Strong earthquake data fusion and engineering safety monitoring method and system

Abstract

The present application relates to the technical field of structural health monitoring, and particularly relates to a strong earthquake data fusion and engineering safety monitoring method and system. The method comprises the following steps: acquiring vibration response data streams of multiple source sensing nodes and performing time-frequency coupling feature analysis to extract a feature vector set representing structural dynamic characteristics; determining a weight distribution relationship according to signal quality and historical consistency and performing cross-source data fusion on multiple time scales to obtain a fusion feature vector reflecting the structural state; comparing the fusion feature vector with a structural damage discrimination standard to identify the damage location and degree and obtain a structural safety state evaluation result; generating a graded early warning instruction in combination with a time-varying constraint boundary representing the degradation process of the bearing capacity of the structure; triggering a monitoring response strategy and adaptively adjusting the weight distribution relationship by using execution state information, thereby improving the accuracy of engineering structural safety evaluation under strong earthquakes and the timeliness of early warning.

Description

Technical Field

[0001] This invention relates to the field of structural health monitoring technology, and in particular to a method and system for strong earthquake data fusion and engineering safety monitoring. Background Technology

[0002] In the field of structural safety monitoring, especially in structural response analysis under strong earthquakes, the conventional approach is to rely on a monitoring network composed of various sensors (such as accelerometers, strain gauges, and displacement gauges) to collect vibration response data of the structure during a strong earthquake. Existing technologies typically perform single-dimensional analysis of this data in the time or frequency domain, such as extracting statistical indicators like peak ground acceleration and Fourier spectra, or using modal parameter identification methods to obtain the structure's natural frequencies and damping ratios. Subsequently, the analysis results from single-source sensors or the results of simple arithmetic mean fusion are compared with pre-set thresholds to determine whether structural damage has occurred and to trigger an alarm. Some more complex methods introduce weighted fusion of data from multiple sensors, but the weights are usually set empirically or fixed based on sensor calibration accuracy and are not dynamically adjusted during the monitoring process. Early warning strategies often use a single threshold or fixed-level classification, lacking consideration for the degradation of structural bearing capacity over time.

[0003] The aforementioned conventional approach has significant drawbacks. Firstly, relying solely on single time-domain or frequency-domain features is insufficient to comprehensively capture the time-varying, non-stationary characteristics of strong vibration responses, easily overlooking time-frequency coupling information related to structural damage, leading to inaccurate feature extraction and consequently affecting the reliability of subsequent damage assessment. Secondly, fixed or empirical weights cannot adapt to signal quality changes caused by sensor aging, malfunctions, environmental interference, etc., and are also insufficient to reflect the consistency and reliability differences of historical data from different sensors. When the performance of individual sensors degrades, the fusion results will introduce bias, reducing the robustness of structural condition assessment. Furthermore, the early warning grading strategy lacks time-varying constraints on the structural performance degradation process, failing to achieve a refined match between graded early warnings and response measures, thus limiting the effectiveness of the monitoring system in practical engineering. Summary of the Invention

[0004] The present invention provides a method and system for strong earthquake data fusion and engineering safety monitoring, which can solve the problems in the prior art.

[0005] A first aspect of the present invention provides a method for strong earthquake data fusion and engineering safety monitoring, comprising:

[0006] The vibration response data stream from multiple source sensing nodes is acquired, and the vibration response data stream is subjected to time-frequency coupling feature analysis to extract a set of feature vectors characterizing the dynamic properties of the structure.

[0007] Based on the signal quality indicators and historical consistency indicators of each data source, a weight allocation relationship reflecting the reliability of the data source is determined. Based on the weight allocation relationship, the feature vector set is fused across multiple time scales to obtain a fused feature vector reflecting the structural state.

[0008] By comparing and analyzing the fused feature vector with the structural damage discrimination criteria, the damage location and degree of the engineering structure are identified, the structural safety status assessment result is obtained, and a graded early warning instruction is generated based on the structural safety status assessment result and the time-varying constraint boundary characterizing the degradation process of the structural bearing capacity.

[0009] Based on the hierarchical early warning command, the corresponding monitoring and response strategy is triggered and the execution status information is recorded. The execution status information is then used to adaptively adjust the weight allocation relationship.

[0010] Acquire vibration response data streams from multiple source sensing nodes, perform time-frequency coupled feature analysis on the vibration response data streams, and extract a set of feature vectors characterizing the dynamic properties of the structure, including:

[0011] The vibration response data stream is decomposed in the time domain to obtain multiple vibration sub-sequences within a time window, and each vibration sub-sequence is transformed in the frequency domain to obtain the corresponding spectral distribution characteristics.

[0012] Calculate the ratio of the energy amplitude of each frequency point in each of the aforementioned spectral distribution features to the average energy amplitude in adjacent frequency bands, and identify the frequency peak positions where the ratio exceeds a preset multiple relationship and their corresponding energy amplitudes as frequency domain feature components;

[0013] The time-domain statistical characteristics of each vibrational subsequence are extracted as time-domain feature components, and the time-domain statistical characteristics include peak response value and response duration;

[0014] Based on the temporal position of each time window in the entire vibration response, time sequence identifiers are assigned to the corresponding frequency domain feature components and time domain feature components.

[0015] The frequency domain feature components are coupled and combined with the time domain feature components according to the time sequence identifier to form a feature vector that reflects the frequency component change characteristics of the vibration response during the time evolution process, thus obtaining the feature vector set.

[0016] Based on the signal quality indicators and historical consistency indicators of each data source, a weight allocation relationship reflecting the reliability of the data source is determined. Then, based on this weight allocation relationship, the feature vector set is fused across multiple time scales to obtain a fused feature vector reflecting the structural state, including:

[0017] For each data source, the signal quality index is calculated for the current monitoring period. The signal quality index is used to quantify the signal-to-noise ratio characteristics and data integrity characteristics of the vibration response data stream.

[0018] The deviation values ​​between the feature vectors output by each data source and the feature vectors output by other data sources during the historical monitoring period are extracted, and the deviation values ​​are statistically analyzed to obtain a historical consistency index reflecting the stability of the data source output.

[0019] The weighting coefficients for each data source are determined based on the relative relationship between the signal quality index and the historical consistency index for each data source in all data sources, thus obtaining the weighting distribution relationship.

[0020] The feature vector set is divided into a short-time response feature subset and a long-time response feature subset according to the time scale. The short-time response feature subset and the long-time response feature subset are weighted and fused according to the weight allocation relationship to obtain a short-time scale fused feature vector and a long-time scale fused feature vector.

[0021] The fused feature vector is obtained by concatenating the short-timescale fused feature vector with the long-timescale fused feature vector.

[0022] The weighting coefficients for each data source are determined based on the relative relationship between the signal quality index and the historical consistency index among all data sources, thus obtaining the weighting allocation relationship, including:

[0023] A two-dimensional evaluation space for signal quality indicators and historical consistency indicators is determined, and the signal quality indicators and historical consistency indicators corresponding to each data source are mapped to evaluation points in the two-dimensional evaluation space.

[0024] Calculate the distance measure between each evaluation point and the reference position in the two-dimensional evaluation space that characterizes the target performance to obtain the performance deviation corresponding to each data source;

[0025] The performance stratification boundary is determined based on the distribution characteristics of each performance deviation in all data sources, and each data source is divided into different trust levels according to the positional relationship between the performance deviation corresponding to each data source and the performance stratification boundary.

[0026] Determine the nonlinear mapping relationship between the performance deviation and the weighting coefficient, wherein the performance deviation is inversely proportional to the rate of change of the corresponding weighting coefficient in the nonlinear mapping relationship;

[0027] The performance deviation of each data source is converted into a corresponding weight coefficient through the nonlinear mapping relationship. The weight coefficient of the corresponding data source is adjusted and corrected according to the trust level of each data source to obtain the corrected weight coefficient, which constitutes the weight allocation relationship.

[0028] By comparing and analyzing the fused feature vector with the structural damage discrimination criteria, the location and degree of damage to the engineering structure are identified, resulting in a structural safety status assessment. Based on this assessment and the time-varying constraint boundary characterizing the structural bearing capacity degradation process, graded early warning instructions are generated, including:

[0029] Identify the feature components in the fused feature vector that exceed the corresponding limits of the structural damage discrimination criteria, and extract the structural monitoring locations corresponding to the feature components as the damage locations;

[0030] Calculate the excess range between the feature component corresponding to each damage location and the structural damage discrimination criterion, and quantify the damage degree of each damage location based on the excess range to obtain the structural safety status assessment result;

[0031] The bearing capacity degradation trajectory is determined based on historical monitoring data of the engineering structure, and the time-varying constraint boundary is determined at the current monitoring time based on the bearing capacity degradation trajectory.

[0032] The degree of damage at each damage location in the structural safety status assessment results is matched and compared with the time-varying constraint boundary to determine the warning level corresponding to each damage location;

[0033] The overall warning level of the engineering structure is determined based on the distribution of the warning levels corresponding to each of the damage locations across all damage locations, and a graded warning instruction containing the overall warning level and the warning levels corresponding to each of the damage locations is generated.

[0034] The bearing capacity degradation trajectory is determined based on historical monitoring data of the engineering structure, and the time-varying constraint boundary is determined at the current monitoring time based on the bearing capacity degradation trajectory, including:

[0035] Structural response feature values ​​reflecting the structural bearing capacity status at multiple historical moments are extracted from the historical monitoring data of the engineering structure, and a time-series correlation is established between the structural response feature values ​​at each historical moment and the corresponding monitoring moment.

[0036] Calculate the temporal decay trend of the structural response characteristic values ​​at each historical moment as a function of the monitoring time, and fit the bearing capacity degradation trajectory characterizing the structural bearing capacity as a function of time based on the temporal decay trend.

[0037] Multiple degradation stage boundary points characterizing different load-bearing capacity levels are marked on the load-bearing capacity degradation trajectory, and the constraint threshold corresponding to each degradation stage boundary point is determined according to the load-bearing capacity level corresponding to each degradation stage boundary point.

[0038] The degradation stage at the current monitoring time is determined based on the position of the current monitoring time on the degradation trajectory, and the constraint threshold corresponding to the degradation stage is extracted as the time-varying constraint boundary at the current monitoring time.

[0039] Based on the tiered early warning command, a corresponding monitoring and response strategy is triggered and execution status information is recorded. The execution status information is then used to adaptively adjust the weight allocation relationship, including:

[0040] The monitoring response strategy to be executed is determined based on the overall warning level and the warning level corresponding to each damage location contained in the graded warning instruction. The monitoring response strategy includes a data acquisition frequency adjustment scheme and a data transmission priority configuration scheme for each data source.

[0041] The monitoring and response strategy is executed, and the data acquisition response latency and data transmission integrity information of each data source are recorded in real time during the execution process to obtain the execution status information;

[0042] The performance indicators of each data source during the execution of the monitoring and response strategy are extracted from the execution status information. The performance indicators include the response latency deviation and transmission integrity deviation of each data source.

[0043] Calculate the performance adjustment coefficient of each data source based on the performance indicators corresponding to each data source, and use the performance adjustment coefficient to adjust the weight coefficient of each data source in the weight allocation relationship to obtain the adjusted weight allocation relationship.

[0044] A second aspect of the present invention provides a strong earthquake data fusion and engineering safety monitoring system, comprising:

[0045] The time-frequency feature extraction unit is used to acquire vibration response data streams from multi-source sensing nodes, perform time-frequency coupled feature analysis on the vibration response data streams, and extract a set of feature vectors characterizing the dynamic properties of the structure.

[0046] The weighted fusion unit is used to determine the weight allocation relationship reflecting the credibility of the data source based on the signal quality index and historical consistency index of each data source, and to perform cross-source data fusion on the feature vector set at multiple time scales based on the weight allocation relationship to obtain a fused feature vector reflecting the structural state.

[0047] The structural safety early warning unit is used to compare and analyze the fused feature vector with the structural damage discrimination standard, identify the damage location and damage degree of the engineering structure, obtain the structural safety status assessment result, and generate graded early warning instructions based on the structural safety status assessment result and the time-varying constraint boundary characterizing the structural bearing capacity degradation process.

[0048] The weight adaptive unit is used to trigger the corresponding monitoring response strategy based on the hierarchical early warning instruction and record the execution status information, and to adaptively adjust the weight allocation relationship using the execution status information.

[0049] A third aspect of the present invention provides an electronic device, comprising:

[0050] processor;

[0051] Memory used to store processor-executable instructions;

[0052] The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0053] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0054] The vibration response data from the multi-source sensing nodes of this invention, after time-frequency coupled feature analysis, can effectively separate vibration information from different frequency bands and extract the essential features reflecting the dynamic characteristics of the structure, avoiding the limitations of single data source or single frequency band analysis. Based on the weight allocation relationship determined by signal quality indicators and historical consistency indicators, cross-source data fusion is performed at multiple time scales, significantly reducing the negative impact of noise interference and abnormal data on structural state identification, and improving the robustness and reliability of the fusion results. Comparative analysis of the fused feature vector and structural damage discrimination criteria enables accurate identification of damage location and damage degree, providing a reliable quantitative basis for structural safety status assessment.

[0055] The tiered early warning command of this invention combines structural safety status assessment results with time-varying constraint boundaries characterizing the load-bearing capacity degradation process. It can dynamically classify early warning levels according to the severity of damage, avoiding false alarms or missed alarms caused by traditional single-threshold early warning systems. The tiered early warning mechanism enables the monitoring system to adopt differentiated response strategies for different risk levels, ensuring rapid response in emergencies while avoiding resource waste caused by over-response. The recording and feedback of execution status information provides closed-loop data support for verifying the effectiveness of the early warning command.

[0056] The adaptive adjustment mechanism of the weight allocation relationship in this invention utilizes the execution status information of the monitoring response strategy for dynamic optimization, enabling the data fusion process to continuously adapt to changes in structural state and environmental conditions. This adaptive capability ensures the rationality of weight allocation during long-term monitoring, avoiding a decrease in fusion accuracy due to sensor performance drift or environmental changes. By continuously correcting the weight relationship, the system can maintain high sensitivity to changes in structural state, providing continuous and effective technical support for engineering safety monitoring. Attached Figure Description

[0057] Figure 1 This is a flowchart illustrating the strong earthquake data fusion and engineering safety monitoring method according to an embodiment of the present invention;

[0058] Figure 2 This is a schematic diagram illustrating the process of determining the set of feature vectors characterizing the dynamic properties of a structure according to an embodiment of the present invention. Detailed Implementation

[0059] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0060] The technical solution of the present invention will be described in detail below with reference to specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0061] Figure 1 This is a flowchart illustrating the strong earthquake data fusion and engineering safety monitoring method according to an embodiment of the present invention. Figure 1 As shown, this strong earthquake data fusion and engineering safety monitoring method includes:

[0062] The vibration response data stream from multiple source sensing nodes is acquired, and the vibration response data stream is subjected to time-frequency coupling feature analysis to extract a set of feature vectors characterizing the dynamic properties of the structure.

[0063] Based on the signal quality indicators and historical consistency indicators of each data source, a weight allocation relationship reflecting the reliability of the data source is determined. Based on the weight allocation relationship, the feature vector set is fused across multiple time scales to obtain a fused feature vector reflecting the structural state.

[0064] By comparing and analyzing the fused feature vector with the structural damage discrimination criteria, the damage location and degree of the engineering structure are identified, the structural safety status assessment result is obtained, and a graded early warning instruction is generated based on the structural safety status assessment result and the time-varying constraint boundary characterizing the degradation process of the structural bearing capacity.

[0065] Based on the hierarchical early warning command, the corresponding monitoring and response strategy is triggered and the execution status information is recorded. The execution status information is then used to adaptively adjust the weight allocation relationship.

[0066] Figure 2 This is a schematic diagram illustrating the process of determining the set of feature vectors characterizing the dynamic properties of a structure according to an embodiment of the present invention. Figure 2 As shown, a vibration response data stream from multiple source sensing nodes is acquired, and time-frequency coupled feature analysis is performed on the vibration response data stream to extract a set of feature vectors characterizing the dynamic properties of the structure, including:

[0067] The vibration response data stream is decomposed in the time domain to obtain multiple vibration sub-sequences within a time window, and each vibration sub-sequence is transformed in the frequency domain to obtain the corresponding spectral distribution characteristics.

[0068] Calculate the ratio of the energy amplitude of each frequency point in each of the aforementioned spectral distribution features to the average energy amplitude in adjacent frequency bands, and identify the frequency peak positions where the ratio exceeds a preset multiple relationship and their corresponding energy amplitudes as frequency domain feature components;

[0069] The time-domain statistical characteristics of each vibrational subsequence are extracted as time-domain feature components, and the time-domain statistical characteristics include peak response value and response duration;

[0070] Based on the temporal position of each time window in the entire vibration response, time sequence identifiers are assigned to the corresponding frequency domain feature components and time domain feature components.

[0071] The frequency domain feature components are coupled and combined with the time domain feature components according to the time sequence identifier to form a feature vector that reflects the frequency component change characteristics of the vibration response during the time evolution process, thus obtaining the feature vector set.

[0072] After acquiring the vibration response data stream output from the multi-source sensing nodes, the continuously acquired raw signals need to be decomposed in the time domain. Specifically, according to the preset time window length and sliding step size, the complete vibration response data stream is divided into several continuous vibration subsequences. A certain proportion of overlapping areas can be set between adjacent time windows to avoid the loss of transient features due to window boundary effects. The length of the time window needs to take into account the periodic characteristics of the low-order vibration modes of the structure, and should generally cover several times the period corresponding to the lowest frequency in the target frequency band to ensure sufficient frequency resolution for subsequent frequency domain analysis. For strong earthquake scenarios, the duration of the vibration response varies from several seconds to tens of seconds. The time window length can be dynamically configured according to the actual sampling rate and the target analysis frequency band to adapt to the differences in signal characteristics under different magnitude events.

[0073] Performing frequency domain transformation on each vibration subsequence separately, using Fast Fourier Transform or Short-Time Fourier Transform to map the time-domain signal to the frequency domain, yields the spectral distribution characteristics corresponding to each time window. These spectral distribution characteristics describe the distribution of vibration energy at different frequency points within that time period, reflecting the main vibration modes excited by the structure at that moment. For non-stationary strong earthquake signals, the spectral characteristics within each time window differ significantly depending on the stage of vibration development. Therefore, performing independent frequency domain transformation window by window is a necessary means to capture time-varying frequency characteristics.

[0074] After obtaining the spectral distribution characteristics of each time window, it is necessary to identify physically significant frequency peak locations. For each frequency point in the spectrum, the ratio of its energy amplitude to the average energy amplitude of all frequency points in the adjacent frequency band is calculated. Let the energy amplitude at a certain frequency point be... The average energy amplitude in its adjacent frequency band is The local salience at that frequency point satisfy .when Exceeding the preset multiple threshold At that time, the frequency point is identified as the frequency peak position, and its corresponding energy amplitude is... The frequency bands are recorded as components of the frequency domain features. The range of adjacent frequency bands can be adaptively adjusted according to the density of structural modes. For structures with small modal spacing, the width of adjacent frequency bands should be appropriately narrowed to improve the resolution of frequency peak identification; for structures with sparse modes, the frequency band range can be appropriately widened to enhance noise immunity. A preset multiple threshold is used. The value of directly affects the sensitivity and false detection rate of frequency peak identification. It can be calibrated by combining the statistical laws of historical vibration data, or it can be dynamically adjusted according to the overall noise level of the current signal.

[0075] Meanwhile, temporal statistical characteristics are extracted from each vibration subsequence as temporal feature components. These characteristics include two core indices: peak response magnitude and response duration. The peak response magnitude reflects the maximum amplitude of vibration acceleration, velocity, or displacement within the time window, serving as a direct quantitative indicator of the impact intensity experienced by the structure. The response duration describes the cumulative time during which the vibration response exceeds a certain amplitude threshold, reflecting the time span during which the structure endures strong excitation. The peak response magnitude can be obtained by taking the absolute value of the vibration subsequence and then finding its maximum value. The response duration is obtained by counting the number of sampling points in the vibration subsequence whose amplitude continuously exceeds a set threshold, and then multiplying this number by the sampling time interval. These two types of temporal feature components describe the intensity and duration characteristics of vibration events from different dimensions, effectively supplementing the limitations of frequency domain features in expressing the time dimension.

[0076] After extracting the frequency and time domain feature components, it is necessary to assign time-series identifiers to the features within each time window to record their temporal position information throughout the vibration response. The time-series identifier can be the absolute timestamp of the start time of each time window or a relative time offset relative to the trigger time of the vibration event. The introduction of time-series identifiers enables the extracted feature components within different time windows to be comparable and traceable in the time dimension, providing a unified time coordinate reference for subsequent feature fusion across time scales. In scenarios where multiple sensing nodes work collaboratively, there may be clock deviations in the data acquisition of each node. Therefore, before assigning time-series identifiers, the time axes of each node need to be synchronously calibrated to ensure that the time-series identifiers from different data sources are within a unified time reference framework.

[0077] Based on the time sequence identifier, the frequency domain feature components and time domain feature components extracted within the same time window are coupled and combined. The frequency domain feature components contain the peak positions of each frequency identified within the time window and their corresponding energy amplitudes, while the time domain feature components contain the peak response value and response duration. These two types of feature components are then concatenated according to a unified vector arrangement rule to form a feature vector that comprehensively describes the vibration response characteristics within the time window. This feature vector carries both modal excitation information in the frequency domain and amplitude intensity information in the time domain, enabling a more comprehensive characterization of the structure's dynamic response state at that moment.

[0078] The above processing procedure is repeated for all time windows, generating feature vectors for each time window sequentially. These feature vectors are then arranged according to the order of their time sequence identifiers, ultimately forming a set of feature vectors reflecting the frequency component changes in the vibration response over time. This set, with time as its axis, comprehensively records the evolution trajectory of the structural dynamic characteristics from the occurrence of the vibration event to its decay. In strong earthquake scenarios, structural damage often occurs at specific vibration stages. Abrupt changes in frequency components or abnormal shifts in energy distribution are important signals of damage occurrence. The feature vector set provides a systematic quantitative expression of this evolution process, offering a reliable input foundation for subsequent cross-source data fusion and structural damage identification. The dimension of the feature vector set is determined by the number of time windows and the dimension of individual feature vectors. The dimension of an individual feature vector depends on the number of identified frequency peaks and the number of time-domain feature components. The number of frequency peaks varies under different vibration events. Therefore, in practical engineering applications, a maximum upper limit for the number of peaks can be preset, and insufficient positions can be filled with zero values ​​to ensure that the feature vector set has a uniform dimensional structure, facilitating subsequent matrix operations.

[0079] Based on the signal quality indicators and historical consistency indicators of each data source, a weight allocation relationship reflecting the reliability of the data source is determined. Then, based on this weight allocation relationship, the feature vector set is fused across multiple time scales to obtain a fused feature vector reflecting the structural state, including:

[0080] For each data source, the signal quality index is calculated for the current monitoring period. The signal quality index is used to quantify the signal-to-noise ratio characteristics and data integrity characteristics of the vibration response data stream.

[0081] The deviation values ​​between the feature vectors output by each data source and the feature vectors output by other data sources during the historical monitoring period are extracted, and the deviation values ​​are statistically analyzed to obtain a historical consistency index reflecting the stability of the data source output.

[0082] The weighting coefficients for each data source are determined based on the relative relationship between the signal quality index and the historical consistency index for each data source in all data sources, thus obtaining the weighting distribution relationship.

[0083] The feature vector set is divided into a short-time response feature subset and a long-time response feature subset according to the time scale. The short-time response feature subset and the long-time response feature subset are weighted and fused according to the weight allocation relationship to obtain a short-time scale fused feature vector and a long-time scale fused feature vector.

[0084] The fused feature vector is obtained by concatenating the short-timescale fused feature vector with the long-timescale fused feature vector.

[0085] When calculating the signal quality index of each data source within the current monitoring period, a quantitative evaluation of the vibration response data stream quality is required from two dimensions. The signal-to-noise ratio (SNR) is calculated based on the ratio of effective signal power to noise power in the original vibration response data stream. By performing a short-time Fourier transform on the acquired time-domain signal, the signal is decomposed into frequency bands, identifying the effective frequency band energy and background noise energy related to the structural vibration frequency, thus obtaining the estimated SNR value for the data source within the current monitoring period. The data integrity characteristic is obtained by statistically analyzing the ratio of the actual number of valid samples acquired within the current monitoring period to the theoretically required number of samples. If a sensor node loses data packets due to communication interruption or hardware failure, the integrity ratio will be lower than normal. The estimated SNR value and the data integrity ratio are then normalized according to a preset combination rule to obtain the comprehensive signal quality index of the data source within the current monitoring period. ,in Index the data source. The range of values ​​is normalized to The higher the value in the range, the more reliable the signal quality of the data source within the current time period.

[0086] The extraction process of historical consistency indicators requires a systematic statistical analysis of the deviations between the feature vectors output by various data sources during the historical monitoring period. The feature vector sequence output by the data source over several historical monitoring periods is extracted, and the Euclidean distance between this feature vector and the mean of the feature vectors output by all other data sources in the same period is calculated for each period. This distance is denoted as the deviation value. ,in This is a historical time period index. It includes all deviation values ​​within the historical time period. Statistical analysis is performed to calculate the mean and standard deviation, comprehensively reflecting the degree and amplitude of systematic deviation and fluctuation between this data source and other data sources during historical monitoring. A larger mean deviation indicates a systematic shift in the data source, while a larger standard deviation indicates poor output stability. The mean and standard deviation are linearly combined according to preset weights and then normalized to obtain the historical consistency index. , The values ​​are also normalized to The higher the value in the range, the better the consistency and stability of the historical output of the data source with other data sources.

[0087] Based on the signal quality indicators of each data source Consistency index with history When determining the weighting coefficients for each data source based on their relative relationships across all data sources, a normalized weighting strategy is employed. A combined score is set for the signal quality metric and the historical consistency metric. ,in and These are the preset proportional coefficients for signal quality indicators and historical consistency indicators, respectively. The values ​​of both can be configured according to the importance placed on real-time signal quality and historical stability in the actual engineering scenario. After obtaining all... Overall score of each data source Then, the data is normalized to obtain the weight coefficients corresponding to each data source. This establishes a complete weighting distribution relationship. This weighting distribution relationship ensures that data sources with high signal quality and good historical consistency contribute a larger proportion in the subsequent fusion process, while the impact of data sources with poor quality or large historical deviations on the fusion result is correspondingly compressed, thereby improving the overall reliability of the fused feature vector.

[0088] When dividing the feature vector set into short-time response feature subsets and long-time response feature subsets according to the time scale, a preset time window length is used as the dividing criterion. The short-time response feature subset corresponds to a shorter time window, typically covering vibration responses within a few seconds to tens of seconds, and can capture the transient dynamic response characteristics of the structure during strong earthquakes, providing high temporal resolution for sudden damage events. The long-time response feature subset corresponds to vibration responses within a window of several minutes or even longer, reflecting the changing trends of the steady-state dynamic characteristics of the structure under continuous excitation, and is of significant value for identifying slowly accumulating damage. For the short-time response feature subset, the feature vectors extracted from each data source within this time window are weighted linearly combined according to a weight allocation relationship, i.e., for the... Feature vectors extracted from data sources within a short time window Multiply by the corresponding weighting coefficient After summing, the short-time scale fused feature vector is obtained. For the long-time response feature subset, the same weighted fusion strategy is used to extract feature vectors from each data source within the long-time window. Weighted summation is performed to obtain the long-term scale fusion feature vector. .

[0089] Short-time scale fusion feature vector Feature vector fused with long-term scale In a physical sense, they complement each other. The former focuses on characterizing the instantaneous dynamic response of a structure under strong seismic excitation, including peak acceleration response and transient frequency shift; the latter focuses on characterizing the overall dynamic characteristics of the structure over the entire monitoring period, including natural frequency drift trends and damping ratio changes. Directly combining the two... and Concatenate the data in dimensional order to form a fused feature vector containing information from multiple time scales. This fused feature vector carries both short-term transient response information and long-term steady-state characteristic information, providing more comprehensive and robust input features for subsequent structural damage identification and safety status assessment, effectively avoiding the problems of missed detections or misjudgments that may occur when relying solely on features of a single time scale.

[0090] In actual engineering deployments, the aforementioned weight allocation relationship is not statically fixed, but rather adaptively adjusted based on feedback from execution status information. When the signal quality of a certain data source continuously declines over multiple monitoring periods, or its historical consistency indicators deteriorate significantly, the corresponding weight coefficients will change. The weight will be dynamically reduced, thereby minimizing the negative impact of this data source on the fusion result. Through this dynamic weight adjustment mechanism, the entire cross-source data fusion process can maintain stable fusion quality in complex monitoring environments where the working status of sensor nodes changes, ensuring that the fused feature vector can always accurately reflect the true state of the engineering structure.

[0091] The weighting coefficients for each data source are determined based on the relative relationship between the signal quality index and the historical consistency index among all data sources, thus obtaining the weighting allocation relationship, including:

[0092] A two-dimensional evaluation space for signal quality indicators and historical consistency indicators is determined, and the signal quality indicators and historical consistency indicators corresponding to each data source are mapped to evaluation points in the two-dimensional evaluation space.

[0093] Calculate the distance measure between each evaluation point and the reference position in the two-dimensional evaluation space that characterizes the target performance to obtain the performance deviation corresponding to each data source;

[0094] The performance stratification boundary is determined based on the distribution characteristics of each performance deviation in all data sources, and each data source is divided into different trust levels according to the positional relationship between the performance deviation corresponding to each data source and the performance stratification boundary.

[0095] Determine the nonlinear mapping relationship between the performance deviation and the weighting coefficient, wherein the performance deviation is inversely proportional to the rate of change of the corresponding weighting coefficient in the nonlinear mapping relationship;

[0096] The performance deviation of each data source is converted into a corresponding weight coefficient through the nonlinear mapping relationship. The weight coefficient of the corresponding data source is adjusted and corrected according to the trust level of each data source to obtain the corrected weight coefficient, which constitutes the weight allocation relationship.

[0097] In determining the weight allocation relationship, a two-dimensional evaluation space needs to be constructed that can comprehensively reflect the performance status of each data source. This two-dimensional evaluation space uses signal quality indicators as the horizontal axis and historical consistency indicators as the vertical axis, with the values ​​of both axes normalized to a certain range. This allows for the comparison of metrics with different dimensions. For each data source, its signal quality metric and historical consistency metric are used as coordinate components and mapped to an evaluation point in this two-dimensional evaluation space. This two-dimensional mapping method can simultaneously characterize the performance characteristics of the data source in two dimensions: real-time acquisition quality and long-term stability, avoiding the bias that can occur when evaluating a single metric.

[0098] In the two-dimensional evaluation space, a reference position representing the target performance is defined. This reference position corresponds to the state point where both the signal quality index and the historical consistency index have reached ideal levels, i.e., the coordinates are... The ideal corner point. For the first There are 1 data source, whose evaluation point coordinates in the two-dimensional evaluation space are denoted as . ,in The normalized signal quality index This is a normalized historical consistency metric. The Euclidean distance between this evaluation point and the reference location is calculated to obtain the performance deviation corresponding to this data source. The calculation method is as follows . The range of values ​​is The smaller the value, the closer the overall performance of the data source is to the ideal state, and the higher the reliability; the larger the value, the greater the deviation of the data source in terms of signal quality or historical consistency, and it should be assigned a lower weight.

[0099] After obtaining the performance deviations of all data sources, the performance stratification boundaries are determined based on the distribution characteristics of these deviation values ​​across all data sources. Specifically, the performance deviations of all data sources are... Sort the data and calculate its statistical distribution characteristics, including the mean. and standard deviation .by and As two hierarchical boundary thresholds, the data source is divided into three trust levels: performance deviation less than The data source is classified as high trust level, with a performance deviation between [missing information]. and The data sources between them are classified into a trusted level, and the performance deviation is greater than [missing information]. The data sources are classified as low trust levels. This dynamic stratification method based on distributional statistical characteristics can adapt to changes in the overall performance level of data sources under different monitoring scenarios, avoiding the problem of stratification failure due to scenario differences when using fixed thresholds.

[0100] After determining the trustworthiness level of each data source, a non-linear mapping relationship needs to be established between performance deviation and weighting coefficients. The core constraint of this non-linear mapping relationship is that the greater the performance deviation, the smaller the rate of change of the corresponding weighting coefficient; that is, in regions with large deviations, the weighting coefficients are less sensitive to deviation. The physical significance of this design is that for data sources with extremely poor performance, even if their deviation increases further, their weights will not infinitely approach zero, thus avoiding the risk of information loss from completely discarding any data source. Conversely, for data sources with better performance, the weighting coefficients are more sensitive to changes in deviation, enabling a finer distinction between the relative differences between high-quality data sources. The non-linear mapping function that satisfies the above constraints... An exponential decay form can be used, for the first... Each data source has its initial weight coefficient. pass The calculation yielded, where This is a positive real parameter used to control the decay rate. Since the derivative of the exponential function decreases as the independent variable increases, it satisfies the requirement that the rate of change is inversely proportional to the degree of deviation. The value can be set according to the requirements for data source differentiation in the actual engineering scenario, and the typical value range is [range missing]. .

[0101] After obtaining the initial weight coefficients of each data source Then, the initial weighting coefficients are adjusted and corrected according to the trust level of each data source. Specifically, for data sources with high trust levels, the initial weighting coefficients are multiplied by a correction factor greater than 1. To further enhance the contribution of high-quality data sources; for data sources with medium confidence levels, the initial weight coefficients remain unchanged, i.e., the correction factor. For data sources with low trust levels, a correction factor less than 1 is applied to the initial weighting coefficients. This is to suppress the interference of low-quality data sources on the fusion results. The corrected weighting coefficients are denoted as... ,satisfy ,in Indicates the first The trust level of each data source This is the correction factor for the corresponding level. A typical value for the correction factor is... , The specific values ​​can be calibrated based on the tolerance level of each level of data source in engineering practice.

[0102] After completing the grade correction, the corrected weight coefficients of all data sources are normalized to ensure that the sum of the weight coefficients of all data sources is 1, thus forming the final weight allocation relationship. The normalized weight coefficients are... By The weights are obtained by dividing by the sum of the corrected weight coefficients from all data sources. The final weight allocation is stored in vector form for use in subsequent cross-source data fusion steps.

[0103] In real-world engineering monitoring scenarios, when a strong earthquake occurs, some sensor nodes may experience a sharp drop in signal quality due to the intense impact, leading to performance deviations. The weighting will increase significantly. Through nonlinear mapping and a level correction mechanism, the weight coefficient of the node will be automatically reduced, while the weight of nodes with stable performance will be increased accordingly, thus ensuring that the reliability of the fusion result is not affected by a few abnormal nodes. At the same time, since the layer boundary is dynamically calculated based on the distribution characteristics of all current data sources, even if the overall data quality level changes systematically, the layering result can still accurately reflect the relative performance differences between the data sources, and has strong adaptability.

[0104] By comparing and analyzing the fused feature vector with the structural damage discrimination criteria, the location and degree of damage to the engineering structure are identified, resulting in a structural safety status assessment. Based on this assessment and the time-varying constraint boundary characterizing the structural bearing capacity degradation process, graded early warning instructions are generated, including:

[0105] Identify the feature components in the fused feature vector that exceed the corresponding limits of the structural damage discrimination criteria, and extract the structural monitoring locations corresponding to the feature components as the damage locations;

[0106] Calculate the excess range between the feature component corresponding to each damage location and the structural damage discrimination criterion, and quantify the damage degree of each damage location based on the excess range to obtain the structural safety status assessment result;

[0107] The bearing capacity degradation trajectory is determined based on historical monitoring data of the engineering structure, and the time-varying constraint boundary is determined at the current monitoring time based on the bearing capacity degradation trajectory.

[0108] The degree of damage at each damage location in the structural safety status assessment results is matched and compared with the time-varying constraint boundary to determine the warning level corresponding to each damage location;

[0109] The overall warning level of the engineering structure is determined based on the distribution of the warning levels corresponding to each of the damage locations across all damage locations, and a graded warning instruction containing the overall warning level and the warning levels corresponding to each of the damage locations is generated.

[0110] After obtaining the fused feature vector, it needs to be compared component by component with a pre-established structural damage discrimination standard to identify the locations of abnormal responses in the engineering structure. The structural damage discrimination standard typically consists of a baseline feature vector of the engineering structure in a healthy state and its allowable fluctuation range. Each feature component corresponds to an upper and lower limit, which comprehensively considers sensor measurement errors, normal drift caused by changes in environmental temperature and humidity, and the nonlinear response characteristics of the structure itself. Each component in the fused feature vector is compared one by one with its corresponding limit; any component exceeding the limit is marked as an abnormal component. Each abnormal feature component has a mapping relationship with a specific structural monitoring location. This mapping relationship is pre-established during the system deployment phase based on the sensor node layout scheme and structural modal analysis results, and stored in the form of a location-component lookup table. By querying the location-component lookup table, the structural monitoring locations corresponding to all abnormal feature components are extracted, and these locations are confirmed as the set of damage locations at the current moment.

[0111] After determining the damage location, it is necessary to further quantify the damage severity at each location. For each damage location, the difference between the measured value of its corresponding abnormal characteristic component and the threshold value of the structural damage discrimination standard is taken as the exceedance amplitude. The larger the exceedance amplitude, the further the structural response at that location deviates from the normal state, and the higher the damage severity. To ensure the comparability of damage severity across locations, the exceedance amplitude is normalized according to the range of the corresponding threshold of each component, resulting in a dimensionless damage severity index. The calculation method is as follows: The first... The difference between the measured value and the threshold value of the characteristic component corresponding to each damage location, divided by the total range of allowable fluctuations of that component under healthy conditions, is calculated as follows: For the number index of the damage location, The value of is a real number greater than 0, with a larger value indicating more severe damage. When multiple anomalous characteristic components exist at a certain damage location, the maximum value of the normalized exceedance of each component is taken as the final damage severity index for that location, reflecting the most unfavorable structural response state. The numbers, coordinate information, and corresponding damage severity indices of all damage locations are summarized to form the structural safety status assessment result.

[0112] Determining the load-bearing capacity degradation trajectory relies on historical monitoring data of the engineering structure. Key structural performance parameters, including natural frequency drift, damping ratio change, and modal deviation, are extracted from the historical database at various historical moments since monitoring began. Regression analysis is then used to fit the trends of these parameters over time, establishing a load-bearing capacity degradation trajectory model. This model uses monitoring time as the independent variable and a comprehensive load-bearing capacity index as the input. As the dependent variable, it describes the gradual degradation of the structure's load-bearing capacity throughout its service life. At the current monitoring time... ,Will Substituting the values ​​into the degradation trajectory model, the predicted carrying capacity at the current moment is calculated. In conjunction with the safety margin specified in the structural design code, the time-varying constraint boundary at the current moment is determined. The time-varying constraint boundary contains multiple levels of thresholds, each corresponding to a structural state of different severity. These thresholds dynamically tighten as the structure's service time increases, reflecting the physical nature of the structure's safety margin gradually decreasing with the aging process.

[0113] The damage severity index at each damage location in the structural safety status assessment results. Match and compare with the thresholds of each level in the time-varying constraint boundary, based on The threshold range within which the damage falls determines the corresponding warning level for that location. The warning level system is divided into four levels: when... When the value is below the first-level threshold, it is considered a normal state and no warning is triggered; when When the threshold exceeds the first-level threshold but not the second-level threshold, a yellow alert is issued, indicating a minor structural anomaly, requiring increased monitoring frequency; when When the threshold exceeds the secondary threshold but not the tertiary threshold, an orange alert is issued, indicating significant structural damage and requiring the initiation of a specialized testing procedure; when... When the threshold value exceeds level three, a red alert is issued, indicating a serious risk of structural damage and requiring immediate emergency response measures. This is because the threshold values ​​for each level in the time-varying constraint boundary vary with... The dynamic adjustment means that the same damage level index corresponds to different warning levels in the early and late stages of structural service, which ensures that the warning judgment can accurately reflect the true safety margin of the structure in the current aging state.

[0114] After determining the individual warning levels for each damage location, it is necessary to comprehensively consider the warning status of all damage locations to determine the overall warning level of the engineering structure. The determination of the overall warning level adopts a weighted distribution strategy: the proportion of each warning level among all damage locations is statistically analyzed, assuming the total number of damage locations is 1. The number of locations under orange alert is The number of locations under red alert is Calculate the percentage of red alerts and the proportion of orange alerts .when When the concentration of red light exceeds the preset threshold, or when any single damage location reaches a red alert level, the overall alert level is directly determined to be red; when Not exceeding the red concentration threshold but When the concentration threshold of orange is exceeded, the overall warning level is determined to be orange; in other cases, the overall warning level is the warning level that appears most frequently among all damage locations. This distribution weighting strategy can capture the decisive impact of local severe damage on overall safety, and also reflect the potential threat of large-scale minor damage to the overall structural performance under the cumulative effect, avoiding the defect of ignoring the damage distribution characteristics when simply taking the highest level.

[0115] The generated tiered early warning instructions contain complete structured information: the overall early warning level, the number and coordinates of each damage location, the early warning level corresponding to each damage location, and the damage severity index for each damage location. The threshold values ​​for each level of the time-varying constraint boundary at the current moment, and the monitoring time. The timestamps are used to encapsulate this information in a standardized data format before transmitting it to the downstream monitoring and response strategy module. This ensures that subsequent response actions can take precise measures based on the specific location and extent of damage, while also providing complete data support for archiving historical early warning records and subsequent trend analysis.

[0116] The bearing capacity degradation trajectory is determined based on historical monitoring data of the engineering structure, and the time-varying constraint boundary is determined at the current monitoring time based on the bearing capacity degradation trajectory, including:

[0117] Structural response feature values ​​reflecting the structural bearing capacity status at multiple historical moments are extracted from the historical monitoring data of the engineering structure, and a time-series correlation is established between the structural response feature values ​​at each historical moment and the corresponding monitoring moment.

[0118] Calculate the temporal decay trend of the structural response characteristic values ​​at each historical moment as a function of the monitoring time, and fit the bearing capacity degradation trajectory characterizing the structural bearing capacity as a function of time based on the temporal decay trend.

[0119] Multiple degradation stage boundary points characterizing different load-bearing capacity levels are marked on the load-bearing capacity degradation trajectory, and the constraint threshold corresponding to each degradation stage boundary point is determined according to the load-bearing capacity level corresponding to each degradation stage boundary point.

[0120] The degradation stage at the current monitoring time is determined based on the position of the current monitoring time on the degradation trajectory, and the constraint threshold corresponding to the degradation stage is extracted as the time-varying constraint boundary at the current monitoring time.

[0121] Continuous monitoring records covering the entire lifecycle or typical service stages of the structure are retrieved from the historical monitoring database of the engineering structure. For each historical moment, structural response feature values ​​that directly reflect the structural bearing capacity state are extracted. These feature values ​​typically include the structure's fundamental frequency drift, modal damping ratio change, peak strain at key sections, and displacement response amplitude. Their physical significance lies in quantifying the stiffness degradation and energy dissipation capacity of the structure at each historical moment from different perspectives. After extraction, a time-series correlation is established between the structural response feature values ​​of each historical moment and the corresponding monitoring moment, forming a time-series dataset with time as the horizontal axis and structural response feature values ​​as the vertical axis. This time-series dataset is the basic input for subsequent fitting of the bearing capacity degradation trajectory, and its completeness and accuracy directly affect the fitting quality of the degradation trajectory.

[0122] After establishing the time-series dataset, the temporal decay trend of the structural response characteristic values ​​at each historical moment is calculated as a function of the monitoring time. Specifically, the time-series dataset is smoothed to eliminate the interference of short-term fluctuation noise on trend judgment. Subsequently, least squares fitting or piecewise polynomial fitting methods are used to perform curve fitting on the structural response characteristic value sequence to obtain the bearing capacity degradation trajectory characterizing the structural bearing capacity over time. Let the historical time sequence be... The corresponding structural response eigenvalue sequence is Then the trajectory of bearing capacity degradation can be expressed as a fitting function. ,in For historical moments, This represents the estimated characteristic value of the corresponding structural bearing capacity. The fitting function should accurately capture the monotonic decreasing trend or phased decay characteristics of the structural bearing capacity throughout its service life. After a structure experiences multiple strong earthquakes, the degradation trajectory will exhibit a complex pattern of alternating step-like decline and steady decay. In this case, a segmented fitting strategy can be adopted, establishing local fitting models for different stages, and then piecing together the segments into a complete degradation trajectory through continuity constraints.

[0123] After establishing the load-bearing capacity degradation trajectory, multiple degradation stage boundary points characterizing different load-bearing capacity levels are marked along the trajectory. The determination of these degradation stage boundary points is based on factors including the ultimate limit state of load-bearing capacity specified in structural design codes, the engineering safety judgment criteria for structures under different degrees of damage, and the critical damage threshold determined by the material properties of the structure itself. For example, the moments on the degradation trajectory corresponding to the structural load-bearing capacity decreasing to 90%, 75%, 60%, and 45% of the initial load-bearing capacity can be used as degradation stage boundary points. These four boundary points divide the structural service life into four intervals: slight degradation, moderate degradation, significant degradation, and severe degradation. Let the first... The monitoring time corresponding to each degradation stage boundary point is The corresponding load-bearing capacity level is Based on the carrying capacity level corresponding to the boundary point of each degradation stage, the constraint threshold corresponding to each degradation stage can be determined respectively. Constraint threshold The physical meaning is: when the structure is in the first position At each degradation stage, the maximum allowable deviation of each damage-sensitive feature in the fused feature vector is determined. Exceeding this threshold triggers a warning command of the corresponding level. The constraint thresholds for different degradation stages tighten as the structural bearing capacity decreases. That is, structures in more severe degradation stages have a lower tolerance for feature deviations, thereby achieving a dynamic adjustment effect where the warning sensitivity automatically increases with the structural health status.

[0124] Based on the current monitoring time To determine the current stage of degradation at the monitoring moment, the location on the load-bearing capacity degradation trajectory is determined by: Substitute into the fitting function This yields the estimated carrying capacity at the current moment. Then Bearing capacity level at each degradation stage boundary Compare and determine Which degradation interval it falls into, thus determining the current degradation stage number. Extract the first The constraint thresholds corresponding to each degradation stage This serves as the time-varying constraint boundary for the current monitoring moment. This time-varying constraint boundary will be directly used in the subsequent generation of graded early warning commands, serving as a dynamic baseline for determining whether the structural safety status assessment result triggers an early warning.

[0125] In practical engineering applications, the fitting accuracy of the load-bearing capacity degradation trajectory is significantly affected by the sampling density and data quality of historical monitoring data. When there are missing segments in the historical monitoring data due to sensor failure or communication interruption, it is necessary to perform interpolation to complete the missing data before fitting. Common methods include linear interpolation, spline interpolation, and spatial interpolation based on data from adjacent sensor nodes. For engineering structures that have been in service for a long time, a large number of continuous observation records have been accumulated in the historical monitoring database. In this case, a sliding window mechanism can be used to select only the historical data within the time window most recent to the current monitoring time to participate in the dynamic update of the degradation trajectory. This ensures that the degradation trajectory can reflect the recent state change trend of the structure in a timely manner and avoids excessive interference from early data in the judgment of the current degradation trend.

[0126] The determination of degradation stage boundary points needs to be differentiated based on structural type and usage function. For infrastructure engineering structures such as bridges and dams, due to their high requirements for load-bearing capacity margin, the setting of degradation stage boundary points should be relatively conservative, and the corresponding constraint thresholds should be tightened accordingly. For general industrial plants or temporary structures, the setting standards for degradation stage boundary points can be appropriately relaxed based on their design service life and importance level. By combining time-varying constraint boundaries with structural type, usage function, and current degradation stage, personalized dynamic early warning benchmark management for different engineering scenarios can be achieved, significantly improving the adaptability and early warning accuracy of strong earthquake monitoring systems in complex engineering environments.

[0127] Based on the tiered early warning command, a corresponding monitoring and response strategy is triggered and execution status information is recorded. The execution status information is then used to adaptively adjust the weight allocation relationship, including:

[0128] The monitoring response strategy to be executed is determined based on the overall warning level and the warning level corresponding to each damage location contained in the graded warning instruction. The monitoring response strategy includes a data acquisition frequency adjustment scheme and a data transmission priority configuration scheme for each data source.

[0129] The monitoring and response strategy is executed, and the data acquisition response latency and data transmission integrity information of each data source are recorded in real time during the execution process to obtain the execution status information;

[0130] The performance indicators of each data source during the execution of the monitoring and response strategy are extracted from the execution status information. The performance indicators include the response latency deviation and transmission integrity deviation of each data source.

[0131] Calculate the performance adjustment coefficient of each data source based on the performance indicators corresponding to each data source, and use the performance adjustment coefficient to adjust the weight coefficient of each data source in the weight allocation relationship to obtain the adjusted weight allocation relationship.

[0132] After triggering a tiered early warning command, the monitoring response strategy to be executed within the current monitoring cycle is comprehensively determined based on the overall early warning level information carried in the command and the local early warning level information corresponding to each damage location. The overall early warning level reflects the urgency of the overall safety status of the engineering structure, while the local early warning level of each damage location corresponds to the dangerous situation of a specific component or area. When the overall early warning level is low, all data sources maintain the normal acquisition frequency; when the overall early warning level rises to medium or high, the acquisition frequency of data sources corresponding to sensor nodes near the damage location is increased, while the acquisition frequency of data sources far from the damage area is appropriately reduced, in order to concentrate resources to ensure the data quality of key monitoring points. The data transmission priority configuration scheme is based on the early warning level of each damage location, with higher early warning levels corresponding to higher data source transmission priority, ensuring that the most valuable monitoring data is transmitted first when network bandwidth is limited.

[0133] During the execution of the aforementioned monitoring and response strategy, the operational status of each data source is recorded in real time. Specifically, for each data source, the time difference between the moment it actually begins collecting data at the new frequency after receiving the frequency adjustment instruction and the moment the instruction is issued is recorded as the data acquisition response latency of that data source. Simultaneously, the ratio between the number of data packets successfully transmitted by that data source within the specified transmission window and the total number of data packets that should be transmitted is calculated as data transmission integrity information. These two types of information together constitute the execution status information, fully reflecting the actual operational performance of each data source during strategy execution. The recording granularity of the execution status information is consistent with the execution time window of the monitoring and response strategy, ensuring sufficient time correspondence for subsequent performance analysis.

[0134] When extracting performance metrics from the execution status information, the difference between the measured response latency of each data source and the baseline response latency of that type of sensor node under normal operating conditions is used as the response latency deviation, and the difference between the measured transmission integrity of each data source and the preset transmission integrity target value is used as the transmission integrity deviation. Let the data source... The corresponding response delay deviation is The transmission integrity deviation is ,but A positive value indicates that the data source is experiencing a response delay. A negative value indicates that packet loss or delay occurred during the transmission of the data source. These two types of deviations together describe the degree to which the data source deviates from the expected performance standard during actual execution, providing a quantitative basis for subsequent weight adjustments.

[0135] Calculate the performance adjustment coefficient for each data source based on the response delay deviation and transmission integrity deviation. Let the data source be... The corresponding performance adjustment factor is Its calculation comprehensively considers the weighted impact of response delay deviation and transmission integrity deviation. A larger response delay deviation indicates a weaker response capability of the data source under emergency conditions, and the corresponding performance adjustment coefficient should be reduced accordingly. Similarly, a larger absolute value of the transmission integrity deviation indicates poorer data integrity under high-load transmission conditions, and the corresponding performance adjustment coefficient should also be reduced. Specifically, let the normalized value of the response delay deviation be... The normalized value of the transmission integrity deviation is The normalization operation uses the maximum absolute value of the deviation across all data sources as a benchmark, mapping each deviation to... Range. Performance adjustment factor. By combining the two types of normalized bias, when and When both are close to zero, A value close to 1 indicates that the data source's performance meets expectations, and the weights will not be adjusted; when the deviation between the two types is large, A value significantly less than 1 indicates that the weight of this data source in the fusion process needs to be reduced.

[0136] When adjusting the weight coefficients of each data source in the weight allocation relationship using performance adjustment factors, the intermediate weight value before adjustment is obtained by multiplying the current weight coefficient by the corresponding performance adjustment factor. Then, the intermediate weight values ​​of all data sources are normalized to ensure that the sum of the adjusted weight coefficients is still equal to 1. Let the data sources before adjustment... The corresponding weighting coefficient is The adjusted intermediate weight value is After normalization, the adjusted weight coefficients are obtained. The calculation formula is:

[0137] ;

[0138] in, To sum and traverse the indices, Total number of data sources. Adjusted weighting coefficients. The weight coefficients of the corresponding data sources in the original weight allocation relationship will be replaced to form an adjusted weight allocation relationship, which will be used for cross-source data fusion calculation in subsequent monitoring periods.

[0139] To prevent drastic fluctuations in the weighting distribution caused by excessively large single adjustments, the performance adjustment coefficient is adjusted. Set upper and lower limit constraints. When the calculated... When it exceeds the preset upper limit, it is truncated to the upper limit value; when When the data falls below a preset lower limit, it is truncated to that lower limit to ensure that the change in the weight coefficient of any single data source in a single adjustment does not exceed a preset proportion. Furthermore, for data sources that experience complete disconnection or interruption of data acquisition during the execution of the monitoring response strategy, their performance adjustment coefficient is directly set to a preset minimum value, reducing their weight coefficient to near zero. This effectively shields the impact of the data source in the fusion calculation until it returns to normal operation and accumulates sufficient performance records in subsequent monitoring cycles, at which point its weight coefficient is gradually restored.

[0140] The adjusted weight allocation relationship officially takes effect at the start of the next monitoring cycle and is jointly updated with the weight results calculated based on signal quality indicators and historical consistency indicators. The joint update mechanism uses a moving average approach, mixing the current adjustment result based on execution status information with the historical weight allocation relationship in a certain proportion. This avoids excessive influence of short-term performance fluctuations on the weight allocation relationship, ensuring the stability and robustness of the entire adaptive weight adjustment process. Through this closed-loop adaptive adjustment mechanism, the weight allocation relationship can continuously track the actual performance of each data source under different warning states, ensuring that the calculation of the fused feature vector always prioritizes data sources with stable performance and timely response, thereby improving the reliability of engineering safety monitoring results during strong earthquake events.

[0141] A second aspect of the present invention provides a strong earthquake data fusion and engineering safety monitoring system, comprising:

[0142] The time-frequency feature extraction unit is used to acquire vibration response data streams from multi-source sensing nodes, perform time-frequency coupled feature analysis on the vibration response data streams, and extract a set of feature vectors characterizing the dynamic properties of the structure.

[0143] The weighted fusion unit is used to determine the weight allocation relationship reflecting the credibility of the data source based on the signal quality index and historical consistency index of each data source, and to perform cross-source data fusion on the feature vector set at multiple time scales based on the weight allocation relationship to obtain a fused feature vector reflecting the structural state.

[0144] The structural safety early warning unit is used to compare and analyze the fused feature vector with the structural damage discrimination standard, identify the damage location and damage degree of the engineering structure, obtain the structural safety status assessment result, and generate graded early warning instructions based on the structural safety status assessment result and the time-varying constraint boundary characterizing the structural bearing capacity degradation process.

[0145] The weight adaptive unit is used to trigger the corresponding monitoring response strategy based on the hierarchical early warning instruction and record the execution status information, and to adaptively adjust the weight allocation relationship using the execution status information.

[0146] A third aspect of the present invention provides an electronic device, comprising:

[0147] processor;

[0148] Memory used to store processor-executable instructions;

[0149] The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0150] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0151] This invention can be a method, apparatus, system, and / or computer program product. The computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for performing various aspects of the invention.

[0152] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for strong earthquake data fusion and engineering safety monitoring, characterized in that, include: The vibration response data stream from multiple source sensing nodes is acquired, and the vibration response data stream is subjected to time-frequency coupling feature analysis to extract a set of feature vectors characterizing the dynamic properties of the structure. Based on the signal quality indicators and historical consistency indicators of each data source, a weight allocation relationship reflecting the reliability of the data source is determined. Based on the weight allocation relationship, the feature vector set is fused across multiple time scales to obtain a fused feature vector reflecting the structural state. By comparing and analyzing the fused feature vector with the structural damage discrimination criteria, the damage location and degree of the engineering structure are identified, the structural safety status assessment result is obtained, and a graded early warning instruction is generated based on the structural safety status assessment result and the time-varying constraint boundary characterizing the degradation process of the structural bearing capacity. Based on the hierarchical early warning command, the corresponding monitoring and response strategy is triggered and the execution status information is recorded. The execution status information is then used to adaptively adjust the weight allocation relationship.

2. The method according to claim 1, characterized in that, Acquire vibration response data streams from multiple source sensing nodes, perform time-frequency coupled feature analysis on the vibration response data streams, and extract a set of feature vectors characterizing the dynamic properties of the structure, including: The vibration response data stream is decomposed in the time domain to obtain multiple vibration sub-sequences within a time window, and each vibration sub-sequence is transformed in the frequency domain to obtain the corresponding spectral distribution characteristics. Calculate the ratio of the energy amplitude of each frequency point in each of the aforementioned spectral distribution features to the average energy amplitude in adjacent frequency bands, and identify the frequency peak positions where the ratio exceeds a preset multiple relationship and their corresponding energy amplitudes as frequency domain feature components; The time-domain statistical characteristics of each vibrational subsequence are extracted as time-domain feature components, and the time-domain statistical characteristics include peak response value and response duration; Based on the temporal position of each time window in the entire vibration response, time sequence identifiers are assigned to the corresponding frequency domain feature components and time domain feature components. The frequency domain feature components are coupled and combined with the time domain feature components according to the time sequence identifier to form a feature vector that reflects the frequency component change characteristics of the vibration response during the time evolution process, thus obtaining the feature vector set.

3. The method according to claim 1, characterized in that, Based on the signal quality indicators and historical consistency indicators of each data source, a weight allocation relationship reflecting the reliability of the data source is determined. Then, based on this weight allocation relationship, the feature vector set is fused across multiple time scales to obtain a fused feature vector reflecting the structural state, including: For each data source, the signal quality index is calculated for the current monitoring period. The signal quality index is used to quantify the signal-to-noise ratio characteristics and data integrity characteristics of the vibration response data stream. The deviation values ​​between the feature vectors output by each data source and the feature vectors output by other data sources during the historical monitoring period are extracted, and the deviation values ​​are statistically analyzed to obtain a historical consistency index reflecting the stability of the data source output. The weighting coefficients for each data source are determined based on the relative relationship between the signal quality index and the historical consistency index for each data source in all data sources, thus obtaining the weighting distribution relationship. The feature vector set is divided into a short-time response feature subset and a long-time response feature subset according to the time scale. The short-time response feature subset and the long-time response feature subset are weighted and fused according to the weight allocation relationship to obtain a short-time scale fused feature vector and a long-time scale fused feature vector. The fused feature vector is obtained by concatenating the short-timescale fused feature vector with the long-timescale fused feature vector.

4. The method according to claim 3, characterized in that, The weighting coefficients for each data source are determined based on the relative relationship between the signal quality index and the historical consistency index among all data sources, thus obtaining the weighting allocation relationship, including: A two-dimensional evaluation space for signal quality indicators and historical consistency indicators is determined, and the signal quality indicators and historical consistency indicators corresponding to each data source are mapped to evaluation points in the two-dimensional evaluation space. Calculate the distance measure between each evaluation point and the reference position in the two-dimensional evaluation space that characterizes the target performance to obtain the performance deviation corresponding to each data source; The performance stratification boundary is determined based on the distribution characteristics of each performance deviation in all data sources, and each data source is divided into different trust levels according to the positional relationship between the performance deviation corresponding to each data source and the performance stratification boundary. Determine the nonlinear mapping relationship between the performance deviation and the weighting coefficient, wherein the performance deviation is inversely proportional to the rate of change of the corresponding weighting coefficient in the nonlinear mapping relationship; The performance deviation of each data source is converted into a corresponding weight coefficient through the nonlinear mapping relationship. The weight coefficient of the corresponding data source is adjusted and corrected according to the trust level of each data source to obtain the corrected weight coefficient, which constitutes the weight allocation relationship.

5. The method according to claim 1, characterized in that, By comparing and analyzing the fused feature vector with the structural damage discrimination criteria, the location and degree of damage to the engineering structure are identified, resulting in a structural safety status assessment. Based on this assessment and the time-varying constraint boundary characterizing the structural bearing capacity degradation process, graded early warning instructions are generated, including: Identify the feature components in the fused feature vector that exceed the corresponding limits of the structural damage discrimination criteria, and extract the structural monitoring locations corresponding to the feature components as the damage locations; Calculate the excess range between the feature component corresponding to each damage location and the structural damage discrimination criterion, and quantify the damage degree of each damage location based on the excess range to obtain the structural safety status assessment result; The bearing capacity degradation trajectory is determined based on historical monitoring data of the engineering structure, and the time-varying constraint boundary is determined at the current monitoring time based on the bearing capacity degradation trajectory. The degree of damage at each damage location in the structural safety status assessment results is matched and compared with the time-varying constraint boundary to determine the warning level corresponding to each damage location; The overall warning level of the engineering structure is determined based on the distribution of the warning levels corresponding to each of the damage locations across all damage locations, and a graded warning instruction containing the overall warning level and the warning levels corresponding to each of the damage locations is generated.

6. The method according to claim 5, characterized in that, The bearing capacity degradation trajectory is determined based on historical monitoring data of the engineering structure, and the time-varying constraint boundary is determined at the current monitoring time based on the bearing capacity degradation trajectory, including: Structural response feature values ​​reflecting the structural bearing capacity status at multiple historical moments are extracted from the historical monitoring data of the engineering structure, and a time-series correlation is established between the structural response feature values ​​at each historical moment and the corresponding monitoring moment. Calculate the temporal decay trend of the structural response characteristic values ​​at each historical moment as a function of the monitoring time, and fit the bearing capacity degradation trajectory characterizing the structural bearing capacity as a function of time based on the temporal decay trend. Multiple degradation stage boundary points characterizing different load-bearing capacity levels are marked on the load-bearing capacity degradation trajectory, and the constraint threshold corresponding to each degradation stage boundary point is determined according to the load-bearing capacity level corresponding to each degradation stage boundary point. The degradation stage at the current monitoring time is determined based on the position of the current monitoring time on the degradation trajectory, and the constraint threshold corresponding to the degradation stage is extracted as the time-varying constraint boundary at the current monitoring time.

7. The method according to claim 1, characterized in that, Based on the tiered early warning command, a corresponding monitoring and response strategy is triggered and execution status information is recorded. The execution status information is then used to adaptively adjust the weight allocation relationship, including: The monitoring response strategy to be executed is determined based on the overall warning level and the warning level corresponding to each damage location contained in the graded warning instruction. The monitoring response strategy includes a data acquisition frequency adjustment scheme and a data transmission priority configuration scheme for each data source. The monitoring and response strategy is executed, and the data acquisition response latency and data transmission integrity information of each data source are recorded in real time during the execution process to obtain the execution status information; The performance indicators of each data source during the execution of the monitoring and response strategy are extracted from the execution status information. The performance indicators include the response latency deviation and transmission integrity deviation of each data source. Calculate the performance adjustment coefficient of each data source based on the performance indicators corresponding to each data source, and use the performance adjustment coefficient to adjust the weight coefficient of each data source in the weight allocation relationship to obtain the adjusted weight allocation relationship.

8. A strong earthquake data fusion and engineering safety monitoring system, used to implement the method as described in any one of claims 1-7, characterized in that, include: The time-frequency feature extraction unit is used to acquire vibration response data streams from multi-source sensing nodes, perform time-frequency coupled feature analysis on the vibration response data streams, and extract a set of feature vectors characterizing the dynamic properties of the structure. The weighted fusion unit is used to determine the weight allocation relationship reflecting the credibility of the data source based on the signal quality index and historical consistency index of each data source, and to perform cross-source data fusion on the feature vector set at multiple time scales based on the weight allocation relationship to obtain a fused feature vector reflecting the structural state. The structural safety early warning unit is used to compare and analyze the fused feature vector with the structural damage discrimination standard, identify the damage location and damage degree of the engineering structure, obtain the structural safety status assessment result, and generate graded early warning instructions based on the structural safety status assessment result and the time-varying constraint boundary characterizing the structural bearing capacity degradation process. The weight adaptive unit is used to trigger the corresponding monitoring response strategy based on the hierarchical early warning instruction and record the execution status information, and to adaptively adjust the weight allocation relationship using the execution status information.

9. An electronic device, characterized in that, include: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having computer program instructions stored thereon, characterized in that, When the computer program instructions are executed by the processor, they implement the method described in any one of claims 1 to 7.

Images