An online testing and evaluation method for the performance of energy-dissipating components used in seismic resistance of buildings.

Abstract

This invention discloses an online testing and evaluation method for the performance of energy-dissipating components in seismic-resistant buildings. The method includes: deploying response acquisition devices at both ends of the component to collect structural response data and preprocessing it to obtain nodal response data; calculating the relative displacement and velocity sequences at both ends, establishing interface mapping relationships, constructing an energy domain perspective matrix and inverting interface forces; calculating energy characteristic parameters, establishing a dynamic energy closed cone constraint interval, and performing energy projection correction; generating hysteresis frames and power residual frames, constructing a multi-channel energy time-series spectrum, and inputting improved MaxViT to extract features; separating reversible and irreversible response components to obtain equivalent stiffness yield and energy dissipation characteristic quantities; calculating performance degradation indices, comparing them with degradation threshold intervals, and determining the component performance degradation level. This invention achieves online evaluation of the performance of seismic-resistant energy-dissipating components in buildings through structural response inversion and energy constraint analysis.

Description

Technical Field

[0001] This invention relates to the field of structural health monitoring and structural component performance testing technology, and in particular to an online testing and evaluation method for the performance of energy-dissipating components used in building seismic resistance. Background Technology

[0002] Under seismic loading, building structures typically dissipate incoming seismic energy by incorporating energy-dissipating components, thereby reducing the dynamic response of the main structure and improving overall seismic performance. With the development of structural health monitoring technology, engineering practice increasingly requires long-term monitoring of the performance of energy-dissipating components under actual service conditions to promptly understand their stiffness changes, yielding behavior, and key mechanical properties such as energy dissipation capacity. Currently, common testing methods mainly rely on laboratory loading tests or obtaining force-displacement hysteresis relationships by arranging force sensors and displacement sensors on the components, thereby assessing the performance status of energy-dissipating components.

[0003] However, in practical engineering environments, the above methods have significant limitations. On the one hand, laboratory loading tests usually require disassembling the components and conducting them on specialized testing equipment, making it difficult to reflect the stress state of the structure under actual service conditions and impossible to achieve long-term online monitoring. On the other hand, directly placing force sensors in actual building structures is not only complex and costly to install, but also susceptible to factors such as temperature changes, environmental vibrations, and sensor drift in long-term service environments, resulting in poor stability of test data. Under weak excitation conditions such as wind vibration or minor earthquakes, the structural response amplitude is small, and the measured mechanical signals are often submerged by noise interference, making it difficult to accurately reconstruct the force-displacement hysteresis relationship and affecting the identification of stiffness degradation and changes in energy dissipation capacity of energy-consuming components.

[0004] Existing technologies struggle to reliably obtain mechanical response information of energy-dissipating components under actual service conditions. In particular, without the need for direct installation of force sensors, it is difficult to accurately reconstruct the force-displacement relationship and hysteresis behavior of energy-dissipating components, thus making it difficult to reliably assess the degree of component performance degradation.

[0005] Therefore, how to provide an online testing and evaluation method for the performance of energy-dissipating components used in seismic resistance of buildings is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0006] One objective of this invention is to propose an online testing and evaluation method for the performance of energy-dissipating components in seismic-resistant buildings. This invention achieves online reconstruction of the force-displacement relationship of energy-dissipating components without the need for direct force sensor installation by constructing an interface force inversion method based on structural response data, a dynamic energy closed cone constraint correction mechanism, and a hysteretic response feature identification method combined with an improved MaxViT network. Furthermore, it identifies and analyzes the equivalent stiffness characteristics, yield characteristics, and energy dissipation characteristics of the components, thereby enabling online evaluation of the performance degradation state of energy-dissipating components. This invention fully utilizes structural dynamic response analysis, energy constraint correction, and visual feature extraction technologies, possessing advantages such as high test data stability, strong identification capability under weak vibration conditions, and high reliability of online evaluation in engineering sites.

[0007] An online testing and evaluation method for the performance of energy-dissipating components for seismic resistance of buildings, according to an embodiment of the present invention, includes: Response acquisition devices are installed at the structural nodes at both ends of the seismic energy dissipation components of the building to collect structural response data at both ends of the nodes. The structural response data is then preprocessed to obtain node response data. The relative displacement and relative velocity sequences between the nodes at both ends of the energy-consuming component are calculated based on the node response data. The interface degree of freedom mapping relationship is established and the energy domain perspective matrix is ​​constructed. The interface force of the energy-consuming component is inverted by solving the minimum energy residual and the initial interface force sequence is obtained. The energy characteristic parameters of the energy-consuming component are calculated based on the initial interface force sequence and relative displacement sequence. A dynamic energy closed cone constraint interval is established, and the initial interface force sequence and relative displacement sequence are corrected by energy projection to obtain the corrected interface force sequence and corrected relative displacement sequence. Force-displacement hysteresis frames, instantaneous power frames, and energy closure residual frames are generated based on the calibration interface force sequence and calibration relative displacement sequence. These are then merged into a multi-channel energy-time series map and input into an improved MaxViT network to obtain visual feature vectors. Based on the corrected interface force sequence, corrected relative displacement sequence, and visual feature vector, the reversible and irreversible response components of the hysteresis response of the energy dissipation component are separated to obtain the equivalent stiffness characteristic quantity, yield characteristic quantity, and energy dissipation characteristic quantity. The performance degradation index of the energy-consuming component is calculated based on the equivalent stiffness characteristic, yield characteristic and energy dissipation characteristic. The performance degradation index is compared with the preset degradation threshold range to determine the performance degradation level of the energy-consuming component.

[0008] Optionally, the structural response data includes nodal acceleration data, nodal displacement data, and nodal strain data, wherein the nodal acceleration data is collected by acceleration sensors installed at the structural nodes at both ends of the energy dissipation component, the nodal displacement data is obtained by displacement measuring devices installed at the structural nodes, and the nodal strain data is collected by strain sensors deployed in the connection area of ​​the energy dissipation component.

[0009] Optionally, the preprocessing of the structural response data includes time synchronization calibration of the acquired structural response data to unify the data time reference of each response acquisition device, detrending processing of the time-synchronized structural response data to eliminate low-frequency drift components in the data, and noise filtering processing of the detrending structural response data to remove high-frequency noise interference and obtain node response data.

[0010] Optionally, obtaining the initial interface force sequence includes: Displacement and velocity data of structural nodes at both ends of the energy-consuming component are extracted from the node response data. The relative displacement sequence is determined by the difference in displacement of the nodes at both ends, and the relative velocity sequence is determined by the difference in velocity of the nodes at both ends. Determine the connection interfaces between the two ends of the energy-consuming component and the structural substructure. Select a set of interface nodes that correspond one-to-one with the connection interfaces in the left and right substructures respectively. Encode the translational and rotational degrees of freedom of each interface node according to the preset degree of freedom order to obtain the left interface degree of freedom sequence and the right interface degree of freedom sequence. The interface degree-of-freedom mapping relationship is constructed based on the left-side interface degree-of-freedom sequence and the right-side interface degree-of-freedom sequence, specifically as follows: A one-to-one pairing of the left-side interface degree-of-freedom sequence with the right-side interface degree-of-freedom sequence is formed to create an interface pairing table; For each pair of paired degrees of freedom, a symbolic consistency marker is set to distinguish relative motions in the same direction and opposite directions; A closed-loop consistency check is performed on the interface pairing table to eliminate duplicate and missing pairings, thereby obtaining an interface degree-of-freedom mapping relationship that satisfies the completeness of the interface degrees of freedom on both sides. The equivalent inertial parameter sets corresponding to the interface node sets are extracted from the left and right substructures respectively. Based on the symbol consistency mark, the equivalent inertial parameter sets on both sides are aligned and fused in the interface degree of freedom dimension to form the interface energy weight matrix. The energy weight projection is applied to each degree of freedom component of the relative displacement sequence according to the interface energy weight matrix to obtain the energy domain perspective matrix. Based on the energy domain perspective matrix, relative displacement sequence, and relative velocity sequence, an energy residual evaluation quantity of the interface force is constructed. The interface force sequence of the energy-consuming component is obtained by minimizing the energy residual evaluation quantity, which serves as the initial interface force sequence.

[0011] Optionally, obtaining the corrected interface force sequence and the corrected relative displacement sequence includes: Based on the initial interface force sequence and relative displacement sequence, the product of the interface force and the relative displacement change is calculated point by point in time sequence. The product results are accumulated step by step to obtain the input energy change sequence of the energy dissipating component at each sampling time. Based on the maximum displacement amplitude of the relative displacement sequence within the current time window and the equivalent stiffness characteristics of the energy-consuming component, calculate the maximum energy dissipation sequence of the energy-consuming component at each sampling time. The irreversible energy dissipation sequence is calculated based on the input energy change sequence and the maximum dissipable energy change sequence, wherein the irreversible energy dissipation sequence represents the unrecoverable energy generated by the energy dissipating component during loading and unloading through material yielding, friction or damping. A dynamic energy closed cone constraint interval is established based on the input energy change sequence, the maximum dissipable energy change sequence, and the irreversible energy dissipation change sequence, specifically: The lower boundary of energy is determined by the irreversible energy consumption change sequence, the upper boundary of energy is determined by the maximum energy consumption change sequence, and the time evolution trend of the input energy change sequence is used as the energy central axis. By progressively updating the upper boundary, lower boundary, and central axis of energy within a continuous time window, the three elements form a conical energy constraint region with a convergent trend in the time dimension, thus obtaining the dynamic energy closed cone constraint interval. The input energy change sequence is compared with the dynamic energy closed cone constraint interval at each time step. When the input energy change sequence is detected to exceed the dynamic energy closed cone constraint interval, the initial interface force sequence and relative displacement sequence are subjected to energy projection correction processing so that the corrected interface force sequence and relative displacement sequence fall back into the dynamic energy closed cone constraint interval, thus obtaining the corrected interface force sequence and the corrected relative displacement sequence.

[0012] Optionally, obtaining the visual feature vector includes: A force-displacement hysteresis frame sequence is generated based on the corrected interface force sequence and the corrected relative displacement sequence, and an instantaneous power frame sequence is generated based on the corrected interface force sequence and the corrected relative velocity sequence. At the same time, an energy closure residual frame sequence is generated based on the input energy change sequence and the irreversible energy dissipation change sequence. The force-displacement hysteresis frame sequence, instantaneous power frame sequence, and energy closure residual frame sequence are spatially aligned according to the same time index and superimposed in the channel dimension to form a multi-channel energy-time sequence, wherein each sequence frame contains hysteresis information channel, power information channel, and energy closure residual information channel. The multi-channel energy-temporal map sequence is input into the improved MaxViT network. The improved MaxViT network includes, in sequence, a map patch embedding layer, a first multi-axis attention block, an energy consistency enhancement layer, a second multi-axis attention block, a cross-frame temporal fusion layer, third to fifth multi-axis attention blocks, a global feature aggregation layer, and a feature output layer. In the map patch embedding layer, the multi-channel energy-temporal map is divided into map patches of fixed size. Feature mapping is performed on each map patch to obtain a patch feature sequence. All patch feature sequences are arranged in spatial order to form a feature vector sequence. In the energy consistency enhancement layer, energy channel weights are calculated on the feature vector sequence, and energy weight coefficients are generated based on the numerical values ​​of the energy closure residual channels. The feature vectors corresponding to the hysteresis information channel and the power information channel are then updated with weights to obtain the energy consistency enhancement feature sequence. In the cross-frame temporal fusion layer, the patch features of adjacent time frames are spliced ​​in chronological order, and the evolution features of hysteresis morphology over time are extracted through joint operation of local convolution and multi-axis attention. The obtained fused features are then input into the third to fifth multi-axis attention blocks for global feature extraction, and feature compression is performed through the global feature aggregation layer. Finally, a visual feature vector representing the hysteresis evolution state of the energy-consuming component is output in the feature output layer.

[0013] Optionally, obtaining the equivalent stiffness characteristic, yield characteristic, and energy dissipation characteristic includes: Align the calibration interface force sequence and the calibration relative displacement sequence in chronological order, and synchronously associate the visual feature vector with the interface force data and relative displacement data of the corresponding time step to form a joint response data sequence containing mechanical response information and visual feature information; The hysteresis loading cycle interval is identified based on the joint response data sequence. Specifically, the loading and unloading transition points are determined based on the directional changes of the relative displacement sequence, and the time interval between two adjacent loading transition points is taken as a hysteresis cycle interval, thereby obtaining multiple hysteresis cycle data segments. For each hysteresis cycle interval, the correction interface force sequence and correction relative displacement sequence within the corresponding time interval are paired point by point in chronological order. The energy dissipation of the energy-consuming component within the hysteresis cycle interval is obtained by accumulating the work done by the interface force during the relative displacement change process. In the calculation process, a hysteresis morphology weight based on visual feature vectors is introduced. Different weight coefficients are assigned to the yielding section, stable energy-consuming section and degradation section identified in the hysteresis curve to form a weighted energy accumulation process and obtain the energy dissipation of a single cycle. The irreversible response component is determined based on the energy dissipation of a single cycle and the relative displacement amplitude within the corresponding cycle interval, and the reversible response component is determined based on the near-linear relationship between the interface force and the relative displacement during the loading and unloading stages, thus completing the separation of the reversible and irreversible response components of the hysteresis response. The equivalent stiffness characteristic of the energy-consuming component is determined based on the reversible response component, the yield characteristic of the energy-consuming component is determined based on the irreversible response component, and the energy dissipation characteristic of the energy-consuming component is determined based on the single-cycle energy dissipation.

[0014] Optionally, the step of comparing the performance degradation index with a preset degradation threshold range to determine the performance degradation level of the energy-consuming component includes: Read the equivalent stiffness characteristic, yield characteristic and energy dissipation characteristic, and obtain the initial equivalent stiffness parameter, initial yield parameter and initial energy dissipation parameter of the energy dissipation component under the initial reference state. The stiffness retention degree is calculated based on the equivalent stiffness characteristic quantity and the initial equivalent stiffness parameter, the energy dissipation capacity retention degree is calculated based on the energy dissipation characteristic quantity and the initial energy dissipation parameter, and the yield capacity retention degree is calculated based on the yield characteristic quantity and the initial yield parameter, thus obtaining a multidimensional degradation characteristic quantity that reflects the current mechanical performance change of the energy dissipation component. The multidimensional degradation characteristics are statistically processed by time window. By calculating the moving average of stiffness retention, energy dissipation capacity retention and yield capacity retention in several consecutive hysteresis cycle intervals, a stable degradation characteristic sequence is obtained. A performance degradation index is constructed based on a stable degradation feature sequence. In the construction process, a degradation weight coefficient based on the evolution of hysteresis morphology is introduced. The degradation weight coefficient quantifies the degree of change of hysteresis morphology based on visual feature vectors. The degradation weight coefficient is applied to the degree of stiffness retention, the degree of energy dissipation capacity retention, and the degree of yield capacity retention to form a performance degradation index. The performance degradation index is compared step by step with the preset degradation threshold range. When the performance degradation index falls into the corresponding threshold range, the performance degradation level of the energy-consuming component is determined, and the performance degradation level is output as the online performance evaluation result of the energy-consuming component.

[0015] The beneficial effects of this invention are: This invention acquires structural response data from both ends of seismic-resistant energy-dissipating components in buildings and constructs an energy domain perspective matrix using interface degree-of-freedom mapping relationships. This enables the inversion calculation of interface forces on energy-dissipating components without the need for force sensors, thus obtaining stable and reliable force-displacement response information under actual service conditions. Compared to traditional methods relying on laboratory loading tests or direct force sensor measurements, this invention can directly utilize response data during structural operation to reconstruct mechanical behavior. This not only reduces the installation complexity and maintenance costs of testing equipment but also allows for the continuous acquisition of key mechanical information during long-term structural operation, improving the feasibility and engineering applicability of online testing of energy-dissipating components.

[0016] This invention establishes a dynamic energy closed-cone constraint interval and performs energy projection correction on the interface force sequence and relative displacement sequence, ensuring that the calculated force-displacement response always conforms to the structural energy evolution law. This effectively suppresses the influence of noise interference and sensor measurement errors on test results under weak vibration environments. By introducing a constraint mechanism based on energy characteristic parameters, abnormal data that does not conform to the energy closure relationship can be automatically corrected, resulting in higher physical consistency and stability of the reconstructed hysteresis response and improving the accuracy of identifying the mechanical behavior of energy-dissipating components under weak excitation conditions such as wind vibration and minor earthquakes.

[0017] This invention utilizes an improved MaxViT network to extract features from hysteresis response spectra and combines this with a method for separating the reversible and irreversible components of the hysteresis response to achieve comprehensive identification of the stiffness, yielding, and energy dissipation characteristics of energy-dissipating components. By constructing a comprehensive performance degradation index and comparing it with a degradation threshold range, the performance degradation level of energy-dissipating components can be accurately determined, enabling online assessment of the service status of these components. This allows for continuous tracking of performance changes in energy-dissipating components in complex engineering environments, providing reliable data support for seismic safety assessment and maintenance decisions for building structures. Attached Figure Description

[0018] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a flowchart of an online testing and evaluation method for the performance of energy-dissipating components used in seismic resistance of buildings, as proposed in this invention. Figure 2 This is a schematic diagram of the improved MaxViT network structure for an online testing and evaluation method of energy-dissipating components for seismic resistance of buildings, as proposed in this invention. Detailed Implementation

[0019] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0020] refer to Figure 1 and Figure 2 An online testing and evaluation method for the performance of energy-dissipating components used in seismic resistance of buildings, comprising: Response acquisition devices are installed at the structural nodes at both ends of the seismic energy dissipation components of the building to collect structural response data at both ends of the nodes. The structural response data is then preprocessed to obtain node response data. The relative displacement and relative velocity sequences between the nodes at both ends of the energy-consuming component are calculated based on the node response data. The interface degree of freedom mapping relationship is established and the energy domain perspective matrix is ​​constructed. The interface force of the energy-consuming component is inverted by solving the minimum energy residual and the initial interface force sequence is obtained. The energy characteristic parameters of the energy-consuming component are calculated based on the initial interface force sequence and relative displacement sequence. A dynamic energy closed cone constraint interval is established, and the initial interface force sequence and relative displacement sequence are corrected by energy projection to obtain the corrected interface force sequence and corrected relative displacement sequence. Force-displacement hysteresis frames, instantaneous power frames, and energy closure residual frames are generated based on the calibration interface force sequence and calibration relative displacement sequence. These are then merged into a multi-channel energy-time series map and input into an improved MaxViT network to obtain visual feature vectors. Based on the corrected interface force sequence, corrected relative displacement sequence, and visual feature vector, the reversible and irreversible response components of the hysteresis response of the energy dissipation component are separated to obtain the equivalent stiffness characteristic quantity, yield characteristic quantity, and energy dissipation characteristic quantity. The performance degradation index of the energy-consuming component is calculated based on the equivalent stiffness characteristic, yield characteristic and energy dissipation characteristic. The performance degradation index is compared with the preset degradation threshold range to determine the performance degradation level of the energy-consuming component.

[0021] In this embodiment, the structural response data includes nodal acceleration data, nodal displacement data, and nodal strain data. The nodal acceleration data is collected by acceleration sensors installed at both ends of the structural nodes of the energy-dissipating component, the nodal displacement data is obtained by displacement measuring devices installed at the structural nodes, and the nodal strain data is collected by strain sensors deployed in the connection area of ​​the energy-dissipating component.

[0022] In this embodiment, the preprocessing of the structural response data includes time synchronization calibration of the collected structural response data to unify the data time reference of each response acquisition device, detrending processing of the time-synchronized structural response data to eliminate low-frequency drift components in the data, and noise filtering processing of the detrending structural response data to remove high-frequency noise interference and obtain node response data.

[0023] In this embodiment, obtaining the initial interface force sequence includes: Displacement and velocity data of structural nodes at both ends of the energy-consuming component are extracted from the node response data. The relative displacement sequence is determined by the difference in displacement of the nodes at both ends, and the relative velocity sequence is determined by the difference in velocity of the nodes at both ends. Determine the connection interfaces between the two ends of the energy-consuming component and the structural substructure. Select a set of interface nodes that correspond one-to-one with the connection interfaces in the left and right substructures respectively. Encode the translational and rotational degrees of freedom of each interface node according to the preset degree of freedom order to obtain the left interface degree of freedom sequence and the right interface degree of freedom sequence. The interface degree-of-freedom mapping relationship is constructed based on the left-side interface degree-of-freedom sequence and the right-side interface degree-of-freedom sequence, specifically as follows: A one-to-one pairing of the left-side interface degree-of-freedom sequence with the right-side interface degree-of-freedom sequence is formed to create an interface pairing table; For each pair of paired degrees of freedom, a symbolic consistency marker is set to distinguish relative motions in the same direction and opposite directions; A closed-loop consistency check is performed on the interface pairing table to eliminate duplicate and missing pairings, thereby obtaining an interface degree-of-freedom mapping relationship that satisfies the completeness of the interface degrees of freedom on both sides. The equivalent inertial parameter sets corresponding to the interface node sets are extracted from the left and right substructures respectively. Based on the symbol consistency mark, the equivalent inertial parameter sets on both sides are aligned and fused in the interface degree of freedom dimension to form the interface energy weight matrix. The energy weight projection is applied to each degree of freedom component of the relative displacement sequence according to the interface energy weight matrix to obtain the energy domain perspective matrix. Based on the energy domain perspective matrix, relative displacement sequence, and relative velocity sequence, an energy residual evaluation quantity for the interface force is constructed. The interface force sequence of the energy-consuming component is obtained by minimizing the energy residual evaluation quantity, and this sequence serves as the initial interface force sequence. Specifically, the process of obtaining the interface force sequence of the energy-consuming component by minimizing the energy residual evaluation quantity is as follows: Energy projection calculations are performed on the relative displacement sequence based on the energy domain perspective matrix to obtain the structural equivalent energy response at the corresponding time. The equivalent energy response is then correlated with the relative velocity sequence to construct an energy residual evaluation quantity expression with the interface force as the unknown variable. The energy residual evaluation quantity is accumulated time-by-time within a continuous time window to form an overall energy residual function. The interface force sequence is used as the variable to be determined, and the energy residual function is gradually reduced by iteratively updating the interface force sequence. When the energy residual function meets the preset convergence condition during continuous iteration, the current iteration result is determined as the interface force sequence of the energy-consuming component, and the interface force sequence is output as the initial interface force sequence.

[0024] In this embodiment, obtaining the corrected interface force sequence and the corrected relative displacement sequence includes: Based on the initial interface force sequence and relative displacement sequence, the product of the interface force and the relative displacement change is calculated point by point in time sequence. The product results are accumulated step by step to obtain the input energy change sequence of the energy dissipating component at each sampling time. Based on the maximum displacement amplitude of the relative displacement sequence within the current time window and the equivalent stiffness characteristics of the energy-dissipating component, the maximum dissipable energy change sequence of the energy-dissipating component at each sampling time is calculated. Specifically, the calculation of the maximum dissipable energy change sequence of the energy-dissipating component at each sampling time is as follows: Amplitude analysis is performed on the relative displacement sequence within a preset time window to extract the maximum displacement amplitude within the current time window, and the displacement envelope curve is determined based on the change in the maximum displacement amplitude between adjacent time windows. Based on the displacement envelope curve and the equivalent stiffness characteristics of the energy-consuming component, the maximum elastic energy that the energy-consuming component can store under the corresponding displacement amplitude is calculated, and the maximum elastic energy is taken as the maximum energy that can be consumed at the corresponding sampling time. The maximum energy consumption values ​​calculated at consecutive sampling times are arranged in chronological order to form a sequence of maximum energy consumption changes over time. The irreversible energy dissipation sequence is calculated based on the input energy change sequence and the maximum dissipable energy change sequence. This irreversible energy dissipation sequence represents the unrecoverable energy dissipated by the energy-consuming component during loading and unloading processes through material yielding, friction, or damping. The calculation of the irreversible energy dissipation sequence is as follows: By aligning the input energy change sequence and the maximum consumable energy change sequence point by point at each sampling time, the input energy value and the maximum consumable energy value at the corresponding time are compared to determine the energy state interval corresponding to the energy-consuming component at that time. When the input energy value exceeds the recoverable energy range corresponding to the maximum expendable energy change sequence, the excess energy is determined as unrecoverable energy, and the increment value of the unrecoverable energy at the current sampling time is recorded. At the same time, the unrecoverable energy is cumulatively updated by combining the displacement change direction of the loading and unloading stages. The irreversible energy increments obtained at each sampling time are accumulated in chronological order to form an irreversible energy consumption change sequence that varies with time. A dynamic energy closed cone constraint interval is established based on the input energy change sequence, the maximum dissipable energy change sequence, and the irreversible energy dissipation change sequence, specifically: The lower boundary of energy is determined by the irreversible energy consumption change sequence, the upper boundary of energy is determined by the maximum energy consumption change sequence, and the time evolution trend of the input energy change sequence is used as the energy central axis. By progressively updating the upper boundary, lower boundary, and central axis of energy within a continuous time window, the three elements form a conical energy constraint region with a convergent trend in the time dimension, thus obtaining the dynamic energy closed cone constraint interval. The input energy change sequence is compared with the dynamic energy closed cone constraint interval at each time step. When the input energy change sequence is detected to exceed the dynamic energy closed cone constraint interval, the initial interface force sequence and relative displacement sequence are subjected to energy projection correction processing so that the corrected interface force sequence and relative displacement sequence fall back into the dynamic energy closed cone constraint interval, thus obtaining the corrected interface force sequence and the corrected relative displacement sequence.

[0025] In this embodiment, obtaining the visual feature vector includes: A force-displacement hysteresis frame sequence is generated based on the corrected interface force sequence and the corrected relative displacement sequence, and an instantaneous power frame sequence is generated based on the corrected interface force sequence and the corrected relative velocity sequence. At the same time, an energy closure residual frame sequence is generated based on the input energy change sequence and the irreversible energy dissipation change sequence. The force-displacement hysteresis frame sequence, instantaneous power frame sequence, and energy closure residual frame sequence are spatially aligned according to the same time index and superimposed in the channel dimension to form a multi-channel energy-time sequence, wherein each sequence frame contains hysteresis information channel, power information channel, and energy closure residual information channel. A multi-channel energy-temporal map sequence is input into an improved MaxViT network. The improved MaxViT network sequentially includes a map patch embedding layer, a first multi-axis attention block, an energy consistency enhancement layer, a second multi-axis attention block, a cross-frame temporal fusion layer, third to fifth multi-axis attention blocks, a global feature aggregation layer, and a feature output layer. In the map patch embedding layer, the multi-channel energy-temporal map is divided into fixed-size map patches. Feature mapping is performed on each map patch to obtain a patch feature sequence. All patch feature sequences are arranged in spatial order to form a feature vector sequence, where: The first multi-axis attention block is used to perform multi-axis attention calculation within a local window on the feature vector sequence, extracting local correlation features between each map patch in the single-frame multi-channel energy-time map. The local correlation features include local bending features of the hysteresis curve, concentrated distribution features of instantaneous power, and local fluctuation features of energy closure residual. During the calculation process, the feature vector sequence is grouped according to a preset window size to form multiple local patch windows. Then, attention calculation is performed along the row and column directions in each local patch window to obtain the correlation weights of each map patch in different spatial directions. Finally, the local patch features are weighted and fused according to the correlation weights to output the feature vector sequence after local morphological enhancement. In the energy consistency enhancement layer, energy channel weights are calculated on the feature vector sequence, and energy weight coefficients are generated based on the numerical values ​​of the energy closure residual channels. The feature vectors corresponding to the hysteresis information channel and the power information channel are then updated with weights to obtain the energy consistency enhancement feature sequence. The second multi-axis attention block performs multi-axis attention calculations across channels and across spaces on the energy consistency enhancement feature sequence to extract the coupling correlation features between multi-channel energy information. During the calculation process, the feature vector sequence is reorganized according to the channel indices of the hysteresis information channel, instantaneous power channel, and energy closure residual channel in the multi-channel energy-time sequence map. Then, attention calculations are performed along the spatial axis and channel axis in each channel group to obtain the correlation weights between different channel features. Subsequently, the feature vectors of the corresponding channels are weighted and fused according to the correlation weights to form the second-stage feature vector sequence with cross-channel coupling enhancement. In the cross-frame temporal fusion layer, the second-stage feature vector sequence is grouped according to the time order. In the cross-frame temporal fusion layer, the patch features of adjacent time frames are spliced ​​in chronological order, and the evolution features of hysteresis morphology over time are extracted through joint operation of local convolution and multi-axis attention. The obtained fused features are then input into the third to fifth multi-axis attention blocks for global feature extraction, and feature compression is performed through the global feature aggregation layer. Finally, a visual feature vector representing the hysteresis evolution state of the energy-consuming component is output in the feature output layer.

[0026] In this embodiment, obtaining the equivalent stiffness characteristic, yield characteristic, and energy dissipation characteristic includes: Align the calibration interface force sequence and the calibration relative displacement sequence in chronological order, and synchronously associate the visual feature vector with the interface force data and relative displacement data of the corresponding time step to form a joint response data sequence containing mechanical response information and visual feature information; The hysteresis loading cycle interval is identified based on the joint response data sequence. Specifically, the loading and unloading transition points are determined based on the directional changes of the relative displacement sequence, and the time interval between two adjacent loading transition points is taken as a hysteresis cycle interval, thereby obtaining multiple hysteresis cycle data segments. For each hysteresis cycle interval, the correction interface force sequence and correction relative displacement sequence within the corresponding time interval are paired point by point in chronological order. The energy dissipation of the energy-consuming component within the hysteresis cycle interval is obtained by accumulating the work done by the interface force during the relative displacement change process. In the calculation process, a hysteresis morphology weight based on visual feature vectors is introduced. Different weight coefficients are assigned to the yielding section, stable energy-consuming section and degradation section identified in the hysteresis curve to form a weighted energy accumulation process and obtain the energy dissipation of a single cycle. The irreversible response component is determined based on the energy dissipation of a single cycle and the relative displacement amplitude within the corresponding cycle interval, and the reversible response component is determined based on the near-linear relationship between the interface force and the relative displacement during the loading and unloading stages, thus completing the separation of the reversible and irreversible response components of the hysteresis response. The equivalent stiffness characteristic of the energy-dissipating component is determined based on the reversible response component, the yield characteristic of the energy-dissipating component is determined based on the irreversible response component, and the energy dissipation characteristic of the energy-dissipating component is determined based on the single-cycle energy dissipation, wherein: The determination of the equivalent stiffness characteristic quantity is as follows: Within each hysteresis cycle interval, extract the reversible response data segment in which the interface force and relative displacement change nearly linearly during the loading phase, and use the reversible response data segment as the elastic response interval. The ratio of the change in interface force to the corresponding change in relative displacement within the elastic response interval is calculated to obtain the instantaneous stiffness value of each sampling segment, and the instantaneous stiffness values ​​obtained within the same hysteresis cycle interval are statistically processed. The instantaneous stiffness values ​​obtained from multiple hysteresis cycle intervals are averaged to obtain the equivalent stiffness characteristic quantity that characterizes the overall elastic response capability of the energy dissipation component. The determination of the yield characteristic quantity is as follows: Within each hysteresis cycle interval, identify the response segment where the relationship between interface force and relative displacement changes from a near-linear change to a non-linear change, and determine the starting position of the response segment as the yield initiation segment; The interface force value at the corresponding moment is extracted in the yield initiation section, and the stability of the interface force value is screened by combining the irreversible response change trend of several adjacent sampling points. The yield initiation interface force values ​​identified in multiple hysteresis cycle intervals are statistically averaged to obtain the yield characteristic quantity of the energy dissipation component. The determination of energy dissipation characteristic quantities is as follows: Extract the single-cycle energy dissipation corresponding to each hysteresis cycle interval within the continuous monitoring time window, and form a single-cycle energy dissipation data set in chronological order; Stability screening is performed on the single-cycle energy dissipation dataset to remove abnormal energy consumption values ​​that deviate significantly from the overall trend, while retaining valid data that reflects the true energy consumption status. Statistical calculations were performed on the effective single-cycle energy dissipation data to obtain energy dissipation characteristic quantities that characterize the overall energy dissipation capacity of energy-consuming components.

[0027] In this embodiment, the step of comparing the performance degradation index with a preset degradation threshold range to determine the performance degradation level of the energy-consuming component includes: Read the equivalent stiffness characteristic, yield characteristic and energy dissipation characteristic, and obtain the initial equivalent stiffness parameter, initial yield parameter and initial energy dissipation parameter of the energy dissipation component under the initial reference state. The stiffness retention rate is calculated based on the equivalent stiffness characteristic quantity and the initial equivalent stiffness parameter; the energy dissipation capacity retention rate is calculated based on the energy dissipation characteristic quantity and the initial energy dissipation parameter; and the yield capacity retention rate is calculated based on the yield characteristic quantity and the initial yield parameter. This yields a multidimensional degradation characteristic quantity reflecting the current mechanical performance changes of the energy-dissipating component, where: The calculation of stiffness retention is as follows: Within the current monitoring time window, read the equivalent stiffness characteristic quantity calculated from the reversible response components, and obtain the initial equivalent stiffness parameters of the energy dissipation component under the initial reference state; The ratio of the equivalent stiffness characteristic quantity to the initial equivalent stiffness parameter is calculated to obtain the stiffness retention value, which characterizes the degree of change in the current structural stiffness relative to the initial state. The stiffness retention values ​​calculated from several consecutive hysteresis cycle intervals are statistically averaged to obtain the degree of stiffness retention that reflects the current overall stiffness state of the energy-dissipating component. The calculation of the degree of energy consumption capacity retention is as follows: Within the current monitoring time window, read the energy dissipation characteristic quantity obtained from the single-cycle energy dissipation statistics, and obtain the initial energy dissipation parameters corresponding to the energy-consuming component under the initial reference state; The ratio of the energy dissipation characteristic quantity to the initial energy dissipation parameter is calculated to obtain the energy retention value, which characterizes the degree of change of the current energy dissipation capacity relative to the initial state. The energy consumption retention values ​​obtained from several consecutive hysteresis loop intervals are statistically averaged to obtain the degree of energy consumption retention that reflects the overall energy dissipation capacity of the energy-consuming component. The calculation of the degree of yield strength retention is as follows: Within the current monitoring time window, read the yield characteristic quantity obtained from the irreversible response component analysis, and obtain the initial yield parameter corresponding to the energy dissipation component under the initial reference state; The yield characteristic quantity is calculated as a ratio to the initial yield parameter to obtain the yield retention value, which characterizes the degree of change of the current yield capacity relative to the initial state. The yield retention values ​​obtained from several consecutive hysteresis cycle intervals are statistically averaged to obtain the degree of yield retention that reflects the current overall yield capacity state of the energy-consuming component. The multidimensional degradation characteristics are statistically processed by time window. By calculating the moving average of stiffness retention, energy dissipation capacity retention and yield capacity retention in several consecutive hysteresis cycle intervals, a stable degradation characteristic sequence is obtained. A performance degradation index is constructed based on a stable degradation feature sequence. In the construction process, a degradation weight coefficient based on the evolution of hysteresis morphology is introduced. The degradation weight coefficient quantifies the degree of change of hysteresis morphology based on visual feature vectors. The degradation weight coefficient is applied to the degree of stiffness retention, the degree of energy dissipation capacity retention, and the degree of yield capacity retention to form a performance degradation index. The performance degradation index is compared step by step with the preset degradation threshold range. When the performance degradation index falls into the corresponding threshold range, the performance degradation level of the energy-consuming component is determined, and the performance degradation level is output as the online performance evaluation result of the energy-consuming component.

[0028] Example 1: To verify the feasibility of this invention in practice, it was applied to a high-rise office building in a coastal city that is currently in use. The building is approximately 96 meters high, has 24 floors, and its structural form is a steel frame-reinforced concrete core tube structure. To improve the energy dissipation capacity of the structure under seismic loading, yield-type metal energy-dissipating components were arranged between the 6th and 18th floors, with two sets of energy-dissipating components on each floor, totaling 26 sets. In this example, a set of yield-type steel plate energy-dissipating components on the east side of the 10th floor was selected as the monitoring object, and its mechanical properties under actual service conditions were tested online and its degradation was evaluated.

[0029] During the operation of the building structure, response acquisition devices are deployed at the steel beam connection nodes at both ends of the energy-dissipating components. These devices include triaxial accelerometers, micro-displacement sensors, and strain sensors to collect acceleration, displacement, and strain data of the structural nodes. The sampling frequency is set to 100Hz and synchronized with a unified clock module. The acquired structural response data is first preprocessed in the monitoring system, including time synchronization calibration, detrending processing, and bandpass filtering, to obtain stable node response data. A relative displacement sequence is calculated based on the displacement data of the nodes at both ends, and a relative velocity sequence is calculated based on the displacement changes at adjacent sampling times.

[0030] Subsequently, an interface degree-of-freedom mapping relationship was established based on the connection relationship between the energy-dissipating component and the substructures on both sides, and an energy domain perspective matrix was constructed. The interface forces of the energy-dissipating component were inverted through the minimum energy residual solution method, thus obtaining the initial interface force sequence. In actual building environments, due to factors such as wind load, equipment vibration, and environmental noise, some monitoring data show energy anomalies. Therefore, the energy characteristic parameters of the energy-dissipating component were calculated based on the initial interface force sequence and relative displacement sequence, and a dynamic energy closed cone constraint interval was established. Energy projection correction was performed on the data that did not meet the energy constraint conditions to obtain the corrected interface force sequence and corrected relative displacement sequence, so that the obtained force-displacement response conforms to the actual energy evolution law of the structure.

[0031] After obtaining the corrected interface force sequence and relative displacement sequence, force-displacement hysteresis frames are generated in chronological order, and instantaneous power frames are generated based on the interface force and relative velocity. Simultaneously, energy residual frames are generated by combining the energy closure residual. These three types of data are superimposed in the time dimension to form a multi-channel energy-time series map, which is then input into an improved MaxViT network for feature extraction. The improved MaxViT network performs block mapping of the map through a patch embedding layer, extracts the spatial features of the hysteresis response through a multi-axis attention structure, and identifies the evolution characteristics of the hysteresis morphology over time through a cross-frame temporal fusion layer to obtain a visual feature vector. Subsequently, based on the corrected interface force sequence, corrected relative displacement sequence, and visual feature vector, the reversible and irreversible response components of the hysteresis response are separated, thereby obtaining the equivalent stiffness feature, yield feature, and energy dissipation feature of the energy-consuming component.

[0032] In this embodiment, the energy-dissipating component was continuously monitored from March to September. During the monitoring period, the building experienced multiple wind-induced vibrations and two minor earthquakes, with a maximum structural acceleration of approximately 0.05g. The hysteretic response curve obtained by the method of this invention is stable and continuous, clearly reflecting the energy dissipation characteristics of the component during loading and unloading. Performance degradation indices were calculated based on the identified equivalent stiffness characteristics, yield characteristics, and single-cycle energy dissipation characteristics, and compared with degradation threshold ranges to evaluate the performance status of the energy-dissipating component.

[0033] Table 1. Identification results of performance parameters of energy-consuming components during the monitoring period.

[0034] As shown in Table 1, the main mechanical performance parameters of the energy-consuming component exhibited a slow and continuous changing trend during the continuous monitoring period from April 10th to September 20th. At the initial monitoring stage, the equivalent stiffness was 162 kN / mm, the yield strength was 86 kN, and the single-cycle energy dissipation was 3.6 kJ. Both the stiffness retention rate and energy dissipation retention rate were 1.00, indicating that the component was in its baseline operating state. At the two monitoring nodes of May 12th and June 18th, the equivalent stiffness decreased to 160 kN / mm and 158 kN / mm, respectively, and the yield strength decreased to 85 kN and 84 kN. The single-cycle energy dissipation remained around 3.5 kJ, corresponding to stiffness retention rates of 0.99 and 0.98, and energy dissipation retention rates of 0.97 and 0.96. This indicates that only slight changes occurred in the component's performance, and the overall performance remained stable.

[0035] During the mid-to-late monitoring phases, the structural parameters continued to show a slight downward trend. At 07:22, the equivalent stiffness was 156 kN / mm, the yield strength was 83 kN, and the single-cycle energy dissipation was 3.4 kJ, corresponding to a stiffness retention rate of 0.96 and an energy dissipation retention rate of 0.95. Subsequently, at 08:26 and 09:20, the equivalent stiffness decreased to 154 kN / mm and 152 kN / mm, respectively, the yield strength decreased to 82 kN and 81 kN, and the single-cycle energy dissipation was 3.3 kJ. The stiffness retention rates were 0.95 and 0.94, respectively, and the energy dissipation retention rates were 0.93 and 0.92. The component condition was determined to be slightly degraded, but still within the safe operating range.

[0036] The data changes over the entire monitoring period show that the equivalent stiffness of the energy-dissipating component decreased by about 6%, the yield strength by about 5%, and the energy dissipation capacity by about 8% within approximately six months. The degradation process was stable and did not exhibit any abrupt changes. The method of this invention can stably reconstruct the hysteretic response of the energy-dissipating component in a real-world operating environment and continuously track changes in key mechanical parameters, thereby promptly identifying trends in component performance changes and providing a reliable basis for seismic safety monitoring and maintenance decisions for building structures.

[0037] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for online testing and evaluation of performance of energy dissipation components for seismic resistance of buildings, characterized in that, include: Response acquisition devices are installed at the structural nodes at both ends of the seismic energy dissipation components of the building to collect structural response data at both ends of the nodes. The structural response data is then preprocessed to obtain node response data. The relative displacement and relative velocity sequences between the nodes at both ends of the energy-consuming component are calculated based on the node response data. The interface degree of freedom mapping relationship is established and the energy domain perspective matrix is ​​constructed. The interface force of the energy-consuming component is inverted by solving the minimum energy residual and the initial interface force sequence is obtained. The energy characteristic parameters of the energy-consuming component are calculated based on the initial interface force sequence and relative displacement sequence. A dynamic energy closed cone constraint interval is established, and the initial interface force sequence and relative displacement sequence are corrected by energy projection to obtain the corrected interface force sequence and corrected relative displacement sequence. Force-displacement hysteresis frames, instantaneous power frames, and energy closure residual frames are generated based on the calibration interface force sequence and calibration relative displacement sequence. These are then merged into a multi-channel energy-time series map and input into an improved MaxViT network to obtain visual feature vectors. Based on the corrected interface force sequence, corrected relative displacement sequence, and visual feature vector, the reversible and irreversible response components of the hysteresis response of the energy dissipation component are separated to obtain the equivalent stiffness characteristic quantity, yield characteristic quantity, and energy dissipation characteristic quantity. The performance degradation index of the energy-consuming component is calculated based on the equivalent stiffness characteristic, yield characteristic and energy dissipation characteristic. The performance degradation index is compared with the preset degradation threshold range to determine the performance degradation level of the energy-consuming component.

2. The performance online testing and evaluation method of the energy dissipation component for building seismic resistance according to claim 1, characterized in that, The structural response data includes nodal acceleration data, nodal displacement data, and nodal strain data. The nodal acceleration data is collected by acceleration sensors installed at both ends of the structural nodes of the energy-dissipating component, the nodal displacement data is obtained by displacement measuring devices installed at the structural nodes, and the nodal strain data is collected by strain sensors deployed in the connection area of ​​the energy-dissipating component.

3. The performance online testing and evaluation method of the energy dissipation component for building seismic resistance according to claim 1, characterized in that, The preprocessing of the structural response data includes time synchronization calibration of the acquired structural response data to unify the data time reference of each response acquisition device, detrending processing of the time-synchronized structural response data to eliminate low-frequency drift components in the data, and noise filtering processing of the detrending structural response data to remove high-frequency noise interference and obtain node response data.

4. The performance online testing and evaluation method of energy dissipation component for building seismic resistance according to claim 1, characterized in that, The process of obtaining the initial interface force sequence includes: Displacement and velocity data of structural nodes at both ends of the energy-consuming component are extracted from the node response data. The relative displacement sequence is determined by the difference in displacement of the nodes at both ends, and the relative velocity sequence is determined by the difference in velocity of the nodes at both ends. Determine the connection interfaces between the two ends of the energy-consuming component and the structural substructure. Select a set of interface nodes that correspond one-to-one with the connection interfaces in the left and right substructures respectively. Encode the translational and rotational degrees of freedom of each interface node according to the preset degree of freedom order to obtain the left interface degree of freedom sequence and the right interface degree of freedom sequence. The interface degree-of-freedom mapping relationship is constructed based on the left-side interface degree-of-freedom sequence and the right-side interface degree-of-freedom sequence, specifically as follows: A one-to-one pairing of the left-side interface degree-of-freedom sequence with the right-side interface degree-of-freedom sequence is formed to create an interface pairing table; For each pair of paired degrees of freedom, a symbolic consistency marker is set to distinguish relative motions in the same direction and opposite directions; A closed-loop consistency check is performed on the interface pairing table to eliminate duplicate and missing pairings, thereby obtaining an interface degree-of-freedom mapping relationship that satisfies the completeness of the interface degrees of freedom on both sides. The equivalent inertial parameter sets corresponding to the interface node sets are extracted from the left and right substructures respectively. Based on the symbol consistency mark, the equivalent inertial parameter sets on both sides are aligned and fused in the interface degree of freedom dimension to form the interface energy weight matrix. The energy weight projection is applied to each degree of freedom component of the relative displacement sequence according to the interface energy weight matrix to obtain the energy domain perspective matrix. Based on the energy domain perspective matrix, relative displacement sequence, and relative velocity sequence, an energy residual evaluation quantity of the interface force is constructed. The interface force sequence of the energy-consuming component is obtained by minimizing the energy residual evaluation quantity, which serves as the initial interface force sequence.

5. The online performance test and evaluation method of energy dissipation component for building seismic resistance according to claim 1, characterized in that, The process of obtaining the corrected interface force sequence and the corrected relative displacement sequence includes: Based on the initial interface force sequence and relative displacement sequence, the product of the interface force and the relative displacement change is calculated point by point in time sequence. The product results are accumulated step by step to obtain the input energy change sequence of the energy dissipating component at each sampling time. Based on the maximum displacement amplitude of the relative displacement sequence within the current time window and the equivalent stiffness characteristics of the energy-consuming component, calculate the maximum energy dissipation sequence of the energy-consuming component at each sampling time. The irreversible energy dissipation sequence is calculated based on the input energy change sequence and the maximum dissipable energy change sequence, wherein the irreversible energy dissipation sequence represents the unrecoverable energy generated by the energy dissipating component during loading and unloading through material yielding, friction or damping. A dynamic energy closed cone constraint interval is established based on the input energy change sequence, the maximum dissipable energy change sequence, and the irreversible energy dissipation change sequence, specifically: The lower boundary of energy is determined by the irreversible energy consumption change sequence, the upper boundary of energy is determined by the maximum energy consumption change sequence, and the time evolution trend of the input energy change sequence is used as the energy central axis. By progressively updating the upper boundary, lower boundary, and central axis of energy within a continuous time window, the three elements form a conical energy constraint region with a convergent trend in the time dimension, thus obtaining the dynamic energy closed cone constraint interval. The input energy change sequence is compared with the dynamic energy closed cone constraint interval at each time step. When the input energy change sequence is detected to exceed the dynamic energy closed cone constraint interval, the initial interface force sequence and relative displacement sequence are subjected to energy projection correction processing so that the corrected interface force sequence and relative displacement sequence fall back into the dynamic energy closed cone constraint interval, thus obtaining the corrected interface force sequence and the corrected relative displacement sequence.

6. The method for online testing and evaluation of the performance of energy-dissipating components for seismic resistance of buildings according to claim 1, characterized in that, The obtained visual feature vector includes: A force-displacement hysteresis frame sequence is generated based on the corrected interface force sequence and the corrected relative displacement sequence, and an instantaneous power frame sequence is generated based on the corrected interface force sequence and the corrected relative velocity sequence. At the same time, an energy closure residual frame sequence is generated based on the input energy change sequence and the irreversible energy dissipation change sequence. The force-displacement hysteresis frame sequence, instantaneous power frame sequence, and energy closure residual frame sequence are spatially aligned according to the same time index and superimposed in the channel dimension to form a multi-channel energy-time sequence, wherein each sequence frame contains hysteresis information channel, power information channel, and energy closure residual information channel. The multi-channel energy-temporal map sequence is input into the improved MaxViT network. The improved MaxViT network includes, in sequence, a map patch embedding layer, a first multi-axis attention block, an energy consistency enhancement layer, a second multi-axis attention block, a cross-frame temporal fusion layer, third to fifth multi-axis attention blocks, a global feature aggregation layer, and a feature output layer. In the map patch embedding layer, the multi-channel energy-temporal map is divided into map patches of fixed size. Feature mapping is performed on each map patch to obtain a patch feature sequence. All patch feature sequences are arranged in spatial order to form a feature vector sequence. In the energy consistency enhancement layer, energy channel weights are calculated on the feature vector sequence, and energy weight coefficients are generated based on the numerical values ​​of the energy closure residual channels. The feature vectors corresponding to the hysteresis information channel and the power information channel are then updated with weights to obtain the energy consistency enhancement feature sequence. In the cross-frame temporal fusion layer, the patch features of adjacent time frames are spliced ​​in chronological order, and the evolution features of hysteresis morphology over time are extracted through joint operation of local convolution and multi-axis attention. The obtained fused features are then input into the third to fifth multi-axis attention blocks for global feature extraction, and feature compression is performed through the global feature aggregation layer. Finally, a visual feature vector representing the hysteresis evolution state of the energy-consuming component is output in the feature output layer.

7. The method for online testing and evaluation of the performance of energy-dissipating components for seismic resistance of buildings according to claim 1, characterized in that, The obtained equivalent stiffness characteristic, yield characteristic, and energy dissipation characteristic include: Align the calibration interface force sequence and the calibration relative displacement sequence in chronological order, and synchronously associate the visual feature vector with the interface force data and relative displacement data of the corresponding time step to form a joint response data sequence containing mechanical response information and visual feature information; The hysteresis loading cycle interval is identified based on the joint response data sequence. Specifically, the loading and unloading transition points are determined based on the directional changes of the relative displacement sequence, and the time interval between two adjacent loading transition points is taken as a hysteresis cycle interval, thereby obtaining multiple hysteresis cycle data segments. For each hysteresis cycle interval, the correction interface force sequence and correction relative displacement sequence within the corresponding time interval are paired point by point in chronological order. The energy dissipation of the energy-consuming component within the hysteresis cycle interval is obtained by accumulating the work done by the interface force during the relative displacement change process. In the calculation process, a hysteresis morphology weight based on visual feature vectors is introduced. Different weight coefficients are assigned to the yielding section, stable energy-consuming section and degradation section identified in the hysteresis curve to form a weighted energy accumulation process and obtain the energy dissipation of a single cycle. The irreversible response component is determined based on the energy dissipation of a single cycle and the relative displacement amplitude within the corresponding cycle interval, and the reversible response component is determined based on the near-linear relationship between the interface force and the relative displacement during the loading and unloading stages, thus completing the separation of the reversible and irreversible response components of the hysteresis response. The equivalent stiffness characteristic of the energy-consuming component is determined based on the reversible response component, the yield characteristic of the energy-consuming component is determined based on the irreversible response component, and the energy dissipation characteristic of the energy-consuming component is determined based on the single-cycle energy dissipation.

8. The method for online testing and evaluation of the performance of energy-dissipating components for seismic resistance of buildings according to claim 1, characterized in that, The step of comparing performance degradation indicators with a preset degradation threshold range to determine the performance degradation level of energy-consuming components includes: Read the equivalent stiffness characteristic, yield characteristic and energy dissipation characteristic, and obtain the initial equivalent stiffness parameter, initial yield parameter and initial energy dissipation parameter of the energy dissipation component under the initial reference state. The stiffness retention degree is calculated based on the equivalent stiffness characteristic quantity and the initial equivalent stiffness parameter, the energy dissipation capacity retention degree is calculated based on the energy dissipation characteristic quantity and the initial energy dissipation parameter, and the yield capacity retention degree is calculated based on the yield characteristic quantity and the initial yield parameter, thus obtaining a multidimensional degradation characteristic quantity that reflects the current mechanical performance change of the energy dissipation component. The multidimensional degradation characteristics are statistically processed by time window. By calculating the moving average of stiffness retention, energy dissipation capacity retention and yield capacity retention in several consecutive hysteresis cycle intervals, a stable degradation characteristic sequence is obtained. A performance degradation index is constructed based on a stable degradation feature sequence. In the construction process, a degradation weight coefficient based on the evolution of hysteresis morphology is introduced. The degradation weight coefficient quantifies the degree of change of hysteresis morphology based on visual feature vectors. The degradation weight coefficient is applied to the degree of stiffness retention, the degree of energy dissipation capacity retention, and the degree of yield capacity retention to form a performance degradation index. The performance degradation index is compared step by step with the preset degradation threshold range. When the performance degradation index falls into the corresponding threshold range, the performance degradation level of the energy-consuming component is determined, and the performance degradation level is output as the online performance evaluation result of the energy-consuming component.

Images