A multi-factor abnormal grading alarm method for bridge health state

Abstract

The present application relates to the technical field of bridge monitoring, and more particularly to a multi-factor abnormal grading alarm method for bridge health state, which comprises: collecting response time series data through a bridge monitoring sensor to obtain basic monitoring data of the bridge; obtaining a first optimization factor by dynamically matching and reconstructing the response waveform of the adjacent sensor; obtaining a second optimization factor by measuring the spatio-temporal consistency of the transmission parameter difference of the upstream and downstream sensors; obtaining a dynamic equivalent damage value by jointly weighting the first and second optimization factors; and obtaining a multi-factor abnormal grading alarm result by multi-source fusion analysis of the dynamic equivalent damage value, thereby solving the problem that the prior art cannot identify the time series transmission rule distortion under a dynamic load scenario, resulting in early damage missed reports.

Description

Technical Field

[0001] This invention relates to the field of bridge monitoring technology, and in particular to a multi-factor anomaly classification alarm method for bridge health status. Background Technology

[0002] As a crucial infrastructure in transportation networks, the safe operation of bridges directly impacts public travel and the stable functioning of the economy and society. To monitor the operational health of bridges in real time and prevent major accidents caused by sudden structural damage, large bridges typically deploy structural health monitoring systems. These systems continuously collect dynamic response data of the structure under environmental and traffic loads by deploying various sensors, including strain sensors, displacement sensors, and acceleration sensors, at key locations on the bridge. Accurately identifying early signs of structural damage or anomalies from this massive amount of monitoring data has long been a technical challenge in the field of bridge health monitoring. To improve the accuracy and reliability of alarms, existing technologies have proposed alarm methods based on multi-source information fusion. This method comprehensively assesses bridge damage by fusing monitoring data from sensors at different locations at the same time. The basic idea is as follows: First, the real-time measurements of each sensor are compared with a preset health baseline value to obtain the corresponding damage value; then, the damage value is converted into a basic probability assignment as evidence input; finally, a comprehensive damage probability is calculated using evidence fusion rules, and an alarm is triggered accordingly. Compared with traditional single-point threshold alarm methods, this type of method utilizes cross-validation of multi-source information, which effectively reduces the risk of false alarms caused by accidental failures or noise interference of a single sensor, and improves the reliability of the alarm to a certain extent.

[0003] However, existing alarm methods based on information fusion still have significant limitations when dealing with dynamic scenarios under the load of moving vehicles. The fundamental reason is that existing methods generally employ analysis based on instantaneous data snapshots, relying solely on sensor readings at a single, isolated moment, completely ignoring the temporal transmission patterns of vehicle load response along the sensor deployment path. In practical engineering, when early damage occurs in a bridge, such as fatigue crack propagation, loosening of critical connection nodes, or bearing performance degradation, the impact usually does not directly manifest as an abnormal increase in response amplitude, but rather as a change in local mechanical transmission characteristics. This change often manifests in the response data as a delay in transmission time or an abnormal attenuation of amplitude during transmission. Because such abnormal instantaneous amplitudes may still be within a set threshold range, existing instantaneous data-based judgment methods are completely insensitive to them, making it difficult to identify such hidden damage signs in a timely manner, potentially leading to missed detections and missed opportunities for optimal structural maintenance. Summary of the Invention

[0004] In view of this, the present invention aims to propose a multi-factor anomaly classification alarm method for bridge health status, in order to solve the problem that the existing technology cannot identify the distortion of the time sequence transmission law under dynamic load scenarios, resulting in the missed early damage.

[0005] This invention provides a multi-factor anomaly classification alarm method for bridge health status, which includes the following steps:

[0006] Step S1: Collect response time series data through bridge monitoring sensors to obtain the corresponding basic monitoring data of the bridge;

[0007] Step S2: Obtain the first optimization factor by dynamically matching and reconstructing the response waveforms of neighboring sensors;

[0008] Step S3: Obtain the second optimization factor by measuring the spatiotemporal consistency of the differences in parameters transmitted by upstream and downstream sensors;

[0009] Step S4: Obtain the dynamic equivalent damage value by jointly weighting the first optimization factor and the second optimization factor;

[0010] Step S5: Obtain multi-factor anomaly classification alarm results by performing multi-source fusion analysis on the dynamic equivalent damage value.

[0011] Preferably, the step of acquiring basic monitoring data by collecting bridge monitoring sensor response time series data includes:

[0012] The sampling frequency of the sensors is set; response time series data of multiple types of sensors deployed on the bridge structure are continuously collected and stored in a database for later retrieval. The multiple types of sensors include strain sensors, displacement sensors and acceleration amplitude sensors, and the response time series data includes strain sensor data, displacement sensor data and acceleration amplitude sensor data.

[0013] Preferably, the step of obtaining the first optimization factor by dynamically matching and reconstructing the response waveforms of neighboring sensors includes:

[0014] The optimal transfer parameter pair is obtained by performing parameter optimization processing on the time series data of the response of neighboring sensors; the first optimization factor is obtained by performing waveform reconstruction processing on the optimal transfer parameter pair of the time series data of the response of neighboring sensors.

[0015] Preferably, the step of obtaining the optimal transfer parameter pair by performing parameter optimization processing on the time series data of neighboring sensor responses includes:

[0016] Set the length of the sliding time window used for waveform matching; for any target sensor and any neighboring upstream sensor of the target sensor, solve for the minimum error of the error function of the target sensor and its upstream sensor by grid search method, and obtain the optimal transfer parameter pair of the target sensor and its upstream sensor.

[0017] The equations for solving the transfer parameters of the target sensor and its upstream sensor are as follows:

[0018] ;

[0019] in, Indicates the first The sensor and the first The optimal transfer parameter pair for each upstream sensor; Indicates the first The sensor and the first Optimal amplitude scaling factor among upstream sensors; Indicates the first The sensor and the first The optimal time shift between the upstream sensors; This represents the value of the independent variable when the objective function reaches its minimum. Indicates the length of the sliding time window used for waveform matching; Indicates the first The sensor at the first The actual measurement value at each moment; Indicates the first The upstream sensor is at the first The actual measurement value at each moment; This represents the magnitude scaling factor in the objective function.

[0020] Preferably, the step of performing waveform reconstruction processing on the optimal transfer parameter pair of the neighboring sensor response time series data to obtain the first optimization factor includes:

[0021] For any target sensor and any upstream sensor of the target sensor at any target time, the result of subtracting the target time of the target sensor from the optimal time shift between the target sensor and the upstream sensor is taken as the optimal matching time between the target sensor and the upstream sensor; the result of multiplying the optimal amplitude scaling factor of the target sensor and the upstream sensor of the target sensor by the measured value of the upstream sensor at the optimal matching time is taken as the first optimization factor of the target sensor and the upstream sensor of the target sensor at the target time.

[0022] Preferably, the step of obtaining the second optimization factor by measuring the spatiotemporal consistency of the differences in transmission parameters between upstream and downstream sensors includes:

[0023] By performing dynamic waveform matching processing on the response time series data of upstream and downstream sensors, the upstream and downstream transmission parameters are obtained.

[0024] By performing differential processing on the parameters transmitted by upstream and downstream sensors, the time delay normalization difference index and the amplitude scaling normalization difference index are obtained.

[0025] The second optimization factor is obtained by performing spatiotemporal consistency measurement on the normalized difference index.

[0026] Preferably, the step of obtaining upstream and downstream transmission parameters by performing dynamic waveform matching processing on the upstream and downstream sensor response time series data includes:

[0027] For any upstream sensor of any target sensor, obtain the optimal amplitude scaling factor and optimal time shift for the target sensor and the upstream sensor.

[0028] For any target sensor and any downstream target sensor, obtain the optimal amplitude scaling factor and optimal time shift for both the target sensor and the downstream target sensor.

[0029] Preferably, the step of obtaining the time delay normalization difference index and the amplitude scaling normalization difference index by performing differential processing on the parameters transmitted by upstream and downstream sensors includes:

[0030] For any target sensor, the optimal time shift between the target sensor and the upstream target sensor is taken as the first optimal time shift of the target sensor; the optimal time shift between the target sensor and the downstream target sensor is taken as the second optimal time shift of the target sensor; the absolute value of the calculation result of subtracting the first optimal time shift of the target sensor from the second optimal time shift of the target sensor is taken as the numerator, and the calculation result of adding the first optimal time shift of the target sensor to the second optimal time shift of the target sensor is taken as the denominator. The corresponding fraction is taken as the time delay normalization difference index of the target sensor.

[0031] The optimal amplitude scaling factor between the target sensor and the upstream target sensor is taken as the first optimal amplitude scaling factor of the target sensor; the optimal amplitude scaling factor between the target sensor and the downstream target sensor is taken as the second optimal amplitude scaling factor of the target sensor; the absolute value of the calculation result of subtracting the first optimal amplitude scaling factor and the second optimal amplitude scaling factor of the target sensor is taken as the numerator, and the calculation result of adding the first optimal amplitude scaling factor and the second optimal amplitude scaling factor of the target sensor is taken as the denominator. The corresponding fraction is taken as the amplitude scaling normalization difference index of the target sensor.

[0032] Preferably, the step of obtaining the second optimization factor by performing spatiotemporal consistency measurement on the normalized difference index includes:

[0033] For any target sensor, the result of adding the time delay normalization difference index and the amplitude scaling normalization difference index of the target sensor is used as the second optimization factor of the target sensor.

[0034] Preferably, the step of obtaining the dynamic equivalent damage value by jointly weighting the first optimization factor and the second optimization factor includes:

[0035] Set the number of candidate neighbor sensors and the number of dynamic neighbor sensors for the sensor; for any target sensor, determine the candidate neighbor sensors on both sides of the target sensor based on the number of candidate neighbor sensors; obtain the minimum error function value between the target sensor and the candidate neighbor sensors through the optimal time shift and the optimal amplitude scaling factor; select the sensor with the minimum error among the candidate neighbor sensors and use the selected sensor as the dynamic nearest neighbor sensor.

[0036] For any target sensor at any target time, the mean of the first optimization factors of all dynamic nearest neighbor sensors of the target sensor at the target time is used as the first mean evaluation of the target sensor, and the result of subtracting the actual measurement value of the target sensor at the target time from the first mean evaluation of the target sensor is used as the first equivalent damage evaluation of the target sensor.

[0037] The result of adding the second optimization factor of the target sensor to the constant 1 at the target time is used as the first equivalent evaluation weight of the target sensor. The result of multiplying the first equivalent damage evaluation of the target sensor by the first equivalent evaluation weight of the target sensor is used as the dynamic equivalent damage value of the target sensor at the target time.

[0038] The beneficial effects of the embodiments of the present invention compared with the prior art are as follows:

[0039] This invention, by introducing modeling of mechanical transmission laws and designing a two-step optimization factor, enables accurate identification of local bridge damage in complex dynamic scenarios caused by moving vehicle loads. Compared to traditional methods that rely solely on instantaneous data for static threshold judgment, this invention utilizes the correlation of response waveforms between neighboring sensors to dynamically optimize the matching relationship between time delay and amplitude, and compares the differences between the reconstructed results of the theoretical healthy response and the measured data. This approach not only accurately eliminates the response superposition effect caused by multiple concurrent vehicles or traffic flow fluctuations, avoiding false alarms, but also captures abnormal waveform distortions caused by changes in structural mechanical properties even when the response amplitude is not significantly abnormal, thereby improving the sensitivity to early, hidden damage. Simultaneously, this invention measures the continuity and consistency of transmission parameters from upstream and downstream sensors, further transforming the mechanical transmission anomalies caused by structural damage into quantitative indicators, and weightedly fusing them with the dynamic reconstruction residuals to obtain a dynamic equivalent damage value that reflects the true damage state. Based on this, this damage value is introduced into a multi-source information fusion alarm framework, enabling the alarm system to maintain high reliability and stability under complex traffic loads. Therefore, this invention not only improves the early warning accuracy of bridge structural health monitoring, but also enables earlier identification of the development trend of local damage, providing a scientific basis for bridge maintenance and management, and significantly reducing public safety risks. Attached Figure Description

[0040] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0041] Figure 1 This is a flowchart of a multi-factor anomaly classification alarm method for bridge health status provided in Embodiment 1 of the present invention. Detailed Implementation

[0042] Embodiments of this disclosure are described in detail below, with examples of these embodiments illustrated in the accompanying drawings. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this disclosure, and should not be construed as limiting it.

[0043] It should be noted that the terms "first," "second," etc., used in this disclosure and the accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this disclosure described herein can be implemented in orders other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this disclosure. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this disclosure.

[0044] To illustrate the technical solution of the present invention, specific embodiments are described below.

[0045] See Figure 1 This is a flowchart of a multi-factor anomaly classification alarm method for bridge health status provided in Embodiment 1 of the present invention, as follows: Figure 1 As shown, the method may include:

[0046] Step S1: Collect response time series data through bridge monitoring sensors to obtain the corresponding basic monitoring data of the bridge.

[0047] This step aims to obtain basic monitoring data of the bridge. First, the sampling frequency of the sensors is set. In this embodiment of the invention, the sampling frequency of the sensors is set to 50Hz. By continuously collecting response data from multiple types of sensors deployed on the bridge structure, the response time series data of the sensors is obtained and stored in a database for subsequent retrieval. The multiple types of sensors include strain sensors, displacement sensors, and acceleration amplitude sensors. The response time series data includes strain sensor data, displacement sensor data, and acceleration amplitude sensor data.

[0048] This completes the acquisition of response time series data through bridge monitoring sensors, and correspondingly obtains the basic monitoring data of the bridge.

[0049] Step S2: Obtain the first optimization factor by dynamically matching and reconstructing the response waveforms of neighboring sensors.

[0050] This step aims to address the problem that existing technologies cannot effectively distinguish between normal load responses and abnormal responses. When a vehicle load travels along a bridge, the structural response it induces is transmitted spatially in an ordered manner and temporally sequentially along the travel path. Therefore, for any two sensors positioned along the vehicle's travel path, the response signals they measure must exhibit a stable sequential relationship and morphological similarity in the time dimension. This relationship is jointly determined by the vehicle's speed and the healthy mechanical transmission characteristics of the bridge. Existing technologies analyze sensor readings independently at any target moment, completely ignoring the fact that the response signals of upstream sensors before the target moment already contain the response information that will be transmitted to the target sensor. This step aims to utilize this neglected look-ahead information. Specifically, if the structure is healthy, the response waveform of the target sensor within a certain period near the current moment should be reconstructed with a high degree of approximation from the response waveform of its upstream sensor a slightly earlier period, after a reasonable time shift and amplitude scaling transformation. However, in actual engineering scenarios, vehicle speeds are not constant, so a fixed time shift cannot adapt to dynamically changing traffic flow. To address this issue, this invention proposes a dynamic optimization construction method. Instead of pre-setting a fixed temporal transmission relationship, it searches in real-time within a continuously sliding time window for an optimal time shift and amplitude scaling factor, ensuring that the historical waveform of the upstream sensor matches the current waveform of the downstream sensor to the greatest extent possible. This construction yields an optimal theoretical reconstruction value of the downstream sensor's current state based on upstream sensor information. This reconstruction value itself is the first optimization factor, representing the theoretically expected response of the target sensor under the current traffic conditions and healthy structural transmission characteristics.

[0051] In summary, firstly, the optimal transfer parameter pair is obtained by performing parameter optimization processing on the response time series data of neighboring sensors. Specifically, the sliding time window length for waveform matching is set, and in this embodiment of the invention, the sliding time window length for waveform matching is set to 5 seconds. For any target sensor and any neighboring upstream sensor of the target sensor, the minimum error value of the error function of the target sensor and its upstream sensor is solved by the grid search method to obtain the optimal transfer parameter pair of the target sensor and its upstream sensor.

[0052] The equations for solving the transfer parameters of the target sensor and its upstream sensor, which are composed of error functions, are as follows:

[0053]

[0054] in, Indicates the first The sensor and the first The optimal transfer parameter pair for each upstream sensor; Indicates the first The sensor and the first Optimal amplitude scaling factor among upstream sensors; Indicates the first The sensor and the first The optimal time shift between the upstream sensors; This represents the value of the independent variable when the objective function reaches its minimum. Indicates the length of the sliding time window used for waveform matching; Indicates the first The sensor at the first The actual measurement value at each moment; Indicates the first The upstream sensor is at the first The actual measurement value at each moment; This represents the magnitude scaling factor in the objective function.

[0055] It should be noted that, in order to solve the above optimization problem, the embodiments of the present invention adopt the grid search method, which is well known in the field of parameter optimization. The grid search method discretizes the preset optimization interval of the amplitude scaling factor and the time shift to form a parameter grid. Then, it traverses all parameter pairs on the grid and calculates the integral error value corresponding to each parameter pair, which is the sum of squared errors in discrete data. Finally, the parameter pair with the smallest error value is taken as the optimal transfer parameter pair.

[0056] After obtaining the optimal transfer parameter pair, waveform reconstruction processing is performed on the optimal transfer parameter pair of the response time series data of the neighboring sensors to obtain the first optimization factor. Specifically, for any target sensor and any upstream sensor of any target sensor at any target time, the result of subtracting the optimal time shift between the target sensor and the upstream sensor is taken as the optimal matching time between the target sensor and the upstream sensor. The result of multiplying the optimal amplitude scaling factor of the target sensor and the upstream sensor of the target sensor by the measured value of the upstream sensor at the optimal matching time is taken as the first optimization factor of the target sensor and the upstream sensor of the target time.

[0057] In one embodiment, the first The sensor and its first The upstream sensor is at the first The expression for calculating the first optimization factor at time n is:

[0058]

[0059] in, Indicates the first The sensor and its first The upstream sensor is at the first The first optimization factor at each moment; Indicates the first The sensor and the first Optimal amplitude scaling factor among upstream sensors; Indicates the first The upstream sensor is at the first The actual measurement value at each moment; Indicates the first The sensor and the first The optimal time shift between the upstream sensors.

[0060] It should be noted that under vehicle loads, the bridge's response is not chaotic, but follows defined mechanical laws. Response of an upstream sensor With the The response of each sensor They are highly correlated in form, but there is a delay in time. . The optimization construction is designed specifically to address this scenario. The integral term mathematically represents the time window... Within, the geometric similarity between the two waveform curves. When the amplitude scaling factor and time shift are adjusted, so that the transformed first... The waveform of the first upstream sensor and the first The integral value is minimized when the waveforms of the first sensor highly overlap. This process essentially begins from the first... From the signals of the sensors, identify and extract the signals that can be obtained from its first sensor. The portion explained by the signals from the upstream sensors.

[0061] When multiple vehicles pass by in close proximity, it leads to When a high-amplitude peak appears after superposition, this superposition also occurs in the first... On the upstream sensor, the overall waveforms of the two signals remain highly similar. Therefore, The process can find a suitable and This ensures that the two waveforms can still match well. The final calculated... It will be very close The actual value. This shows that, although The amplitude is high, but its shape perfectly conforms to the healthy temporal transmission pattern and is predictable. This provides a basis for subsequent judgment on whether it is normal superposition, thus avoiding false alarms. Conversely, if the first... Damage near the sensor can cause additional oscillations or distortions in its response waveform. In this case, no matter how it is adjusted... and The first healthy The waveforms of each upstream sensor cannot perfectly match the first... The distorted waveform of each sensor will lead to The obtained minimum integral value is still relatively large, and the reconstructed value is... and There are non-negligible residuals. This provides us with an effective way to identify damage signs that are small in amplitude but abnormal in shape. Thus, through the construction of the first optimization factor, the theoretical health response based on the time-series transmission law can be successfully separated from the original signal.

[0062] Thus, the first optimization factor was obtained by dynamically matching and reconstructing the response waveforms of neighboring sensors.

[0063] Step S3: Obtain the second optimization factor by measuring the spatiotemporal consistency of the differences in transmission parameters between upstream and downstream sensors.

[0064] In step S2, for the target sensor and each of its neighboring sensors, a pair of optimal transfer parameters are obtained through dynamic waveform matching. This pair of parameters objectively reflects the time delay and amplitude scaling relationship of the load response as it is transferred from the upstream sensor to the target sensor within the local time window at the current moment. This step is to verify the rationality and consistency of these calculated transfer parameters. The physical basis for this is that, for a healthy bridge structure, the transfer process of vehicle loads should be smooth and continuous. Specifically, when a vehicle travels at a constant or nearly constant speed along a path, the time required for its response to be transferred between adjacent, equidistant sensor segments, and the proportion of change in response amplitude, should be approximately constant. In other words, the transfer pattern from the upstream sensor to the target sensor should be consistent with the transfer pattern from the target sensor to the downstream sensor, which is its immediate neighbor. However, when structural damage occurs in the local area where the target sensor is located, such as cracks causing a decrease in local stiffness, or abnormal settlement of the supports, this consistency of the transfer pattern will be disrupted. When a load passes through a damaged area, its transmission time may abnormally increase, or the response amplitude may abnormally decrease due to abnormal energy dissipation. This distortion of the transmission pattern at the damaged point is a more stable and fundamental damage characteristic than the response amplitude itself. Therefore, this step analyzes the transmission parameters obtained in step S2. A metric is obtained by comparing the differences in transmission parameters between the upstream and downstream paths centered on the target sensor. This metric reflects whether the continuity and consistency of the transmission path have been disrupted, thereby identifying anomalies in the temporal transmission pattern caused by local damage.

[0065] In summary, firstly, by performing dynamic waveform matching processing on the response time series data of upstream and downstream sensors, the upstream and downstream transmission parameters are obtained. Specifically, for any upstream sensor of any target sensor, the optimal amplitude scaling factor and optimal time shift amount between the target sensor and the upstream sensor are obtained; for any downstream sensor of any target sensor, the optimal amplitude scaling factor and optimal time shift amount between the target sensor and the downstream sensor are obtained.

[0066] After obtaining the upstream and downstream transmission parameters, the parameters from the upstream and downstream sensors are further differentiated to obtain time delay normalization difference index and amplitude scaling normalization difference index. Specifically, for any target sensor, the optimal time shift between the target sensor and the upstream target sensor is taken as the first optimal time shift of the target sensor; the optimal time shift between the target sensor and the downstream target sensor is taken as the second optimal time shift of the target sensor; the absolute value of the calculation result of subtracting the first optimal time shift of the target sensor from the second optimal time shift of the target sensor is taken as the numerator, and the first optimal time shift of the target sensor is added to the second optimal time shift of the target sensor. The result is used as the denominator, and the corresponding fraction is used as the time delay normalization difference index of the target sensor; the optimal amplitude scaling factor between the target sensor and the upstream target sensor is used as the first optimal amplitude scaling factor of the target sensor; the optimal amplitude scaling factor between the target sensor and the downstream target sensor is used as the second optimal amplitude scaling factor of the target sensor; the absolute value of the calculation result of subtracting the first optimal amplitude scaling factor and the second optimal amplitude scaling factor of the target sensor is used as the numerator, and the calculation result of adding the first optimal amplitude scaling factor and the second optimal amplitude scaling factor of the target sensor is used as the denominator. The corresponding fraction is used as the amplitude scaling normalization difference index of the target sensor.

[0067] Finally, by performing spatiotemporal consistency measurement on the normalized difference index, a second optimization factor is obtained. Specifically, for any target sensor, the result of adding the target sensor's time delay normalized difference index and amplitude scaling normalized difference index is used as the target sensor's second optimization factor.

[0068] In one implementation, assume the first The upstream sensor to the first The sensor at the first The optimal time shift at each moment is: ;No. The sensor to the first The downstream sensor in the first The optimal time shift at each moment is: ;No. The upstream sensor to the first The sensor at the first The optimal amplitude scaling factor at time is ;No. The sensor to the first The downstream sensor in the first The optimal amplitude scaling factor at time is Then the first The sensor at the first The expression for calculating the second optimization factor at time t is:

[0069]

[0070] in, Indicates the first The sensor at the first The second optimization factor at each time step; Indicates the first The upstream sensor to the first The sensor at the first The optimal time shift at any given moment; Indicates the first The sensor to the first The downstream sensor in the first The optimal time shift at any given moment; Indicates the first The upstream sensor to the first The sensor at the first The optimal amplitude scaling factor at any given moment; Indicates the first The sensor to the first The downstream sensor in the first The optimal amplitude scaling factor at any given moment; This indicates absolute value calculation.

[0071] It should be noted that the formula for calculating the second optimization factor consists of two parts, corresponding to the continuity tests of the two transfer characteristics: time delay and amplitude scaling. Taking the first term... For example, molecules The absolute difference in propagation delay between the upstream and downstream paths was calculated. (Denominator) This is the sum of the two time delays. Dividing the two yields a normalized relative difference rate. Normalization allows it to automatically adapt to different vehicle speeds. For example, when the vehicle is traveling at high speed... and The values ​​themselves are very small; even if their absolute differences are small, the relative difference rate can still be significant. Conversely, when the vehicle is traveling at low speed, A large value indicates that a small absolute difference may not necessarily indicate an anomaly. This construction makes the second optimization factor well-adapted to changes in vehicle speed. When the bridge structure is healthy and the vehicle travels at approximately a constant speed, it can be expected that... and They will be very close. and They will also be very close at this time. The value will approach This indicates that, with the first The transmission path centered on each sensor is uniform and healthy. Even if the vehicle experiences a loss due to normal driving behavior (such as slight acceleration or deceleration)... and Synchronous changes occur, but due to and It will change in sync. and They will also change synchronously, and the differences between them remain very small, therefore The value will remain low, preventing false positives. However, when the... When damage exists near a sensor, the mechanical transmission characteristics change. For example, a decrease in stiffness causes the load to shift from the first sensor to the second sensor. The upstream sensor transmits data to the first... Time of each sensor Significantly longer than from the first The first sensor transmits data to the health area. The time of each downstream sensor This asymmetry will lead to This produces a significant non-zero value, thus making The value increases. Similarly, energy dissipation at the site of damage may also lead to... and Significant differences exist between them. Therefore, As a quantitative indicator, it can capture the disruption of the transmission pattern caused by local damage.

[0072] Thus, the second optimization factor was obtained by measuring the spatiotemporal consistency of the parameters transmitted by upstream and downstream sensors.

[0073] Step S4: Obtain the dynamic equivalent damage value by performing a joint weighted calculation on the first optimization factor and the second optimization factor.

[0074] After obtaining the first and second optimization factors, the damage value is evaluated. Existing technologies calculate damage values ​​by directly comparing the absolute magnitude of measured values, which fails to distinguish the source of the response. The solution of this invention is that a true damage signal should simultaneously possess two characteristics: first, the response pattern of the damage signal cannot be reasonably explained by neighboring healthy sensors based on normal temporal transmission relationships, i.e., waveform matching residuals exist; second, this mismatch is accompanied by a disruption of the temporal transmission pattern itself, i.e., path distortion exists. Based on the above optimization factor design analysis, this step obtains the optimized damage value, using the average waveform matching residual as a measure of basic damage to reflect the degree of abnormality in the response pattern. The second optimization factor... As a penalty coefficient. When A larger value indicates that when the temporal transmission pattern is disrupted, it significantly amplifies the underlying waveform matching residual, achieving intelligent weighting of damage evidence. Even a signal with a small amplitude but accompanied by waveform mismatch and path distortion (a higher second optimization factor) will have its calculated equivalent damage value significantly enhanced; conversely, a signal with a large amplitude but good waveform matching and a smooth transmission path (a lower second optimization factor) will have its calculated equivalent damage value effectively suppressed. This damage assessment, compared to existing techniques, can more accurately pinpoint the true structural anomaly.

[0075] In summary, this step obtains the dynamic equivalent damage value by jointly weighting the first optimization factor and the second optimization factor. Specifically, the number of candidate neighbor sensors and the number of dynamic neighbor sensors of the sensor are set. In this embodiment of the invention, the number of candidate neighbor sensors is set to 4 and the number of dynamic neighbor sensors of the sensor is set to 2. For any target sensor, candidate neighbor sensors of the target sensor are determined on both sides of the target sensor based on the number of candidate neighbor sensors of the sensor. The minimum error function value between the target sensor and the candidate neighbor sensors is obtained by using the optimal time shift and the optimal amplitude scaling factor between the target sensor and the candidate neighbor sensors. The sensor with the minimum error among the candidate neighbor sensors is selected as the dynamic nearest neighbor sensor.

[0076] For any target sensor at any target time, the mean of the first optimization factors of all dynamic nearest neighbor sensors of the target sensor at the target time is used as the first mean evaluation of the target sensor, and the result of subtracting the actual measurement value of the target sensor at the target time from the first mean evaluation of the target sensor is used as the first equivalent damage evaluation of the target sensor.

[0077] The result of adding the second optimization factor of the target sensor to the constant 1 at the target time is used as the first equivalent evaluation weight of the target sensor. The result of multiplying the first equivalent damage evaluation of the target sensor by the first equivalent evaluation weight of the target sensor is used as the dynamic equivalent damage value of the target sensor at the target time.

[0078] In one implementation, assume the first The dynamic nearest neighbor sensor set of the sensors is ;No. The number of sensors in the dynamic nearest neighbor sensor set of a sensor is Then the first The sensor at the first The expression for calculating the dynamic equivalent damage value at time t is:

[0079]

[0080] in, Indicates the first The sensor at the first The dynamic equivalent damage value at each moment; Indicates the first The sensor and its first The nearest neighbor sensor in the first The first optimization factor at each moment; Indicates the first The sensor at the first The actual measurement value at each moment; Indicates the first The number of sensors in the dynamic nearest neighbor sensor set of a sensor; Indicates the first A dynamic nearest neighbor sensor set for each sensor; Indicates the first The sensor at the first The second optimization factor at each time step.

[0081] Thus, the dynamic equivalent damage value is obtained by jointly weighting the first optimization factor and the second optimization factor.

[0082] Step S5: Obtain multi-factor anomaly classification alarm results by performing multi-source fusion analysis on the dynamic equivalent damage values.

[0083] This step utilizes the dynamic equivalent damage value calculated in step S4 and integrates it into the existing multi-source information fusion alarm framework to complete the final anomaly diagnosis and graded alarm, thereby solving the problems raised in the background technology. The processing flow can refer to the method framework disclosed in CN115909686B.

[0084] Specifically, each sensor at every moment Calculated dynamic equivalent damage value This serves as the basic input, replacing the damage value obtained by subtracting the static baseline from the actual measured value in existing technologies. Subsequently, following the subsequent processing flow of existing technologies, based on... The amplitude calculation assigns basic probability values ​​to each sensor to quantify the strength of evidence pointing to an abnormal state from each information source. Then, using evidence theory and other information fusion algorithms, multi-level fusion calculations are performed on evidence from different sensors within the same monitoring section (first fusion) and from different monitoring sections (second fusion) to obtain a comprehensive confidence level regarding the overall health status of the structure or key components. Finally, based on preset graded alarm criteria, the calculated comprehensive confidence level is compared with alarm thresholds for different levels, and corresponding early warning information is output, indicating levels such as normal, attention, warning, or danger.

[0085] By employing the dynamic equivalent damage value calculated using this invention, the entire alarm system can effectively identify and ignore the superposition of responses caused by normal multi-vehicle concurrency, thus avoiding false alarms. Simultaneously, because this damage value is highly sensitive to distortions in the temporal transmission pattern, the system can detect early abnormal signs caused by localized structural damage more quickly, reducing the risk of missed alarms. Therefore, this invention improves the accuracy and reliability of alarms by optimizing the evidence source of the alarm system.

[0086] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. 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 of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A multi-factor anomaly classification alarm method for bridge health status, characterized in that, The multi-factor anomaly classification alarm method for bridge health status includes: Step S1: Collect response time series data through bridge monitoring sensors to obtain the corresponding basic monitoring data of the bridge; Step S2: Obtain the first optimization factor by dynamically matching and reconstructing the response waveforms of neighboring sensors; Step S3: Obtain the second optimization factor by measuring the spatiotemporal consistency of the differences in parameters transmitted by upstream and downstream sensors; Step S4: Obtain the dynamic equivalent damage value by jointly weighting the first optimization factor and the second optimization factor; Step S5: Obtain multi-factor anomaly classification alarm results by performing multi-source fusion analysis on the dynamic equivalent damage value.

2. The multi-factor anomaly classification alarm method for bridge health status according to claim 1, characterized in that, The process of acquiring response time series data through bridge monitoring sensors to obtain corresponding basic monitoring data of the bridge includes: The sampling frequency of the sensors is set; multiple sensor arrays are continuously deployed along the bridge's driving path; the response time series data of the multiple sensor arrays deployed on the bridge structure are continuously collected and stored in a database for later retrieval; the multiple sensors include strain sensors, displacement sensors, and acceleration amplitude sensors; and the response time series data includes strain sensor data, displacement sensor data, and acceleration amplitude sensor data.

3. The multi-factor anomaly classification alarm method for bridge health status according to claim 1, characterized in that, The step of obtaining the first optimization factor by dynamically matching and reconstructing the response waveforms of neighboring sensors includes: The optimal transfer parameter pair is obtained by performing parameter optimization processing on the time series data of the response of neighboring sensors; the first optimization factor is obtained by performing waveform reconstruction processing on the optimal transfer parameter pair of the time series data of the response of neighboring sensors.

4. The multi-factor anomaly classification alarm method for bridge health status according to claim 3, characterized in that, The step of obtaining the optimal transfer parameter pair by performing parameter optimization processing on the time series data of the response of nearby sensors includes: Set the length of the sliding time window used for waveform matching; for any target sensor and any neighboring upstream sensor of the target sensor, solve for the minimum error of the error function of the target sensor and its upstream sensor by grid search method, and obtain the optimal transfer parameter pair of the target sensor and its upstream sensor. The equations for solving the transfer parameters of the target sensor and its upstream sensor are as follows: ； in, Indicates the first The sensor and the first The optimal transfer parameter pair for each upstream sensor; Indicates the first The sensor and the first Optimal amplitude scaling factor among upstream sensors; Indicates the first The sensor and the first The optimal time shift between the upstream sensors; This represents the value of the independent variable when the objective function reaches its minimum. Indicates the length of the sliding time window used for waveform matching; Indicates the first The sensor at the first The actual measurement value at each moment; Indicates the first The upstream sensor is at the first The actual measurement value at each moment; This represents the magnitude scaling factor in the objective function.

5. The multi-factor anomaly classification alarm method for bridge health status according to claim 3, characterized in that, The step of performing waveform reconstruction processing on the optimal transfer parameters of the neighboring sensor response time series data to obtain the first optimization factor includes: For any target sensor and any upstream sensor of the target sensor at any target time, the result of subtracting the target time of the target sensor from the optimal time shift between the target sensor and the upstream sensor is taken as the optimal matching time between the target sensor and the upstream sensor; the result of multiplying the optimal amplitude scaling factor of the target sensor and the upstream sensor of the target sensor by the measured value of the upstream sensor at the optimal matching time is taken as the first optimization factor of the target sensor and the upstream sensor of the target sensor at the target time.

6. The multi-factor anomaly classification alarm method for bridge health status according to claim 1, characterized in that, The step of obtaining a second optimization factor by measuring the spatiotemporal consistency of parameters transmitted by upstream and downstream sensors includes: By performing dynamic waveform matching processing on the response time series data of upstream and downstream sensors, the upstream and downstream transmission parameters are obtained. By performing differential processing on the parameters transmitted by upstream and downstream sensors, the time delay normalization difference index and the amplitude scaling normalization difference index are obtained. The second optimization factor is obtained by performing spatiotemporal consistency measurement on the normalized difference index.

7. The multi-factor anomaly classification alarm method for bridge health status according to claim 6, characterized in that, The process of obtaining upstream and downstream transmission parameters by performing dynamic waveform matching processing on the time series data of upstream and downstream sensor responses includes: For any upstream sensor of any target sensor, obtain the optimal amplitude scaling factor and optimal time shift for the target sensor and the upstream sensor. For any target sensor and any downstream target sensor, obtain the optimal amplitude scaling factor and optimal time shift for both the target sensor and the downstream target sensor.

8. A multi-factor anomaly classification alarm method for bridge health status according to claim 6, characterized in that, The step of differentiating the parameters transmitted from upstream and downstream sensors to obtain time delay normalization difference index and amplitude scaling normalization difference index includes: For any target sensor, the optimal time shift between the target sensor and the upstream target sensor is taken as the first optimal time shift of the target sensor; the optimal time shift between the target sensor and the downstream target sensor is taken as the second optimal time shift of the target sensor; the absolute value of the calculation result of subtracting the first optimal time shift of the target sensor from the second optimal time shift of the target sensor is taken as the numerator, and the calculation result of adding the first optimal time shift of the target sensor to the second optimal time shift of the target sensor is taken as the denominator. The corresponding fraction is taken as the time delay normalization difference index of the target sensor. The optimal amplitude scaling factor between the target sensor and the upstream target sensor is taken as the first optimal amplitude scaling factor of the target sensor; the optimal amplitude scaling factor between the target sensor and the downstream target sensor is taken as the second optimal amplitude scaling factor of the target sensor; the absolute value of the calculation result of subtracting the first optimal amplitude scaling factor and the second optimal amplitude scaling factor of the target sensor is taken as the numerator, and the calculation result of adding the first optimal amplitude scaling factor and the second optimal amplitude scaling factor of the target sensor is taken as the denominator. The corresponding fraction is taken as the amplitude scaling normalization difference index of the target sensor.

9. A multi-factor anomaly classification alarm method for bridge health status according to claim 6, characterized in that, The process of obtaining the second optimization factor by performing spatiotemporal consistency measurement on the normalized difference index includes: For any target sensor, the result of adding the time delay normalization difference index and the amplitude scaling normalization difference index of the target sensor is used as the second optimization factor of the target sensor.

10. A multi-factor anomaly classification alarm method for bridge health status according to claim 1, characterized in that, The step of obtaining the dynamic equivalent damage value by jointly weighting the first optimization factor and the second optimization factor includes: Set the number of candidate neighbor sensors and the number of dynamic neighbor sensors for the sensor; for any target sensor, determine the candidate neighbor sensors on both sides of the target sensor based on the number of candidate neighbor sensors; obtain the minimum error function value between the target sensor and the candidate neighbor sensors through the optimal time shift and the optimal amplitude scaling factor; select the sensor with the minimum error among the candidate neighbor sensors and use the selected sensor as the dynamic nearest neighbor sensor. For any target sensor at any target time, the mean of the first optimization factors of all dynamic nearest neighbor sensors of the target sensor at the target time is used as the first mean evaluation of the target sensor, and the result of subtracting the actual measurement value of the target sensor at the target time from the first mean evaluation of the target sensor is used as the first equivalent damage evaluation of the target sensor. The result of adding the second optimization factor of the target sensor to the constant 1 at the target time is used as the first equivalent evaluation weight of the target sensor. The result of multiplying the first equivalent damage evaluation of the target sensor by the first equivalent evaluation weight of the target sensor is used as the dynamic equivalent damage value of the target sensor at the target time.

Images