A real-time wind speed index calculation method and system for bridge structure health monitoring

Abstract

The application discloses a kind of real-time wind speed index calculation method and system for bridge structure health monitoring.The application extracts real-time bridge health monitoring data from message middleware, pushes to real-time stream calculation engine, engine carries out abnormal correction to data, pushes real-time operator to carry out incremental calculation, and the aggregated intermediate state of sliding window is pushed to memory database for storage, while using the improved bridge structure health monitoring index for calculation, finally by index job, the index calculation result is reorganized and spliced, and pushed to message middleware output.The application can reduce the delay of bridge structure health monitoring index calculation;Ease the influence of instantaneous noise, jump point and other data anomalies caused by bridge environment, hardware equipment and other factors on precision;According to the characteristics of index, delay specified time trigger data processing, improve the accuracy of index calculation;Introduce memory database to store sliding window aggregation value and index value, reduce the time consumption of reading and writing hard disk.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of real-time data processing of bridge structure health monitoring, and in particular to a real-time wind speed index calculation method and system for bridge structure health monitoring. BACKGROUND

[0002] With the continuous upgrading of building materials and the continuous progress of bridge engineering technology, the construction scale and safe service life of bridges are continuously expanding. Therefore, new demands are put forward for the bridge structure health monitoring system. On the one hand, the volume of bridge structure health monitoring data is increasing, which puts forward new requirements for the processing capacity of sensor monitoring data; on the other hand, in order to ensure that the bridge structure health monitoring index can as much as possible reflect the health condition of the bridge, higher standards are put forward for the calculation accuracy of the bridge structure health monitoring index. The bridge structure state is manifested through the structure response, and the working mode of health monitoring is to obtain the structure state of the bridge by analyzing the structure response of the bridge, so as to perform a series of health assessments on the bridge. The bridge health monitoring system is generally divided into three levels, namely long-term online automatic health monitoring system, periodic offline health monitoring system and periodic maintenance health monitoring system. The long-term online automatic health monitoring system, as the highest level of monitoring system, is particularly suitable for use on large-span complex bridge structures and key bridges with important strategic significance. Real-time computing technology, as the core technology in the long-term online automatic health monitoring system, has higher and higher standards for the real-time processing capacity of massive data and the real-time feedback of the bridge health condition, and the traditional solution is to use database calls for timing calculation. The applicable scenario of SQL query is mainly large-scale batch processing, and the performance is not good when performing real-time stream processing.

[0003] In view of how to perform real-time calculation on massive data, the existing Lambda structure and Kappa architecture realize the real-time calculation technical scheme of "batch-flow integration". The Lambda architecture is a big data processing system jointly maintained by a stream computing mode and a batch computing mode, which provides real-time data calculation by the stream computing module and accurate offline calculation by the batch computing module to correct possible errors in real-time calculation. The disadvantage of this architecture is that it needs to maintain two independent code logics of stream computing and batch computing, which increases the maintenance cost, and it is difficult to realize the timing pursuit of batch computing to real-time calculation. The Kappa architecture uses a powerful stream computing mode to provide stable and timely computing power, and needs to wait for the idle period of the service to repair data when an error occurs. In order to ensure the accuracy of calculation, both architectures affect the normal stream computing service and the system performance, and it is difficult to meet the real-time index calculation demand of bridge structure health monitoring.

[0004] In view of the above problems, the current urgent need for a real-time wind speed index calculation method and system for bridge structure health monitoring can overcome the defects of long time consumption and low precision in the existing industry technical solutions, and can simultaneously achieve low delay high throughput data processing and high precision index calculation. SUMMARY

[0005] The purpose of the present application is to overcome the deficiencies of the prior art in the field of real-time calculation of bridge structure health monitoring indicators, and to propose a real-time wind speed index calculation method and system for bridge structure health monitoring.

[0006] The purpose of the present application is achieved by the following technical solution: a real-time wind speed index calculation method for bridge structure health monitoring, comprising the following steps:

[0007] (1) The collected bridge sensor data is cached in real time to the upstream message middleware, and when the wind speed index needs to be calculated, the bridge sensor data cached in the upstream message middleware is extracted and parsed in real time, and pushed to the real-time stream calculation engine in the format of (id, timestamp, data);

[0008] (2) The real-time operator in the real-time stream calculation engine uses incremental calculation and sliding window to aggregate the bridge sensor data; the data aggregation results of the real-time operator are detected by an anomaly detection algorithm, and after the abnormal data is modified to normal, it is stored for a short period of time, and is pushed to the real-time wind speed index job and the real-time operator for calculation;

[0009] (3) The real-time wind speed index job of the bridge structure health monitoring extracts the stored data in the memory database according to the calculation logic, designs a delay trigger mechanism based on data time, aligns, calculates and reorganizes the data, and stores the calculation results in the memory database and pushes them to the downstream message middleware;

[0010] (4) According to the index calculation logic requirement, the calculation results stored in the memory database will be extracted by the index job, and step (3) will be continued until the index calculation logic is completed, and the index calculation results will be pushed to the downstream message middleware.

[0011] Further, in step (2), when the real-time operator uses a sliding window to aggregate bridge data, it uses data-driven incremental calculation, and the time span of the sliding window increases continuously as data enters, and when the aggregated data in the sliding window reaches the maximum span, it is stored and emptied.

[0012] Further, in step (2), the real-time wind speed index includes main wind direction fluctuating wind speed, cross wind direction fluctuating wind speed, vertical direction fluctuating wind speed, average wind speed and turbulence intensity.

[0013] Further, for each bridge sensor data channel, the real-time operator maintains multiple independent sliding windows, each of which maintains an intermediate state value, and the start time intervals of adjacent sliding windows are the same; when new bridge sensor data is extracted from the upstream message middleware into the stream computing engine, all sliding windows perform calculations at the same time.

[0014] Further, in step (3), according to the characteristics of the bridge health monitoring scene, the delay trigger mechanism introduces expert experience, and sets the data processing delay trigger based on the driving clock.

[0015] Further, in step (2), a 3sigma or Z-score anomaly detection algorithm is used to monitor data anomalies based on recent historical data, and abnormal data is modified.

[0016] Further, in step (3), the delay trigger mechanism drives the real-time wind speed index job, and the real-time wind speed index is calculated by the real-time operator according to the index job calculation logic.

[0017] Further, in steps (3) and (4), the data format pushed to the downstream message middleware is {id, name, starttime, endtime, data}.

[0018] On the other hand, the present application also provides a real-time wind speed index calculation system for bridge structure health monitoring, which comprises a data processing module, a real-time stream computing module, an index calculation logic module and an index job extraction module.

[0019] The data processing module is used to extract and analyze the bridge sensor data cached in the upstream message middleware in real time, and push it to the real-time stream computing module in the format of (id, timestamp, data);

[0020] The real-time stream computing module is used to aggregate bridge sensor data by using incremental calculation and sliding windows through real-time operators in the real-time stream computing engine; after abnormal data is modified to normal by an anomaly detection algorithm, the data aggregation result of the real-time operator is detected, and the abnormal data is stored for a short period of time, and each real-time wind speed index job and real-time operator is pushed for calculation;

[0021] The index calculation logic module is used to design a delay trigger mechanism based on data time according to the calculation logic of the real-time wind speed index job for bridge structure health monitoring, extract the data stored in the in-memory database for data alignment, calculation and reorganization, and store the calculation result in the in-memory database and push it to the downstream message middleware.

[0022] The index job extraction module is used for extracting the calculation result stored in the in-memory database according to the index calculation logic requirement of the index calculation logic module, and continuing to execute the index calculation logic module until the index calculation logic is completed, and pushing the index calculation result to the downstream message middleware.

[0023] The beneficial effects of the present application are as follows: first, the periodic characteristics of the bridge sensor data are utilized to correct data anomalies, which can better cope with transient noise, jump points and other data anomalies caused by factors such as physical factors affecting hardware devices, signal interference in data transmission environment, and reduce the influence of abnormal data on index accuracy; at the same time, an incremental calculation method and a sliding window data aggregation technology suitable for real-time calculation of bridge health monitoring indicators are proposed, the average wind speed, fluctuating wind speed and turbulence intensity index calculation method is improved, the storage volume is reduced, not only the resource utilization rate fluctuation caused by traditional SQL batch calculation is reduced, but also the IO pressure and time consumption caused by read-write hard disk is reduced, the index calculation delay is greatly reduced, and the system throughput is increased; in addition, a delay trigger mechanism is designed to relieve the real-time data local disorder and late problem caused by network transmission, signal interference and other reasons of bridge structure health monitoring data, and ensure data calculation accuracy as much as possible. BRIEF DESCRIPTION OF DRAWINGS

[0024] Figure 1 The figure is a system architecture diagram for real-time wind speed index calculation for bridge structure health monitoring;

[0025] Figure 2 The figure is a data aggregation diagram of a single data stream real-time calculation operator sliding window;

[0026] Figure 3 The figure is a data-driven delay calculation scheme diagram;

[0027] Figure 4 The figure is a data processing flowchart of the real-time wind speed index calculation method for bridge structure health monitoring;

[0028] Figure 5 The figure is a real-time calculation flowchart of the average wind speed index;

[0029] Figure 6 The figure is a real-time calculation flowchart of the main wind direction and transverse wind direction fluctuating wind speed index;

[0030] Figure 7 The figure is a real-time calculation flowchart of the turbulence intensity index. DETAILED DESCRIPTION

[0031] The present application will be further described in detail below in combination with the drawings and specific embodiments.

[0032] The present application provides a real-time wind speed index calculation method for bridge structure health monitoring, as shown inFigure 1 and Figure 4 As shown, it includes the following steps:

[0033] (1) The data processing module extracts the latest bridge sensor data from the upstream message middleware in real time, parses the message body data in the upstream message middleware into a (id, timestamp, data) structure, where timestamp is the data time, id is the unique identifier of the data stream, and pushes it to the real-time stream computing engine to calculate the mean and variance in units of data stream for data correction.

[0034] Real-time mean operator increment calculation formula:

[0035]

[0036] Formula for calculating the real-time variance operator increment:

[0037]

[0038] in, and S (n-1) They are x1, x2, ..., x n-1 The mean and variance, and S (n) They are x1, x2, ..., x n-1 ,x n The mean and variance. and S (n) These represent the mean intermediate state and variance intermediate state maintained within the sliding window, respectively. x1, x2, ..., x n-1 ,x n This refers to data calculated within a sliding window. When the real-time operator uses the sliding window to aggregate bridge data, it performs incremental calculations driven by the data. The time span of the sliding window increases continuously as data enters, until it reaches its maximum span, at which point the data is stored and cleared.

[0039] For each data stream, the real-time operator maintains multiple independent sliding windows, each window maintaining an intermediate state value. Adjacent sliding windows have the same start time interval. When new bridge monitoring data enters the upstream message middleware, all sliding windows are calculated simultaneously. If the sliding window span is less than 10 minutes, only the intermediate state needs to be updated. If the sliding window span is 10 minutes, the currently calculated intermediate state is used as the 10-minute data aggregation value, along with relevant parameters, and stored in the in-memory database. The intermediate states of the sliding windows are then cleared to prepare for incremental calculations of new data. The relevant parameters are used in cases of delayed data arrival; incremental calculations are performed using these parameters and the delayed original data to update the aggregation value in the in-memory database. Figure 2As shown, the present application stores the aggregated intermediate state calculated by each operator in the past 60 minutes to the in-memory database, Figure 2 10 independent sliding windows, each window maintains an intermediate state value, and the start time interval of adjacent sliding windows is 1 minute. When new bridge monitoring data enters the upstream message middleware, 10 sliding windows simultaneously perform calculations and update their respective intermediate states. The sliding window stores the aggregated intermediate state every 10 minutes, so the data mean and variance calculation results of the 10-minute span starting at the whole minute can be obtained in the in-memory database. When data is late, the aggregated value of the time span to which the data belongs can be found in the in-memory database according to the time of the data, and the value is updated, for example Figure 3 In this way, data with a time of 9:56 arrives after 10:00, and the time interval to which 9:56 belongs is 0:00 to 10:00. The new mean and variance are calculated using the incremental calculation of the data aggregation value of the interval, and the in-memory database is updated.

[0040] (2) In units of data streams, real-time data correction is performed for each data stream, and a 3sigma or Z-score anomaly detection algorithm is used to monitor data anomalies according to recent historical data, and abnormal data is modified. The present application uses the 3sigma correction method, which requires extracting the mean and variance of the past 10 minutes of data for 3sigma correction. The present application sets a delay trigger mechanism for the operation of extracting the mean and variance for 3sigma correction, as shown in Figure 3 As shown, the extraction operation is driven by the entire data stream, and when the latest data of the entire data stream exceeds the whole minute, it is considered that the data time has reached the whole minute, triggering the corresponding operation. For example, Figure 3The data time of the data in the certain data is 11:01, which is the first data exceeding 11:00, and thus it is considered that the data time has reached 11:00, and the operation of the whole minute can be triggered, that is, the data extraction operation. According to the expert experience, a 1-minute waiting delay is set, when the data time reaches 11:00, it is considered that the data in the interval of 0:00 to 10:00 has basically reached the memory database, and the more accurate mean and variance in the interval of 0:00 to 10:00 can be extracted from the memory database for the 3sigma algorithm to perform anomaly detection; when the latest data of the data stream does not exceed the whole minute, the mean and variance extracted the last time are still used. The 3sigma algorithm sets a normal data range (μ-3σ, μ+3σ) for real-time checking, and the data exceeding the normal range is determined as an anomaly, if the current data value is lower than the μ-3σ boundary, the μ-2σ value is used for replacement, and if the current data value is higher than the μ+3σ boundary, the μ+2σ value is used for replacement. Wherein μ represents the 10-minute mean extracted from the memory database, and σ represents the standard deviation value calculated from the 10-minute variance extracted from the memory database. After the data correction is completed, the corrected data is pushed to the real-time stream computing engine in real time, which is used for clock synchronization and real-time operator calculation of various index job calculations, and the corrected original data in the last 5 minutes is cached to the memory database.

[0041] (3) As shown in Figure 5 , after the original data is corrected, it is re-pushed into the average wind speed index job, and the wind speed and direction index job is based on a real-time operator and uses a delay trigger mechanism to perform index calculation. Wherein the x data stream is the north-south direction wind speed time history, the y data stream is the east-west direction wind speed time history, and the z data stream is the vertical direction wind speed time history. The average wind speed index is triggered once every whole minute, and the average wind speed index job is based on a sliding window and an incremental calculation method, and the real-time wind speed index uses a delay trigger mechanism to obtain the mean aggregation value of the x data stream and the y data stream, performs data time alignment and splicing, and calculates the average wind speed index. The same as step (2), based on the overall data stream data driving, a 1-minute waiting delay is set, as shown in Figure 3 , the 10-minute mean value stored the last time in the memory 1 minute ago is obtained, including the mean value of the data stream x and the mean value of the data stream y, the mean value calculation method is the same as step (1), after time alignment and splicing, square root calculation is performed, the time alignment means that the 10-minute interval of the mean value of x and the 10-minute interval of the mean value of y are consistent, and the calculation result is stored in the memory database, indicating the average wind speed in the 10-minute interval, and the index calculation result is pushed to the downstream message middleware in the format of {id, name, starttime, endtime, data}. The calculation formula of the average wind speed U is:

[0042]

[0043] where, u x represents the north-south direction wind speed time history, represents u x the mean of the past 10 minutes data, u y represents the east-west direction wind speed time history, represents u y the mean of the past 10 minutes data.

[0044] In steps (2) and (3), according to the characteristics of the bridge health monitoring scene, the delay trigger mechanism introduces expert experience, and sets the data processing delay trigger based on the driving clock. When the order of the arrival of the bridge health monitoring data is out of order, the data that is produced earlier but has not yet arrived in the system can be appropriately waited for, so as to ensure that the data before a certain time is basically arrived before calculation, pushing and time alignment splicing of multi-flow data are performed. Among them, the overall data flow is used as the driving clock, the latest data time of the overall data flow is used as the synchronization clock of all operators and indicators, and all operators and indicators are driven to calculate and process data.

[0045] (4) As shown in Figure 6 , after the original data is corrected, it will be pushed into the fluctuating wind speed index job again. The fluctuating wind speed will trigger a calculation for each data entering the job, which is the same as step (2). Based on the overall data flow after data correction, data driving is performed, and a 3-minute waiting delay is set to obtain the data 3 minutes ago from the in-memory database. The specific data is that, with the current overall data time as a reference, the x, y, z data stream corrected original data 3 minutes ago, the mean values of the x, y and z data streams stored 3 minutes ago, and the average wind speed index value are obtained from the memory. The mean value calculation method is the same as step (1), and data time alignment and splicing are performed. Among them, the data time alignment refers to the data time of the x, y, z data stream original data being consistent, the time interval of the mean value and the average wind speed index being the same, and they being the latest results stored in the in-memory database at the same time, and the calculation interval containing the original data data time extracted in this step. For example, the current data time is 13:06, the alignment rule is that the original data of the x, y and z data streams are all 10:06, the interval of the x data stream mean value, the y data stream mean value, the z data stream mean value and the average wind speed is 01:00-11:00, after time alignment and splicing, the trigonometric function, inverse trigonometric function and division calculation are performed to obtain the fluctuating wind speed values of the main wind direction, the cross wind direction and the vertical direction at 10:06, and the calculation results are stored in the in-memory database, and at the same time, the {id, name, starttime, endtime, data} structure is pushed into the downstream message middleware. The calculation formulas of the main wind direction fluctuating wind speed u, the cross wind direction fluctuating wind speed v and the vertical direction fluctuating wind speed w are as follows:

[0046] Main wind direction fluctuating wind speed u:

[0047]

[0048] Crosswind pulsating speed v:

[0049]

[0050] Vertical pulsating wind speed w:

[0051]

[0052] Where u z This represents the raw data of the z-data stream. This represents the average value of the z-data stream over the past 10 minutes.

[0053] (5) Figure 7 As shown, after the pulsed wind speed calculation is completed, the turbulence intensity index operation will be pushed back into the operation. In this operation, the pulsed wind speed will be pushed to the real-time variance operator for incremental calculation and data aggregation, the same as step (1). This operation triggers a calculation once every whole minute, the same as step (2). Based on the overall data stream after data correction, data is driven and a 4-minute waiting delay is set to retrieve the data from the memory database 4 minutes ago. Specifically, with the overall data time of the current data as a reference, the 10-minute variance value and average wind speed index of the main wind direction, crosswind direction and vertical direction pulsed wind speed 4 minutes ago are obtained from memory. After time alignment and splicing, square root and division calculations are performed, and the index calculation results are pushed to the downstream message middleware in the format {id,name,starttime,endtime,data}. The time alignment rule is that the time intervals of the variance and average wind speed index of the original data are the same. For example: Given that the current overall data stream time is 14:00, the average wind speed and the variance of fluctuating wind speed from 0:00 to 10:00 are extracted to calculate the turbulence intensity index for the 0:00 to 10:00 interval. Turbulence intensity I in the prevailing wind direction. u Crosswind turbulence intensity I v and vertical turbulence intensity I w The calculation formulas are as follows:

[0054]

[0055] In the formula σ u σ v σ w The standard deviations of the 10-minute spans for the fluctuating wind speeds u in the prevailing wind direction, v in the crosswind direction, and w in the vertical wind direction are obtained by taking the square root of the variance value provided by the real-time variance operator.

[0056] In steps (2), (3), and (4), the aggregation intermediate state of the real-time operator and the bridge structure health monitoring index calculation result are stored using a memory database.

[0057] In another aspect, the present application also provides a real-time wind speed index calculation system for bridge structure health monitoring, which comprises a data processing module, a real-time stream calculation module, an index calculation logic module, and an index job extraction module.

[0058] The data processing module is used for real-time extraction and analysis of the bridge sensor data cached in the upstream message middleware, and pushes the data to the real-time stream calculation module in the format of (id, timestamp, data); the specific process of the module refers to the steps of the real-time wind speed index calculation method for bridge structure health monitoring provided by the present application.

[0059] The real-time stream calculation module is used for data aggregation of the bridge sensor data by the real-time operator in the real-time stream calculation engine through incremental calculation and sliding window; after the abnormal data is modified to be normal through the abnormality detection algorithm, the data is stored for a short time, and each real-time wind speed index job and the real-time operator are pushed for calculation; the specific process of the module refers to the steps of the real-time wind speed index calculation method for bridge structure health monitoring provided by the present application.

[0060] The index calculation logic module is used for extraction of the data stored in the memory database according to the calculation logic of the real-time wind speed index job for bridge structure health monitoring, design of a delay trigger mechanism based on data time, data alignment, calculation, and reorganization, and storage of the calculation result in the memory database and pushing to the downstream message middleware; the specific process of the module refers to the steps of the real-time wind speed index calculation method for bridge structure health monitoring provided by the present application.

[0061] The index job extraction module is used for extraction of the calculation result stored in the memory database according to the index calculation logic requirement of the index calculation logic module, and continues to execute the index calculation logic module until the index calculation logic is completed, and the index calculation result is pushed to the downstream message middleware; the specific process of the module refers to the steps of the real-time wind speed index calculation method for bridge structure health monitoring provided by the present application.

[0062] The real-time wind speed index calculation method and system for bridge structure health monitoring provided by the present application perform incremental calculation based on sensor time series data, greatly reduce the calculation complexity and read-write IO time delay of complex indexes by saving intermediate aggregation values to a memory database, fundamentally increase the throughput of bridge structure health monitoring index calculation, reduce system calculation time delay, and can be widely applied in the field of bridge structure health monitoring real-time index calculation.

[0063] The above embodiments are only used for illustrating the present application, but not for limiting the present application. Those skilled in the art can make various changes and modifications without departing from the spirit and scope of the present application. Therefore, all equivalent technical solutions belong to the scope of the present application, and the patent protection scope of the present application should be defined by the claims.

Claims

1. A method for calculating real-time wind speed indicators for bridge structural health monitoring, the method comprising the following steps: (1) The collected bridge sensor data is cached in real time in the upstream message middleware. When it is necessary to calculate the wind speed index, the bridge sensor data cached in the upstream message middleware is extracted and parsed in real time and pushed to the real-time stream computing engine in the format of (id, timestamp, data). (2) The real-time operators in the real-time stream computing engine use incremental computing and sliding windows to aggregate bridge sensor data; The data aggregation results of the real-time operator are detected by an anomaly detection algorithm. After the abnormal data is corrected to normal, it is stored for a short period of time and then pushed to the real-time wind speed index operation and the real-time operator for calculation. Main wind direction, pulsating wind speed u: in, This indicates the time history of wind speed in the north-south direction. express The average of the data over the past 10 minutes, The time history of wind speed in the east-west direction. express The average of the data over the past 10 minutes, where U represents the average wind speed; Crosswind pulsating speed v: Vertical pulsating wind speed w: in This represents the raw data of the z-data stream. This represents the average value of the z-data stream over the past 10 minutes. (3) The real-time wind speed index operation for bridge structural health monitoring is based on the calculation logic and a delayed triggering mechanism designed based on data time. The data stored in the memory database is extracted, data is aligned, calculated and reorganized, and the calculation results are stored in the memory database and pushed to the downstream message middleware. (4) According to the requirements of the indicator calculation logic, the calculation results stored in the memory database will be extracted by the indicator job and continue to execute step (3) until the indicator calculation logic is completed and the indicator calculation results are pushed to the downstream message middleware.

2. The method for calculating real-time wind speed index for bridge structural health monitoring according to claim 1, characterized in that, In step (2), when the real-time operator uses a sliding window to aggregate bridge data, it performs incremental calculations driven by data. The time span of the sliding window increases continuously as data enters. When the aggregated data in the sliding window reaches the maximum span, it is stored and cleared.

3. The method for calculating real-time wind speed index for bridge structural health monitoring according to claim 1, characterized in that, In step (2), the real-time wind speed indicators include the pulsating wind speed in the main wind direction, the pulsating wind speed in the crosswind direction, the pulsating wind speed in the vertical direction, the average wind speed, and the turbulence intensity.

4. The method for calculating real-time wind speed index for bridge structural health monitoring according to claim 2, characterized in that, For each bridge sensor data channel, the real-time operator maintains multiple independent sliding windows, each sliding window maintains an intermediate state value, and the starting time interval of adjacent sliding windows is the same; when new bridge sensor data is extracted from the upstream message middleware and enters the stream computing engine, all sliding windows are calculated simultaneously.

5. The method for calculating real-time wind speed index for bridge structural health monitoring according to claim 1, characterized in that, In step (3), based on the characteristics of the bridge health monitoring scenario, the delayed triggering mechanism incorporates expert experience and sets the data processing delay triggering based on the driving clock.

6. The method for calculating real-time wind speed index for bridge structural health monitoring according to claim 1, characterized in that, In step (2), the 3sigma or Z-score anomaly detection algorithm is used to monitor data anomalies based on recent historical data and to modify the abnormal data.

7. The method for calculating real-time wind speed index for bridge structural health monitoring according to claim 1, characterized in that, In step (3), the delayed triggering mechanism drives the real-time wind speed index operation, and the real-time wind speed index is calculated using real-time operators according to the index operation calculation logic.

8. The method for calculating real-time wind speed index for bridge structural health monitoring according to claim 1, characterized in that, In steps (3) and (4), the data format for pushing downstream message middleware is {id,name,starttime,endtime,data}.

9. A real-time wind speed index calculation system for bridge structural health monitoring, implementing the method of any one of claims 1-8, characterized in that, The system includes a data processing module, a real-time stream computing module, an indicator calculation logic module, and an indicator job extraction module; The data processing module is used to extract and parse the bridge sensor data cached in the upstream message middleware in real time, and push it to the real-time stream computing module in the format of (id, timestamp, data). The real-time stream computing module is used to aggregate bridge sensor data using incremental computing and sliding windows through real-time operators in the real-time stream computing engine. The data aggregation results of the real-time operators are detected by an anomaly detection algorithm. After the abnormal data is corrected to normal, it is stored for a short period of time. At the same time, the real-time wind speed indicators and real-time operators are pushed for calculation. The index calculation logic module is used to perform real-time wind speed index operation of bridge structural health monitoring according to the calculation logic, based on the data time design delay trigger mechanism, extract the data stored in the memory database, perform data alignment, calculation and reorganization, and store the calculation results in the memory database and push them to the downstream message middleware. The indicator job extraction module is used to extract the calculation results stored in the memory database according to the indicator calculation logic requirements of the indicator calculation logic module, and continue to execute the indicator calculation logic module until the indicator calculation logic is completed, and then push the indicator calculation results to the downstream message middleware.

Images