle of this invention is as follows: This invention utilizes the strong correlation between "geometry and prestress" in prestressed cable net structures, taking the consistency between form and force as the core constraint for state inversion. It achieves global geometric evolution through the coordinates of a finite number of key measuring points. Through secondary cable force correction using the force density method and the inverse iteration method, it achieves a prestress state where the model shape is consistent with the measured shape under given parameters. Furthermore, it expands sparse temperature measuring points into full-field temporal excitation by reconstructing the temperature field, synchronizing the finite element temporal response with monitoring. Finally, it learns the implicit nonlinear mapping between "unknown parameters and structural temporal response" from the temporal response using a hybrid neural network model, completing the identification of unknown parameters and realizing the quantitative evaluation of structural performance.

[0018] Compared with the prior art, the present invention has the following advantages: This invention proposes a computational method and system for in-service single-layer orthogonal cable net structures, enabling structural service state inversion identification and further load-bearing capacity and stiffness performance assessment under conditions where cable forces are difficult to measure directly, loads and environmental excitations are uncertain, and structural information is limited. This method and system, even with unknown absolute cable forces, can achieve structural state inversion through morphological consistency and secondary cable force correction, reducing reliance on absolute cable force measurements. The method and system establish a nonlinear mapping dataset of unknown parameters and time-series responses (external excitation primarily based on temperature) (a dataset of "unknown parameters - structural time-series responses"). A hybrid neural network model can be established to automatically identify key unknown parameters, making it suitable for scenarios with incomplete data for in-service cable net structures. Through graph-regularized temperature field reconstruction, the method and system can expand sparse temperature measurement points into global temperature excitations while ensuring spatiotemporal smoothness and boundary rationality, significantly reducing the amount of monitoring data required for structural assessment. Finally, the method and system achieve quantitative assessment through a modelable current state, with assessment indicators directly supporting engineering decisions and providing a basis for safety assessment and operation and maintenance strategies. Attached Figure Description

[0019] Figure 1 This is a flowchart of the in-service cable net structure state inversion and performance evaluation method proposed in this invention; Figure 2 is a flowchart of the overall process for state inversion and performance evaluation of in-service cable net structures disclosed in some embodiments of the present invention. Figure 3 is a flowchart of the reverse iterative process for structural state inversion under given parameters disclosed in some embodiments of the present invention. Figure 4 This is a three-dimensional schematic diagram of an in-service cable net structure disclosed in some embodiments of the present invention; Figure 5 This is a schematic diagram of the HNN neural network model structure and information flow disclosed in some embodiments of the present invention; Detailed Implementation To enable those skilled in the art to better understand the technical solutions of the present invention, the present invention will be further described below with reference to embodiments. It should be understood that the following embodiments are only used to explain the present invention and are not intended to limit the scope of protection of the present invention.

[0020] As described in the background section, some existing methods utilize the "force-shape correspondence" or geometry-based inversion concepts to identify cable forces in cable net structures, but these methods often rely on high-fidelity initial models with known load parameters. Furthermore, identification methods based on dynamic response are often difficult to implement in engineering projects when modalities are unclear, parameter coupling is significant, and excitations are complex. To achieve state inversion and quantifiable performance evaluation of in-service cable net structures under conditions of unknown load parameters, difficult-to-measure cable forces, and incomplete data, a systematic method that is feasible for engineering implementation is urgently needed. To address the technical problems existing in the prior art, this invention provides a method for inverting the service status and evaluating the performance of in-service cable net structures. This method is implemented under conditions where the structural geometry is measurable but the cable forces are unknown and the actual loads are uncertain. The method comprises: 1. Performing global geometric evolution and structural state inversion on the cable net structure based on only local detection / monitoring information; 2. Obtaining the structural state under different parameters by changing uncertain load parameters while maintaining the current geometric shape of the structure, and then acquiring time-series response data based on the finite element model of the structural state, creating a dataset of time-series responses for different load parameters; 3. Building a neural network model suitable for identifying time-series response characteristics, completing training, and identifying unknown load parameters; 4. Establishing a current structural state model based on the identified load conditions, and quantitatively evaluating performance indicators such as bearing capacity and stiffness by comparing it with the structural design state.

[0021] Implementation Method 1 This embodiment provides a method for in-service cable net structure service status inversion and performance evaluation. This method addresses the engineering conditions of in-service cable net structures, including "difficulty in directly measuring cable forces, uncertainty of external loads, limited measurement points, and uninterrupted operation." It achieves: global geometric reconstruction and coordinate evolution; service status inversion based on initial force density values ​​and inverse iteration; temperature field reconstruction and generation of time-series response datasets; parameter identification using a hybrid neural network; and load-bearing capacity and stiffness performance evaluation by comparing the design state and in-service state. The method includes the following steps: Figure 2 As shown: Step S1: Measure structural information, i.e., obtain structural inspection information and structural health monitoring information of the in-service cable net structure. The structural inspection obtains local real information of the in-service cable net structure, including but not limited to: three-dimensional coordinates of key structural locations (structural boundaries, key nodes) or convertible geometric information, including cable net span and local coordinates, suspension cable sag ratio, boundary coordinates; component specification information, including cable arrangement, cable type and cross-sectional specifications. Structural response and excitation are obtained through health monitoring, including but not limited to: displacement / deformation time-series response information; cable stress-strain time-series data; temperature field-related time-series excitation information and external meteorological time-series excitation information, such as roof wind load-related time-series excitation information, roof snow load-related time-series excitation information, and other meteorological information.

[0022] Step S2: Determine the reasonable range of unknown parameters, that is, determine the unknown parameters to be identified and their reasonable value range, and establish the unknown parameter space; Based on structural inspection and statistical analysis, the unknown parameters to be identified and their reasonable value ranges are determined. These unknown parameters include at least key parameters that can cause changes in the cable net state, such as structural loads and component specifications. The unknown parameter space Ω (one-dimensional or multi-dimensional) is then given based on the reasonable range of the parameters.

[0023] Step S3: Perform structural state inversion and time-series response analysis based on the current geometric state of the structure. This step involves the following three main technical solutions, the specific process of which is as follows: Figure 3 As shown.

[0024] Given unknown parameters Under the condition of ∈Ω (e.g., load information), the vertical force density equilibrium matrix of the global nodes of the structure is first established based on the force density method. The first calculated cable force / initial strain is then obtained through iteration. Based on the first calculated cable force / initial strain and the global node coordinate set... X meas Given unknown parameters A finite element model of the cable net structure is established, then the equilibrium state of the structure is solved, and the calculated nodal coordinates and the global nodal coordinate set are extracted. X meas The difference is used to adjust the initial input coordinates of the finite element model by employing a reverse iterative geometric correction method of inverse correction + damping update, as shown in the following formula. X (n+1) These coordinates will be used as input for the next round of calculations. X (n) These are the input coordinate values ​​from the previous calculation. It is the damping update coefficient. It is the difference between the coordinates of the previous calculation and the input coordinates. This represents the maximum updated coordinate value for this round. After updating the coordinates, the first calculation of cable force / initial strain and the given unknown parameters are applied again. The global cable force is corrected a second time through multiple inverse iterations until the final calculated stress state of the cable net structure and the global node coordinate set are obtained. X meas If the geometric error is less than a specified threshold, the current structural state is considered to be the current geometric shape of the structure, given unknown parameters. The geometric state under the given conditions.

[0025]

[0026] (3) Based on monitoring external excitation, perform finite element time series analysis on the structural state under the current parameters. It is known that under the service conditions of a structure, external excitation is dominated by temperature, with other excitations such as wind load accounting for a smaller proportion. To ensure that the finite element time-series excitation is consistent with the actual structural excitation, the time-series global temperature field is reconstructed under the condition of sparse temperature measurement points, and the cable net is discretized into an undirected graph. G =( V,E Constructing a graph Laplacian matrix based on topological adjacency relationships. L Structural topology based on Figure 4 A three-dimensional schematic diagram of the structure is shown. A constrained quadratic optimization is constructed, as shown in the optimization objective below. While strictly matching the temperature at the measuring points, the global temperature is made spatially smooth and temporally continuous, and boundary corrections are introduced for edge regions to suppress drift in areas without measuring points. Thus, the full-time temperature field matrix T(t) is obtained. Using the baseline time as a reference, ΔT(t) is formed as a volume load applied to the finite element model. Simultaneously, wind and snow loads are monitored to correct the load conditions, and finally, the displacement / strain time-series response synchronized with monitoring is obtained through the finite element model.

[0027]

[0028] Step S4: Construct a dataset of unknown structural parameters and their corresponding time-series responses under structural states. Based on step S3, it can be evolved into N The group specifies the structural state under unknown parameter conditions. Based on the evolving structural state, the time-series response (displacement / strain) synchronized with monitoring can be calculated. The specified unknown parameters will be used. As a label q k By using the corresponding structural time-series response as input features, a structure can be constructed. N Group of unknown parameters—time series response data. The dataset shape is... N × D in × T . N For the sample size, D in For the displacement, strain, and temperature data of the structure, T This is a time series dataset. Small perturbations to the material / section can be further introduced to cover potential damage or modeling uncertainties. The dataset is randomly divided into training, validation, and test sets in an 8:1:1 ratio. D in × T The dimensions are normalized and Gaussian noise is added. This noise improves the robustness of the subsequent neural network model. A random mask is also added to simulate potential bad detections or short-term missed detections during monitoring.

[0029] Step S5: Identify unknown parameters based on a hybrid neural network model trained with a dataset. Construct a hybrid neural network model suitable for recognizing features of time-series signals, see details. Figure 5 The model input is a structural time-series response dataset, and the output is the unknown parameters to be identified. The network includes: a feature input & feature segmentation preprocessing module, which divides the dataset into structural response features (…). N , T , X s ) and temperature time series characteristics ( N , T , X t ),in D in =[ X s , X t First, temperature baseline processing is used to extract the residuals of the structural response features at the temperature baseline. X s ′= X s - X base ,in X base = W·X t +b; then calculate the first-order time difference to extract the structural response Δ under temperature difference. X = X s ′[t]- X s ′[t-1], and then the processed structural features are spliced ​​together. X aug =[ X s ′,Δ X ], forming new output features ( N , T , X aug ); further, the features ( N , T , X aug The input is fed into a one-dimensional temporal convolutional module (1D-CNN), which extracts temporal local features through three layers of one-dimensional convolutional blocks. These features are then fed into the Squeeze-and-Excitation Module (SE), which sequentially employs temporal average pooling, fully connected squeezing layers, fully connected excitation layers, and a broadcast mechanism layer to focus on information-rich feature channels, thereby forming new processed output features. N , T , H CNN Further, the features (N , T , H CNN The input is processed by the BiLSTM time series information processing module. After processing by LSTM in both forward and reverse time series, the features are concatenated to form (…). N , T , H LSTM In addition, to further amplify the model's focus on structural features at extreme temperatures, a bias calculation is performed on the extreme temperatures. First, the time-series features are input ( N , T , X t Then, the average temperature is calculated along the feature dimension. mag t =Mean(| X t |), and then perform Z-score normalization to form a normalized temperature feature. z t Then specify the standardized temperature characteristic threshold. z thresh The normalized temperature above a threshold is identified by using the ReLU activation function. z t ' Finally, these standardized temperatures with large temperature differences are scaled up. k This creates a temperature bias. Bias ext = k z t ' This results in a temperature bias characteristic ( N , T Further, the temperature bias characteristics ( N , T ) and structural features after LSTM processing ( N , T , H LSTM Input the additive attention module with the original temperature features ( X t ) as the query, with H LSTM As a key-value pair, while adding temperature bias Bias ext Subsequently, the context vector is obtained through a linear layer, a nonlinear fusion layer, bias injection, and softmax normalization, thereby obtaining the output features. N , H attThe above modules are used to adaptively highlight key information related to temperature excitation in the time dimension; finally, the regression output head of the fully connected module layer is used to identify continuous parameters, or the classification output head is used to identify discrete parameters.

[0030] Step S6: Real-world modeling verification and performance evaluation The prediction parameters identified in step S5 Substitute the parameters into the model and reconstruct the finite element structural state model of the in-service cable net structure based on the predicted parameters from step S3. Verify the consistency by monitoring the time-series responses such as displacement and strain. After successful verification, conduct performance evaluation. Analyze cable force changes by comparing the cable forces with the structural design model; and compare the structural bearing capacity and stiffness degradation indices by static nonlinear loading.

[0031] The above method will be further explained below with specific embodiments; (1) Structural information acquisition and preprocessing a) Obtain structural inspection information and structural health monitoring data of in-service cable net structures, wherein the data includes at least: spatial coordinates (or convertible geometric information) of some key nodes, displacement time history and / or strain time history, time history of temperature field measurement points in structural layers, and roof meteorological data.

[0032] b) Time synchronization and baseline unification of monitoring data: Using a reference time t0 as the time baseline, unify the starting time and zero point of each channel; use threshold or statistical methods to remove / correct significant deviations, and correct the monitoring data to time-series response or time-series excitation change values.

[0033] (2) Coordinate evolution and global geometric reconstruction a) Establish a global coordinate system: Determine the origin and direction of the global coordinate axes using the center point of the structure.

[0034] b) Coordinate mapping: Map the measured local node coordinates or field measurement coordinates to the global coordinate system.

[0035] c) Unmeasured node coordinate completion: The coordinates of unmeasured structural nodes are completed by interpolation / fitting, and a smoothing constraint (such as smoothing or regularization based on adjacency relationship) is applied to the completed coordinate field to obtain the global node coordinate set Xmeas.

[0036] (3) Parameter space definition (set of parameters to be identified) The parameter θ to be identified based on the structural conditions may include, but is not limited to, external equivalent uniformly distributed load parameters, component size parameters, and structural cable forces. A reasonable parameter range space Ω is determined based on actual structural monitoring information, and subsequent parameter traversal within Ω is performed to collect data and establish the system.

[0037] (4) Structural state inversion under given parameters (initial force density + reverse iteration) For any specified θ For ∈Ω, perform the following inversion process: a) Solving for the initial value of the force density cable force: Based on the cable net topology and boundary conditions, the theoretical cable force (or equivalent force density) of the cable segment is obtained by the force density method and written into the initial state (e.g., initial strain or initial stress) of the finite element model.

[0038] b) Static nonlinear analysis: Applying specified parameters to the structure θ The corresponding equivalent load conditions are applied, and external excitations (temperature, wind, snow loads, etc.) obtained from monitoring are applied simultaneously. Static nonlinear calculations are then performed to obtain the calculated coordinates. X cal .

[0039] c) Convergence determination and reverse correction based on inverse iteration method: extract the calculated coordinates of the nodes, and... X meas The residual is obtained by subtraction. If the residual meets the threshold, the iteration stops and the result is output. θ The prestress state of the structure is determined; if it is not satisfied, a geometric update strategy of "displacement difference reverse correction + damping write-back" is adopted: the displacement reverse iteration method is performed on the global node coordinates, the initial cable force and load are applied again, the residuals are calculated and iterated repeatedly until the structural geometry and global node coordinates are consistent. X meas The residuals meet the requirements.

[0040] (5) Reconstruction of main external excitation (temperature field) and generation of temporal response (graphical Laplace constraint optimization) After obtaining the prestress equilibrium state that satisfies the current structural geometry, a time-series finite element analysis driven by temperature field reconstruction is used to generate a time-series response consistent with actual monitoring (other secondary excitations can be supplemented). The temperature field reconstruction method mainly discretizes the cable net into an undirected graph and constructs the adjacency matrix and the graph Laplacian matrix. L For each moment t Treating the temperature field as a spatial field under the steady-state heat conduction approximation, we solve a constrained optimization problem to reconstruct the temperature vector. T ( t Simultaneously satisfying the requirements of measurement point fidelity, spatial smoothing, temporal continuity, and boundary correction, global temperature field reconstruction under sparse measurement point conditions can be achieved based on the graph Laplace method. Subsequently, the baseline increment method is used to obtain the full-time domain temperature field variation data Δ. T ( t The time-series response is applied to the finite element model and solved step-by-step, outputting the displacement / strain responses corresponding to the monitoring points. y ( t ).

[0041] (6) Parameter identification - Construction of time series response dataset Traverse the parameter space Ω with a certain step size. θ i Repeat steps (4) and (5) for each set of parameters to obtain... N Group parameters θ i Different structural timing responses By using structural time-history responses such as displacement and strain as the main input features, and simultaneously incorporating temperature time-history data, the neural network model is guided to have better identification capabilities; parameters are used... θ i As a label to be identified.

[0042] (7) HNN training and parameter identification a) Constructing a hybrid neural network: 1D-CNN is used to extract local temporal features, BiLSTM is used to capture bidirectional time dependence, and a temperature-guided additive attention mechanism (TempAttention) is introduced in the temporal dimension. Temperature baseline processing and SE-type channel recalibration modules can be further set to enhance feature expression and noise resistance.

[0043] b) Train the network using the dataset to learn from time-series responses. y ( t ) to parameters θ The mapping relationship allows the real-time monitoring data to be input into the trained neural network model, and the output will be the recognition parameters. During the inference phase, random perturbations are maintained and multiple forward samples are performed. The mean and confidence interval are output to evaluate the reliability and risk boundary of the no-truth identification results.

[0044] (8) Verification and performance evaluation (comparison between design state and in-service state) a) Inversion verification: using identification parameters Reconstruct the finite element model of the in-service state and generate a time-series response consistent with the monitoring conditions; compare the calculated response with the monitoring response, and if the error meets the preset threshold, the in-service state inversion is considered reliable.

[0045] b) Performance evaluation under a uniform loading regime: After verification, a consistent static loading regime (e.g., monotonically increasing uniformly distributed load or graded loading) is applied to both the design model and the in-service model. The ultimate bearing capacity is obtained based on the preset failure criterion. Considering that under vertical load, the cable net is initially controlled by the combined action of the main cables and auxiliary cables, but as the load increases, the auxiliary cables may gradually loosen, causing a decrease in overall stiffness. Therefore, the "early stage when the auxiliary cables are still under tension" can be used as the initial stiffness comparison interval. The comparison results of bearing capacity and initial stiffness are output, and degradation indices (e.g., bearing capacity ratio, stiffness ratio, or relative degradation rate) are calculated to provide a quantitative basis for maintenance and reinforcement decisions.

[0046] Implementation Method Two: This embodiment also provides a system for in-service cable net structure service status inversion and performance evaluation, including: a data acquisition module, an unknown parameter space establishment module, a coordinate evolution module, a state inversion module (two-stage cable force correction of initial force density + inverse iteration), a temperature field reconstruction module, a dataset construction module, a parameter identification module (HNN), and a performance evaluation module (unified loading comparison output of design state / in-service state). Each module is configured to execute the corresponding steps in the method described in Embodiment 1. Each module can be deployed on a local workstation or server, supporting offline evaluation and online updates.

[0047] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for inverting the service status and evaluating the performance of an in-service cable net structure, characterized in that, Includes the following steps: S1, acquire structural inspection information and structural health monitoring information of the in-service cable net structure; the information includes at least: geometric measurement information of some key nodes, structural time-series response information and external excitation information; S2, Based on the structural detection information and engineering constraints, determine the unknown parameters to be identified and their reasonable value ranges, and establish an unknown parameter space; S3, under the condition of only having some key node geometric measurement information, obtain a global node coordinate set that can characterize the current geometric shape of the structure through coordinate evolution; S4. For any specified parameter in the unknown parameter space, combined with the strong correlation between the "geometry and prestress" of the cable net structure, the initial theoretical value is calculated based on the force density method and the cable force is corrected by reverse iteration to obtain the structural equilibrium state that is geometrically consistent with the global node coordinate set, thereby obtaining the structural service state model under the candidate parameter. S5, under the condition of sparse temperature measurement points, reconstruct the global temperature field time history of the structure, and apply the external excitation information to the service state model of the structure, and perform time series analysis to obtain the structure time series response information synchronized with the monitoring; S6, obtain samples by traversing the unknown parameter space. N Set the parameters, repeat steps S4 to S5, and construct the dataset of "unknown parameters - structural time response"; N It is a natural number; S7. Construct a hybrid neural network model for recognizing time-series response features, train it using the dataset, input the real monitored time-series response into the trained model, and output the unknown parameter recognition results of the in-service structure. S8, the identification results are used to establish the current state model of the in-service structure and the response consistency verification is performed; after the verification is passed, the in-service state model and the design state model are compared and analyzed under a unified loading system, and the performance indicators and their degradation evaluation results are output.

2. The method for service status inversion and performance evaluation of in-service cable net structures according to claim 1, characterized in that: The structural health monitoring information also includes displacement time history and / or strain time history, and the external excitation information includes temperature-related time-series excitation information, wind and snow load-related time-series excitation information, and snow load-related time-series excitation information.

3. The method for service status inversion and performance evaluation of in-service cable net structures according to claim 1, characterized in that: The coordinate evolution in step S3 includes: fitting the cable shape based on the known point geometric information to complete the cable node coordinates, and combining the damping write-back of the intersection of the main cable and the secondary cable with global smoothing to realize the global node coordinate evolution until the geometric error meets the preset threshold; or directly obtaining the global geometric information through technical means.

4. The method for service status inversion and performance evaluation of in-service cable net structures according to claim 1, characterized in that: Step S4 further includes a secondary cable force correction process, which includes: first obtaining the calculated cable force or initial strain as the initial state, and then performing a secondary correction on the global cable force through reverse iteration until the geometric error between the model shape and the global node coordinate set obtained in step S3 is less than a specified threshold.

5. The method for service status inversion and performance evaluation of in-service cable net structures according to claim 1, characterized in that: In step S5, reconstructing the global temperature field of the structure includes discretizing the cable net into a graph structure and solving it based on topological adjacency relationships, so that the reconstructed temperature field simultaneously meets the requirements of measurement point consistency, spatial smoothness, temporal continuity, and boundary correction, in order to suppress the drift of areas without measurement points.

6. The method for service status inversion and performance evaluation of in-service cable net structures according to claim 1, characterized in that: In step S6, when constructing the dataset, the input time-series response is normalized and noise perturbation is added to enhance the robustness of the model. At the same time, a random mask can be added to simulate the detection of bad points or short-term missed detections.

7. The method for service status inversion and performance evaluation of in-service cable net structures according to claim 1, characterized in that: The hybrid neural network model in step S7 includes at least a convolutional structure for extracting temporal local features, a recurrent neural network structure for modeling temporal dependencies, and an attention mechanism for highlighting key segments related to temperature excitation, and outputs regression results for identifying continuous parameters and / or classification results for identifying discrete parameters.

8. The method for service status inversion and performance evaluation of in-service cable net structures according to claim 1, characterized in that: Step S7 outputs statistical results through multiple forward calculations during the inference phase, which are used to characterize the reliability and risk boundary of the identification results under the condition of no truth value.

9. The method for service status inversion and performance evaluation of in-service cable net structures according to claim 1, characterized in that: The comparative analysis in step S8 includes: comparing the design state model and the in-service state model under a unified loading regime to obtain the bearing capacity and stiffness degradation indices, and combining the cable force comparison to analyze the cable force variation characteristics.

10. A service status inversion and performance evaluation system for in-service cable net structures, characterized in that, include: The data acquisition module is configured to acquire structural inspection information and structural health monitoring information of in-service cable net structures; The information includes at least: geometric measurement information of some key nodes, structural timing response information, and external excitation information; The unknown parameter space establishment module is configured to determine the unknown parameters to be identified and their reasonable value ranges based on the structural detection information and engineering constraints, and to establish the unknown parameter space. The coordinate evolution module is configured to obtain a global set of node coordinates that can characterize the current geometric shape of the structure by means of coordinate evolution, given only partial key node geometric measurement information. The state inversion module is configured to take any specified parameter in the unknown parameter space, combine the strong correlation characteristics of "geometry-prestress" of the cable net structure, calculate the theoretical initial value based on the force density method and the reverse iterative cable force correction to obtain the structural equilibrium state that is geometrically consistent with the global node coordinate set, thereby obtaining the structural service state model under the candidate parameter. The temperature field reconstruction and time series analysis module is configured to reconstruct the global temperature field time history of the structure under the condition of sparse temperature measurement points, and apply external excitation information to the structure service state model to perform time series analysis to obtain the structure time series response information synchronized with monitoring. The dataset construction module is configured to traverse and sample within the unknown parameter space to obtain... N The parameters are set, and the state inversion module and temperature field reconstruction and time series analysis module are called to construct a dataset of "unknown parameters - structural time series response". The parameter identification module is configured to construct a hybrid neural network model for identifying time-series response features, train it using the dataset, input the real monitored time-series response into the trained model, and output the unknown parameter identification results of the in-service structure. The performance evaluation module is configured to use the identification results to build a current state model of the in-service structure and perform response consistency verification. After the verification is passed, the in-service state model and the design state model are compared and analyzed under a unified loading regime, and the performance indicators and their degradation evaluation results are output.

Images