Bridge safety assessment method and system based on BIM bridge safety monitoring system

Abstract

The invention relates to the technical field of bridge engineering, and discloses a bridge safety assessment method and system based on a BIM bridge safety monitoring system, and the method comprises the steps: obtaining the real-time vehicle weight data in a bridge through a sensor module in the BIM bridge safety monitoring system, carrying out the data cleaning of the multi-modal bridge real-time data, and obtaining the real-time vehicle weight data of the bridge; the method comprises the following steps of: cleaning data, performing feature extraction on the cleaned data, fusing extracted feature bridge data, establishing a two-dimensional convolutional neural network model, and optimizing model parameters of the two-dimensional convolutional neural network model added with an attention mechanism by using an improved whale optimization algorithm, and inputting the fused bridge real-time data into the bridge safety evaluation model for identification, inputting the real-time bridge safety state into the BIM bridge safety monitoring system for block color display, and performing color distinguishing according to different safety levels. Management personnel can be helped to quickly understand the current state of the bridge and make scientific and reasonable maintenance decisions.

Description

Technical Field

[0001] The present invention relates to the technical field of bridge engineering, and in particular to a bridge safety assessment method and system based on a BIM bridge safety monitoring system. Background Art

[0002] With the continuous development of transportation infrastructure, bridges, as important transportation hubs, are of vital importance for their safety. Although BIM technology in existing technologies can integrate data from the entire life cycle of bridges and realize three-dimensional visualization, this method usually has the problems of scattered data, low visualization, and the analysis is not combined with the real-time load-bearing conditions of the bridge. Therefore, how to effectively combine BIM with the safety assessment algorithm of bridges to form a set of accurate and efficient assessment methods is an urgent problem to be solved. Summary of the invention

[0003] The purpose of the present invention is to solve the above problems and design a bridge safety assessment method and system based on BIM bridge safety monitoring system.

[0004] A first aspect of the present invention provides a bridge safety assessment method based on a BIM bridge safety monitoring system, the bridge safety assessment method based on a BIM bridge safety monitoring system comprising the following steps: Based on the sensor module in the BIM bridge safety monitoring system, real-time vehicle weight data in the bridge is obtained, and the real-time vehicle weight data is judged. If the real-time vehicle weight data is greater than the bridge load-bearing threshold, multi-modal bridge real-time data is obtained; Performing data cleaning on the multimodal bridge real-time data, performing feature extraction on the cleaned data, and fusing the extracted feature bridge data to obtain fused bridge real-time data; A 2D-CNN two-dimensional convolutional neural network model is established, an Attention mechanism is added to the 2D-CNN two-dimensional convolutional neural network model, and the model parameters of the 2D-CNN two-dimensional convolutional neural network model after adding the attention mechanism are optimized using an improved IWOA whale optimization algorithm to obtain an IWOA-Attention-2D-CNN bridge safety assessment model; The fused bridge real-time data is input into the IWOA-Attention-2D-CNN bridge safety assessment model for identification to obtain the real-time bridge safety status; The real-time bridge safety status is input into the BIM bridge safety monitoring system for block color display, and the colors are differentiated according to different safety levels.

[0005] Optionally, in a first implementation of the first aspect of the present invention, the sensor module in the BIM-based bridge safety monitoring system obtains real-time vehicle weight data in the bridge, and judges the real-time vehicle weight data. If the real-time vehicle weight data is greater than the bridge load-bearing threshold, multi-modal bridge real-time data is obtained, including: Acquire real-time vehicle weight data in the bridge based on the sensor module in the BIM bridge safety monitoring system, wherein the sensor module includes at least an axle weight sensor, a strain sensor, a displacement sensor, an image sensor, an acceleration sensor, a piezoelectric sensor, and a temperature and humidity sensor; Assumptions , ,..., is real-time vehicle weight data from bridge sensors of independent samples, each independent sample represents data within a period of time, then:

[0006] in, yes The maximum value among the sample values, yes The minimum value among the sample values, then:

[0007] in, represents the probability symbol, Indicates the maximum value of the sample Less than or equal to a value The probability of Indicates the minimum value of the sample Less than or equal to a value probability; is the distribution function of the population, describing the random variable Value less than or equal to The probability of is the overall distribution function of Power, in In, it means when there is When there are independent and identically distributed random variables, their maximum value is less than or equal to The probability is equal to the population distribution function of Power, Indicates that the random variable is greater than The probability of express At least one of the sample values ​​is less than or equal to probability; The statistical threshold value is calculated by the formula The distribution is estimated based on the Fisher-Tippett extreme value type theorem. The threshold extreme value of follows the generalized Pareto distribution, and the calculation formula is as follows:

[0008] in, represents the shape parameter, when When , it is represented as Pareto type I distribution. When When , it is represented as Pareto type III distribution. represents the standard deviation of the Pareto distribution, Indicates the maximum load-bearing capacity of the bridge; and according to Determining the real-time vehicle weight data; If the real-time vehicle weight data is greater than the bridge load-bearing threshold, the multimodal bridge real-time data in the bridge is obtained through the sensor module in the BIM bridge safety monitoring system. The multimodal bridge real-time data at least includes bridge strain data, bridge displacement data, vehicle acceleration data, bridge temperature and humidity data, bridge deck image data and bridge structure image data.

[0009] Optionally, in a second implementation of the first aspect of the present invention, the data cleaning of the multimodal bridge real-time data includes: The outliers of the bridge strain data in the multimodal bridge real-time data are detected by using a linear interpolation method, and reasonable estimation and replacement are performed according to the trend of adjacent normal data points. The outliers of the vehicle acceleration data are determined by combining the vehicle model and the number of axles with the average speed range of the vehicle in the historical data; the outliers of the multimodal bridge real-time data are removed to obtain normal multimodal bridge real-time data; Based on the ARIMA time series model, the missing values ​​in the normal multimodal bridge real-time data are filled to obtain complete multimodal bridge real-time data. The AR autoregressive part of the ARIMA time series model assumes that the time series is , the autoregressive order is , then the expression of the autoregressive part is: ,in is the autoregressive coefficient; The difference part of the ARIMA time series model, the difference order is , by performing The first-order difference is converted into a stationary time series. The calculation formula of the first-order difference is , the second-order difference is to differentiate the sequence after the first-order difference again, that is, ; The sliding average part of the ARIMA time series model has a sliding average order of , whose expression is ,in is a white noise sequence, is the sliding mean coefficient; Combining the AR autoregressive part, the difference part and the moving average part of the ARIMA time series model, the ARIMA ( , , ) The calculation formula of the model is: +

[0010] in, The mean is 0 and the variance is White noise sequence.

[0011] Optionally, in a third implementation of the first aspect of the present invention, the cleaned data is subjected to feature extraction, and the extracted feature bridge data is fused to obtain fused bridge real-time data, including: Acquire the complete multimodal bridge real-time data after data cleaning, extract the daily variation characteristics of temperature in the complete multimodal bridge real-time data, including the daily maximum temperature, daily minimum temperature, daily average temperature and temperature change rate; the mean, variance, maximum and minimum values ​​of relative humidity, and calculate the correlation coefficient between humidity and temperature; the maximum value of wind speed, average wind speed, standard deviation of wind speed, main direction of wind direction and distribution frequency of wind direction, and obtain environmental data characteristic data; Extract the mean, variance, maximum and minimum values ​​of the strain data in the complete multimodal bridge real-time data; extract the mean, variance, kurtosis and skewness of the acceleration in the time domain characteristics of the acceleration data; extract the peak value, peak-to-peak value and root mean square value of the time domain characteristics of the displacement data to obtain the structural response data characteristic data; Calculate the daily flow, hourly flow and peak-hour flow of traffic, and analyze the periodic characteristics of traffic flow to obtain traffic load characteristic data; Obtain image data of key structural parts of the bridge in the complete multimodal bridge real-time data, identify the image data using the YOLOv5 image detection algorithm, obtain crack data and aging data of the key structural parts of the bridge, and obtain characteristic data of the key structural parts of the bridge; After normalizing the environmental data characteristic data, the structural response data characteristic data, the traffic load characteristic data and the bridge key structural component characteristic data, the fused bridge real-time data is obtained, and the fused bridge real-time data at least includes the environmental data characteristic data , structural response data characteristic data , traffic load characteristic data and characteristic data of key structural parts of bridges .

[0012] Optionally, in a fourth implementation manner of the first aspect of the present invention, establishing a 2D-CNN two-dimensional convolutional neural network model, and adding an Attention mechanism to the 2D-CNN two-dimensional convolutional neural network model, includes: Establish a 2D-CNN two-dimensional convolutional neural network model, and the input layer receives the preprocessed two-dimensional data matrix ,matrix The dimension is , each element of the input Indicated in In the time step The normalized value of the fused data; The first convolutional layer uses the convolution kernel , the size is , the number of convolution kernels is , the step length is , filled with , for the input two-dimensional data matrix The convolution formula is:

[0013] in, is the output of the first convolutional layer exist The value of the position, is the first convolution kernel exist The weight of the position, is the bias term; using RELU activation function , nonlinearly change the output of the first convolutional layer, that is , and get the output after activation ; Add a second convolutional layer using the product kernel , the size is , the number of convolution kernels is , the step length is , filled with , the formula is:

[0014] in, is the output of the second convolutional layer exist The value of the position, is the first convolution kernel exist The weight of the position, is the bias term; using RELU activation function , nonlinearly change the output of the second convolutional layer, that is , and get the output after activation ; Assume the pooling window size is , the step length is , for the output of the activated convolutional layer , the pooling operation formula is:

[0015] After multiple convolutional and pooling layers, the feature map is flattened into a one-dimensional vector and output through a fully connected layer; Add the Attention mechanism to the 2D-CNN two-dimensional convolutional neural network model, assuming that the output of the pooling layer Extract Query, Key and Value, and use different weight matrices through linear transformation , , , the dimensions are , , ,in Is the dimension of the input feature, calculate the Query 、Key and Value :

[0016]

[0017]

[0018] The similarity is calculated by the dot product of Query and Key, and then the Attention weight is obtained by Softmax normalization. and Elements and , Attention weight The calculation formula is as follows:

[0019] in, is the dimension of Key, represents the number of time windows, represents the number of data points in a time window, represents the exponential function, Indicates the first The elements at the positions, Indicates the first The elements at the positions are weighted and summed using the calculated Attention weights to get the output of the Attention mechanism.

[0020] Optionally, in a fifth implementation of the first aspect of the present invention, the improved IWOA whale optimization algorithm is used to optimize the model parameters of the 2D-CNN two-dimensional convolutional neural network model after adding the attention mechanism to obtain the IWOA-Attention-2D-CNN bridge safety assessment model, including: The improved IWOA whale optimization algorithm is used to optimize the model parameters of the 2D-CNN two-dimensional convolutional neural network model after adding the attention mechanism; The mean square error function is set as the objective function of the 2D-CNN two-dimensional convolutional neural network model after adding the attention mechanism ,set up is a bridge safety indicator. is the model prediction result. samples, the objective function The expression is:

[0021] The convolution kernel weights, biases, and fully connected layer weights of the 2D-CNN two-dimensional convolutional neural network model are used as individual positions in the whale optimization algorithm , for each individual Assign it to the corresponding parameters of the 2D-CNN two-dimensional convolutional neural network model, use the training data for forward propagation to obtain the prediction results, and calculate each individual Fitness ; Update parameters and update individual positions according to the whale position update formula , find the current optimal individual , Represents the number of iterations. When the number of iterations is 120, the optimal individual is assigned to the 2D-CNN two-dimensional convolutional neural network model.

[0022] Optionally, in a fifth implementation of the first aspect of the present invention, the improved IWOA whale optimization algorithm includes: The Tent chaotic map is used to optimize the initialization population of the WOA whale optimization algorithm in the search space to improve the optimization efficiency and solution accuracy. The WOA whale optimization algorithm population is classified using the K-means++ clustering algorithm, and a whale is randomly selected from the whale population and its position coordinates are used as the initial cluster center. ; Calculate the remaining whales in the population and The Euclidean distance , No. The probability of the first whale being selected as the next cluster center is: , select the next cluster center by roulette wheel method ; Repeat the iteration to get Cluster centers; Calculate the individual whales in the population except the cluster center and Clustering of cluster centers, dividing the remaining whale individuals into the center point with the smallest distance In the subsets where the center points of each subset are located, the iteration is repeated until the result does not change, and the improved IWOA whale optimization algorithm is obtained.

[0023] The second aspect of the present invention provides a bridge safety assessment system based on a BIM bridge safety monitoring system, wherein the bridge safety assessment system based on a BIM bridge safety monitoring system comprises a data analysis and judgment module, a data cleaning and fusion module, an assessment model establishment module, a bridge safety assessment module, and a bridge status display module, wherein: A data analysis and judgment module is used to obtain real-time vehicle weight data in the bridge based on the sensor module in the BIM bridge safety monitoring system, and to judge the real-time vehicle weight data. If the real-time vehicle weight data is greater than the bridge load-bearing threshold, multi-modal bridge real-time data is obtained; A data cleaning and fusion module is used to clean the multi-modal bridge real-time data, extract features from the cleaned data, and fuse the extracted feature bridge data to obtain fused bridge real-time data; An evaluation model establishment module is used to establish a 2D-CNN two-dimensional convolutional neural network model, add an Attention mechanism to the 2D-CNN two-dimensional convolutional neural network model, and use an improved IWOA whale optimization algorithm to optimize the model parameters of the 2D-CNN two-dimensional convolutional neural network model after adding the attention mechanism, so as to obtain an IWOA-Attention-2D-CNN bridge safety evaluation model; A bridge safety assessment module, used for inputting the fused bridge real-time data into the IWOA-Attention-2D-CNN bridge safety assessment model for identification, and obtaining a real-time bridge safety status; The bridge status display module is used to input the real-time bridge safety status into the BIM bridge safety monitoring system for block color display and to distinguish colors according to different safety levels.

[0024] Optionally, in a first implementation of the second aspect of the present invention, the data analysis and judgment module includes the following submodules: An acquisition submodule is used to acquire real-time vehicle weight data in the bridge based on a sensor module in the BIM bridge safety monitoring system, wherein the sensor module includes at least an axle weight sensor, a strain sensor, a displacement sensor, an image sensor, an acceleration sensor, a piezoelectric sensor, and a temperature and humidity sensor; Computational submodule, used to assume , ,..., is real-time vehicle weight data from bridge sensors of independent samples, each independent sample represents data within a period of time, then:

[0025] in, yes The maximum value among the sample values, yes The minimum value among the sample values, then:

[0026] in, represents the probability symbol, Indicates the maximum value of the sample Less than or equal to a value The probability of Indicates the minimum value of the sample Less than or equal to a value The probability of is the distribution function of the population, describing the random variable Value less than or equal to The probability of is the overall distribution function of Power, in In, it means when there is When there are independent and identically distributed random variables, their maximum value is less than or equal to The probability is equal to the population distribution function of Power, Indicates that the random variable is greater than The probability of express At least one of the sample values ​​is less than or equal to The probability of The estimation submodule is used to calculate the threshold value through the statistics of the formula The distribution is estimated based on the Fisher-Tippett extreme value type theorem. The threshold extreme value of follows the generalized Pareto distribution, and the calculation formula is as follows:

[0027] in, represents the shape parameter, when When , it is represented as Pareto type I distribution. When When , it is represented as Pareto type III distribution. represents the standard deviation of the Pareto distribution, Indicates the maximum load-bearing capacity of the bridge; and according to Determining the real-time vehicle weight data; The judgment submodule is used to judge if the real-time vehicle weight data is greater than the bridge load-bearing threshold, and then obtain the multimodal bridge real-time data in the bridge through the sensor module in the BIM bridge safety monitoring system, and the multimodal bridge real-time data at least includes bridge strain data, bridge displacement data, vehicle acceleration data, bridge temperature and humidity data, bridge deck image data and bridge structure image data.

[0028] Optionally, in a second implementation of the second aspect of the present invention, the data cleaning and fusion module includes the following submodules: The acquisition submodule is used to acquire the complete multimodal bridge real-time data after data cleaning, extract the daily variation characteristics of temperature in the complete multimodal bridge real-time data, including the daily maximum temperature, daily minimum temperature, daily average temperature and temperature change rate; the mean, variance, maximum and minimum values ​​of relative humidity, and calculate the correlation coefficient between humidity and temperature; the maximum value of wind speed, average wind speed, standard deviation of wind speed, main direction of wind direction and distribution frequency of wind direction, and obtain environmental data characteristic data; The extraction submodule is used to extract the mean, variance, maximum and minimum values ​​of the strain data in the complete multimodal bridge real-time data; extract the mean, variance, kurtosis and skewness of the acceleration in the time domain characteristics of the acceleration data; extract the peak value, peak-to-peak value and root mean square value of the time domain characteristics of the displacement data to obtain the structural response data characteristic data; The analysis submodule is used to calculate the daily flow, hourly flow and peak-hour flow of traffic, and analyze the periodic characteristics of traffic flow to obtain traffic load characteristic data; An identification submodule is used to obtain image data of key structural parts of the bridge in the complete multimodal bridge real-time data, identify the image data using the YOLOv5 image detection algorithm, obtain crack data and aging data of the key structural parts of the bridge, and obtain characteristic data of the key structural parts of the bridge; The normalization submodule is used to normalize the environmental data characteristic data, the structural response data characteristic data, the traffic load characteristic data and the bridge key structural component characteristic data to obtain the fused bridge real-time data, wherein the fused bridge real-time data at least includes the environmental data characteristic data. , structural response data characteristic data , traffic load characteristic data and characteristic data of key structural parts of bridges .

[0029] The technical solution provided by the present invention has the following beneficial effects: 1. Real-time monitoring and early warning of the weight overload of the bridge: The technical solution obtains vehicle weight data in real time through the sensor module installed on the bridge, and makes an immediate judgment on the overload situation. When the bridge load-bearing threshold is detected, the system automatically collects multi-modal bridge real-time data, including structural stress, displacement and other parameters. It enables managers to respond quickly to abnormal conditions and take necessary restrictions or maintenance measures, thereby effectively preventing the occurrence of safety accidents. 2. Data processing and feature fusion: For the collected multi-modal bridge real-time data, the system performs data cleaning and feature extraction to ensure the quality and reliability of the data. Subsequently, feature fusion is used to obtain information that more comprehensively reflects the state of the bridge, which helps to improve the accuracy of subsequent analysis. It not only enhances the availability of data, but also reduces the waste of computing resources due to redundant information. 3. Model optimization and intelligent evaluation: This solution uses 2D-CNN combined with the Attention mechanism to construct a bridge safety assessment model, and uses the improved IWOA whale optimization algorithm to optimize the model parameters. The Attention mechanism makes the model pay more attention to the features that are critical to bridge safety; and the IWOA algorithm ensures the effectiveness and efficiency of parameter adjustment during model training. It not only improves the accuracy of the model, but also speeds up the model convergence and shortens the training time. 4. Visual feedback and decision support: The system integrates the identified bridge safety status into the BIM platform in the form of block color display, and uses different colors to distinguish and display according to different safety levels. It can help managers quickly understand the current status of the bridge and make scientific and reasonable maintenance decisions. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the following detailed description of the preferred embodiment.The drawings are only for the purpose of illustrating the preferred embodiments and are not to be construed as limiting the invention.

[0031] Figure 1 A schematic diagram of a first embodiment of a bridge safety assessment method based on a BIM bridge safety monitoring system provided by an embodiment of the present invention; Figure 2A schematic diagram of a second embodiment of a bridge safety assessment method based on a BIM bridge safety monitoring system provided by an embodiment of the present invention; Figure 3 A schematic diagram of a third embodiment of a bridge safety assessment method based on a BIM bridge safety monitoring system provided by an embodiment of the present invention; Figure 4 A structural schematic diagram of a bridge safety assessment system based on a BIM bridge safety monitoring system provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0032] The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, device, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0033] For ease of understanding, the specific process of the embodiment of the present invention is described below. Figure 1 A first embodiment schematic diagram of a bridge safety assessment method based on a BIM bridge safety monitoring system provided by an embodiment of the present invention, the method specifically comprises the following steps: Step 101: acquiring real-time vehicle weight data in the bridge based on the sensor module in the BIM bridge safety monitoring system, judging the real-time vehicle weight data, and acquiring multi-modal bridge real-time data if the real-time vehicle weight data is greater than the bridge load-bearing threshold; Specifically, in this embodiment, the BIM bridge safety monitoring system includes a sensor module, an integrated display module, a data processing module and a model recognition module. The real-time data of the bridge, including vehicle weight data and multimodal bridge real-time data, is obtained through the sensor module during operation, and the data is input into the data processing module for data processing. After the data processing, the data is input into the drawing model recognition module, and the recognition and evaluation results are input into the integrated display module.

[0034] Specifically, in this embodiment, the real-time vehicle weight data in the bridge is obtained based on the sensor module in the BIM bridge safety monitoring system, and the sensor module includes at least an axle weight sensor, a strain sensor, a displacement sensor, an image sensor, an acceleration sensor, a piezoelectric sensor, and a temperature and humidity sensor; Assumptions , ,..., is real-time vehicle weight data from bridge sensors of independent samples, each independent sample represents data within a period of time, then:

[0035] in, yes The maximum value among the sample values, yes The minimum value among the sample values, then:

[0036] in, represents the probability symbol, Indicates the maximum value of the sample Less than or equal to a value The probability of Indicates the minimum value of the sample Less than or equal to a value The probability of is the distribution function of the population, describing the random variable Value less than or equal to The probability of is the overall distribution function of Power, in In, it means when there is When there are independent and identically distributed random variables, their maximum value is less than or equal to The probability is equal to the population distribution function of Power, Indicates that the random variable is greater than The probability of express At least one of the sample values ​​is less than or equal to The probability of The statistical threshold value is calculated by the formula The distribution is estimated based on the Fisher-Tippett extreme value type theorem. The threshold extreme value of follows the generalized Pareto distribution, and the calculation formula is as follows:

[0037] in, represents the shape parameter, when When , it is represented as Pareto type I distribution. When When , it is represented as Pareto type III distribution. represents the standard deviation of the Pareto distribution, Indicates the maximum load-bearing capacity of the bridge; and according to Judge the real-time vehicle weight data; If the real-time vehicle weight data is greater than the bridge load-bearing threshold, the multimodal bridge real-time data in the bridge is obtained through the sensor module in the BIM bridge safety monitoring system. The multimodal bridge real-time data at least includes bridge strain data, bridge displacement data, vehicle acceleration data, bridge temperature and humidity data, bridge deck image data and bridge structure image data.

[0038] Specifically, the bridge load-bearing threshold includes the following aspects: Highway bridges (for ordinary vehicles): For newly built highway bridges, the lane load standard is generally in accordance with the provisions of the "Technical Standards for Highway Engineering" (JTG B01-2014): a six-axle truck with an axle weight not exceeding 13 tons and a total weight not exceeding 55 tons is used as the design load.

[0039] Urban bridges (including urban overpasses for light rail or subway): Bridges in cities are usually designed to accommodate pedestrians, bicycles, and various types of motor vehicles, including buses and small trucks. For example, the weight limit for single-axle vehicles on urban bridges is about 20 tons.

[0040] Railway bridges: up to 30 tons per axle or more. For heavy freight railway bridges, the design load may be higher, such as some special railway bridges used for coal transportation, which need to withstand loads of 35 tons or even heavier per axle.

[0041] Pedestrian bridges: These bridges are designed for loads of approximately 250 to 500 kg per square metre (75 to 150 kg per person), depending on the expected density of pedestrian traffic.

[0042] Special purpose bridges: Floating bridges or temporary bridges for military purposes carrying a weight of more than 70 tons.

[0043] Step 102: cleaning the multimodal bridge real-time data, extracting features from the cleaned data, and fusing the extracted feature bridge data to obtain fused bridge real-time data; Specifically, in this embodiment, the linear interpolation method is used to detect the abnormal values ​​of the bridge strain data in the multi-modal bridge real-time data, and reasonable estimation and replacement are performed according to the trend of adjacent normal data points. The abnormal values ​​of the vehicle acceleration data are judged by combining the vehicle model and the number of axles with the average speed range of the vehicle in the historical data; the abnormal values ​​of the multi-modal bridge real-time data are removed to obtain normal multi-modal bridge real-time data; Based on the ARIMA time series model, the missing values ​​in the normal multimodal bridge real-time data are filled to obtain the complete multimodal bridge real-time data. The AR autoregressive part of the ARIMA time series model assumes that the time series is , the autoregressive order is , then the expression of the autoregressive part is: ,in is the autoregressive coefficient; The difference part of the ARIMA time series model, the difference order is , by performing The first-order difference is converted into a stationary time series. The calculation formula of the first-order difference is , the second-order difference is to differentiate the sequence after the first-order difference again, that is, ; The sliding average part of the ARIMA time series model has a sliding average order of , whose expression is ,in is a white noise sequence, is the sliding mean coefficient; Combining the AR autoregressive part, the difference part and the moving average part of the ARIMA time series model, the ARIMA ( , , ) The calculation formula of the model is: +

[0044] in, The mean is 0 and the variance is White noise sequence.

[0045] Specifically, the present embodiment also includes obtaining complete multimodal bridge real-time data after data cleaning, extracting daily temperature variation characteristics in the complete multimodal bridge real-time data, including daily maximum temperature, daily minimum temperature, daily average temperature and temperature change rate; the mean, variance, maximum and minimum values ​​of relative humidity, and calculating the correlation coefficient between humidity and temperature; the maximum value of wind speed, average wind speed, standard deviation of wind speed, main direction of wind direction and distribution frequency of wind direction, and obtaining environmental data characteristic data; Extract the mean, variance, maximum and minimum values ​​of strain data in the complete multi-modal bridge real-time data; extract the mean, variance, kurtosis and skewness of acceleration in the time domain characteristics of acceleration data; extract the peak, peak-to-peak value and root mean square value of the time domain characteristics of displacement data to obtain the characteristic data of structural response data; Calculate the daily flow, hourly flow and peak-hour flow of traffic, and analyze the periodic characteristics of traffic flow to obtain traffic load characteristic data; Obtain image data of key structural parts of bridges in the complete multimodal real-time bridge data, use the YOLOv5 image detection algorithm to identify the image data, obtain crack data and aging data of key structural parts of bridges, and obtain characteristic data of key structural parts of bridges; After normalizing the environmental data characteristic data, the structural response data characteristic data, the traffic load characteristic data and the bridge key structural component characteristic data, the fused bridge real-time data is obtained. The fused bridge real-time data at least includes the environmental data characteristic data. , structural response data characteristic data , traffic load characteristic data and characteristic data of key structural parts of bridges . Two-dimensional data matrix It is the integration of real-time bridge data.

[0046] Step 103: Establish a 2D-CNN two-dimensional convolutional neural network model, add an Attention mechanism to the 2D-CNN two-dimensional convolutional neural network model, and use the improved IWOA whale optimization algorithm to optimize the model parameters of the 2D-CNN two-dimensional convolutional neural network model after adding the attention mechanism, so as to obtain an IWOA-Attention-2D-CNN bridge safety assessment model; Specifically, in this embodiment, a 2D-CNN two-dimensional convolutional neural network model is established, and the input layer receives the preprocessed two-dimensional data matrix ,matrix The dimension is , each element of the input Indicated in In the time step The normalized value of the fused data; The first convolutional layer uses the convolution kernel , the size is , the number of convolution kernels is , the step length is , filled with , for the input two-dimensional data matrix The convolution formula is:

[0047] in, is the output of the first convolutional layer exist The value of the position, is the first convolution kernel exist The weight of the position, is the bias term; using RELU activation function , nonlinearly change the output of the first convolutional layer, that is , and get the output after activation ; Add a second convolutional layer using the product kernel , the size is , the number of convolution kernels is , the step length is , filled with , the formula is:

[0048] in, is the output of the second convolutional layer exist The value of the position, is the first convolution kernel exist The weight of the position, is the bias term; using RELU activation function , nonlinearly change the output of the second convolutional layer, that is , and get the output after activation ; Assume the pooling window size is , the step length is , for the output of the activated convolutional layer , the pooling operation formula is:

[0049] After multiple convolutional and pooling layers, the feature map is flattened into a one-dimensional vector and output through a fully connected layer; Add the Attention mechanism to the 2D-CNN two-dimensional convolutional neural network model, assuming that the output of the pooling layer Extract Query, Key and Value, and use different weight matrices through linear transformation , , , the dimensions are , , ,in Is the dimension of the input feature, calculate the Query 、Key and Value :

[0050]

[0051]

[0052] The similarity is calculated by the dot product of Query and Key, and then the Attention weight is obtained by Softmax normalization. and Elements and , Attention weight The calculation formula is as follows:

[0053] in, is the dimension of Key, represents the number of time windows, represents the number of data points in a time window, represents the exponential function, Indicates the first The elements at the positions, Indicates the first The elements at the positions are weighted and summed using the calculated Attention weights to get the output of the Attention mechanism.

[0054] Specifically, in this embodiment, the improved IWOA whale optimization algorithm is used to optimize the model parameters of the 2D-CNN two-dimensional convolutional neural network model after adding the attention mechanism, and the IWOA-Attention-2D-CNN bridge safety assessment model is obtained, including: The improved IWOA whale optimization algorithm is used to optimize the model parameters of the 2D-CNN two-dimensional convolutional neural network model after adding the attention mechanism; The mean square error function is set as the objective function of the 2D-CNN two-dimensional convolutional neural network model after adding the attention mechanism ,set up is a bridge safety indicator. is the model prediction result. samples, the objective function The expression is:

[0055] The convolution kernel weights, biases, and fully connected layer weights of the 2D-CNN two-dimensional convolutional neural network model are used as individual positions in the whale optimization algorithm , for each individual Assign it to the corresponding parameters of the 2D-CNN two-dimensional convolutional neural network model, use the training data for forward propagation to obtain the prediction results, and calculate each individual Fitness ; Update parameters and update individual positions according to the whale position update formula , find the current optimal individual , Represents the number of iterations. When the number of iterations is 120, the optimal individual is assigned to the 2D-CNN two-dimensional convolutional neural network model.

[0056] Specifically, the improved IWOA whale optimization algorithm in this embodiment includes: The Tent chaotic map is used to optimize the initialization population of the WOA whale optimization algorithm in the search space to improve the optimization efficiency and solution accuracy. The K-means++ clustering algorithm is used to classify the WOA whale optimization algorithm population, and a whale is randomly selected from the whale population and its position coordinates are used as the initial cluster center. ; Calculate the remaining whales in the population and The Euclidean distance , No. The probability of the first whale being selected as the next cluster center is: , select the next cluster center by roulette wheel method ; Repeat the iteration to get Cluster centers; Calculate the individual whales in the population except the cluster center and Clustering of cluster centers, dividing the remaining whale individuals into the center point with the smallest distance In the subsets where the center points of each subset are located, the iteration is repeated until the result does not change, and the improved IWOA whale optimization algorithm is obtained.

[0057] Tent chaos map is a simple and typical chaos map, which is a one-dimensional chaos map. Using Tent chaos map to initialize the position of whales can make whales more evenly distributed in the search space, thereby improving the ability of whale algorithm to search for the global optimal solution. Its mathematical expression is: when hour, ;when hour, , It is The parameter values ​​of the whale optimization algorithm for the iteration, Yes, yes The parameter value of the whale optimization algorithm for the iteration; the Tent chaotic map is obtained by Iterate continuously to generate a series of random but deterministic numerical sequences. If the initial value , in the first iteration, because ,so . In the second iteration, because ,so .

[0058] Step 104: input the fused bridge real-time data into the IWOA-Attention-2D-CNN bridge safety assessment model for identification to obtain the real-time bridge safety status; Specifically, the integrated bridge real-time data includes at least environmental data and characteristic data. , structural response data characteristic data , traffic load characteristic data and characteristic data of key structural parts of bridges . Two-dimensional data matrix It is a fusion of real-time bridge data. Arranged in a two-dimensional matrix according to certain rules In the above, the different data are concatenated into a vector, and then these vectors are arranged into a matrix, for example, the concatenated vectors of multiple time steps are stacked to form a matrix ,matrix The dimension is expressed as ,in is the time step, is the length of the fused data in each time step.

[0059] The input layer of the IWOA-Attention-2D-CNN bridge safety assessment model receives the preprocessed two-dimensional data matrix , the dimension is . Each element of the input Indicated in In the time step The normalized value of the fused data.

[0060] Step 105: Input the real-time bridge safety status into the BIM bridge safety monitoring system for block color display, and differentiate colors according to different safety levels.

[0061] Specifically, in this embodiment, the system converts the real-time bridge safety status into an easy-to-understand form, such as a color-coded map block display. Green represents normal operation, yellow indicates that attention is needed, and red means that there is a potential danger and an immediate response is required. The visual interface is intuitive and practical, greatly improving communication efficiency.

[0062] The technical solution provided by the present invention has the following beneficial effects: 1. Real-time monitoring and early warning of the weight overload of the bridge: The technical solution obtains vehicle weight data in real time through the sensor module installed on the bridge, and makes an immediate judgment on the overload situation. When the bridge load-bearing threshold is detected, the system automatically collects multi-modal bridge real-time data, including structural stress, displacement and other parameters. It enables managers to respond quickly to abnormal conditions and take necessary restrictions or maintenance measures, thereby effectively preventing the occurrence of safety accidents. 2. Data processing and feature fusion: For the collected multi-modal bridge real-time data, the system performs data cleaning and feature extraction to ensure the quality and reliability of the data. Subsequently, feature fusion is used to obtain information that more comprehensively reflects the state of the bridge, which helps to improve the accuracy of subsequent analysis. It not only enhances the availability of data, but also reduces the waste of computing resources due to redundant information. 3. Model optimization and intelligent evaluation: This solution uses 2D-CNN combined with the Attention mechanism to construct a bridge safety assessment model, and uses the improved IWOA whale optimization algorithm to optimize the model parameters. The Attention mechanism allows the model to pay more attention to features that are critical to bridge safety; and the IWOA algorithm ensures the effectiveness and efficiency of parameter adjustment during model training. It not only improves the accuracy of the model, but also speeds up the model convergence and shortens the training time. 4. Visual feedback and decision support: The system integrates the identified bridge safety status into the BIM platform in the form of block color display, and uses different colors to distinguish and display according to different safety levels. It can help managers quickly understand the current status of the bridge and make scientific and reasonable maintenance decisions.

[0063] See also Figure 2 , a schematic diagram of a second embodiment of a bridge safety assessment method based on a BIM bridge safety monitoring system provided by an embodiment of the present invention, wherein the method performs feature extraction on the cleaned data, fuses the extracted feature bridge data, and obtains fused bridge real-time data, including: Step 1021, obtain the complete multimodal bridge real-time data after data cleaning, extract the daily variation characteristics of temperature in the complete multimodal bridge real-time data, including the daily maximum temperature, daily minimum temperature, daily average temperature and temperature change rate; the mean, variance, maximum and minimum values ​​of relative humidity, and calculate the correlation coefficient between humidity and temperature; the maximum value of wind speed, average wind speed, standard deviation of wind speed, main direction of wind direction and distribution frequency of wind direction, and obtain environmental data characteristic data; Step 1022, extracting the mean, variance, maximum and minimum values ​​of strain data in the complete multi-modal bridge real-time data; extracting the mean, variance, kurtosis and skewness of acceleration in the time domain characteristics of acceleration data; extracting the peak value, peak-to-peak value and root mean square value of the time domain characteristics of displacement data to obtain structural response data characteristic data; Step 1023, calculating the daily flow, hourly flow and peak-hour flow of traffic flow, and analyzing the periodic characteristics of traffic flow to obtain traffic load characteristic data; Step 1024, obtaining image data of key structural parts of the bridge in the complete multimodal bridge real-time data, using the YOLOv5 image detection algorithm to identify the image data, obtaining crack data and aging data of the key structural parts of the bridge, and obtaining characteristic data of the key structural parts of the bridge; Step 1025: normalize the environmental data characteristic data, the structural response data characteristic data, the traffic load characteristic data and the bridge key structural component characteristic data to obtain the fused bridge real-time data, which at least includes the environmental data characteristic data. , structural response data characteristic data , traffic load characteristic data and characteristic data of key structural parts of bridges .

[0064] The beneficial effects are as follows: 1. By accurately extracting multiple time-domain features of strain, acceleration and displacement data, the dynamic response of the bridge under different conditions can be fully reflected. Characteristic parameters are crucial for evaluating the health of bridge structures, which helps to detect potential problems earlier and improve early warning capabilities. 2. Detailed statistical analysis of traffic flow is carried out. In-depth analysis of traffic flow characteristics provides a basis for predicting future traffic patterns, and also helps to identify potential safety hazards caused by overload. 3. The advanced YOLOv5 image detection algorithm is used to automatically identify cracks and aging of key structural parts of the bridge, which improves work efficiency, reduces human errors, and ensures the consistency and accuracy of detection, which is of great significance for timely maintenance measures. 4. The environmental factors, structural response, traffic loads, and the status of key structural parts are standardized and integrated to form a comprehensive real-time bridge data set. Ensure that all relevant information can be effectively integrated to avoid the one-sidedness or limitations that may be caused by a single data source.

[0065] See also Figure 3 , a schematic diagram of a third embodiment of a bridge safety assessment method based on a BIM bridge safety monitoring system provided by an embodiment of the present invention, wherein the improved IWOA whale optimization algorithm in the method comprises: Step 1031, using Tent chaotic mapping to optimize the optimization efficiency and solution accuracy of the initialization population of the WOA whale optimization algorithm in the search space; Step 1032: Use the K-means++ clustering algorithm to classify the WOA whale optimization algorithm population, and randomly select a whale from the whale population and use its position coordinates as the initial cluster center; Among them, the initial cluster center is set to .

[0066] Step 1033: Calculate the Euclidean distance between the remaining whales in the population and the initial cluster center, select the next cluster center by the roulette wheel method, repeat the iteration, and obtain Cluster centers; Among them, the remaining whales in the population and the initial cluster center are calculated The Euclidean distance , No. The probability of the first whale being selected as the next cluster center is: , select the next cluster center by roulette wheel method ; Repeat the iteration to get Cluster centers.

[0067] Step 1034: Calculate the number of whales in the population except for the cluster center. The remaining whales are divided into subsets with the smallest distance to the center point, and the center point of each subset is calculated. The iteration is repeated until the result does not change, and the improved IWOA whale optimization algorithm is obtained.

[0068] Among them, let the remaining whale individuals be divided into the center with the smallest distance .

[0069] The beneficial effect is that by using the Tent chaotic map to initialize the position of the whale, the whale can be distributed more evenly in the search space, thereby improving the ability of the whale algorithm to search for the global optimal solution and improving the convergence speed of the algorithm.

[0070] The K-means++ clustering algorithm is used to classify the WOA whale optimization algorithm population. For the elite group, the convergence factor decreases rapidly in the early stage of iteration, which helps the whale individuals with lower fitness to quickly approach the optimal value. The convergence factor decreases slowly in the later stage of iteration, which helps to avoid falling into the local optimum. For the poor individuals, the convergence factor decreases slowly in the early stage of iteration, which helps the whale individuals with higher fitness to find the best in a larger range. It decreases rapidly in the later stage of iteration to complete the convergence. For the ordinary group, the convergence is completed according to the convergence factor of WOA.

[0071] See also Figure 4 , a structural schematic diagram of a bridge safety assessment system based on a BIM bridge safety monitoring system provided by an embodiment of the present invention, the system includes a data analysis and judgment module, a data cleaning and fusion module, an assessment model establishment module, a bridge safety assessment module, and a bridge status display module, wherein, The data analysis and judgment module 401 is used to obtain the real-time vehicle weight data in the bridge based on the sensor module in the BIM bridge safety monitoring system, judge the real-time vehicle weight data, and obtain the multi-modal bridge real-time data if the real-time vehicle weight data is greater than the bridge load-bearing threshold; Specifically, this embodiment also includes the following submodules: An acquisition submodule is used to acquire real-time vehicle weight data in the bridge based on a sensor module in the BIM bridge safety monitoring system, where the sensor module includes at least an axle weight sensor, a strain sensor, a displacement sensor, an image sensor, an acceleration sensor, a piezoelectric sensor, and a temperature and humidity sensor; Computational submodule, used to assume , ,..., is real-time vehicle weight data from bridge sensors of independent samples, each independent sample represents data within a period of time, then:

[0072] in, yes The maximum value among the sample values, yes The minimum value among the sample values, then:

[0073] in, represents the probability symbol, Indicates the maximum value of the sample Less than or equal to a value The probability of Indicates the minimum value of the sample Less than or equal to a value probability; is the distribution function of the population, describing the random variable Value less than or equal to The probability of is the overall distribution function of Power, in In, it means when there is When there are independent and identically distributed random variables, their maximum value is less than or equal to The probability is equal to the population distribution function of Power, Indicates that the random variable is greater than The probability of express At least one of the sample values ​​is less than or equal to probability; The estimation submodule is used to calculate the threshold value through the statistics of the formula The distribution is estimated based on the Fisher-Tippett extreme value type theorem. The threshold extreme value of follows the generalized Pareto distribution, and the calculation formula is as follows:

[0074] in, represents the shape parameter, when When , it is represented as Pareto type I distribution. When When , it is represented as Pareto type III distribution. represents the standard deviation of the Pareto distribution, Indicates the maximum load-bearing capacity of the bridge; and according to Judge the real-time vehicle weight data; The judgment submodule is used to judge if the real-time vehicle weight data is greater than the bridge load-bearing threshold, and then obtain the multimodal bridge real-time data in the bridge through the sensor module in the BIM bridge safety monitoring system. The multimodal bridge real-time data at least includes bridge strain data, bridge displacement data, vehicle acceleration data, bridge temperature and humidity data, bridge deck image data and bridge structure image data.

[0075] The data cleaning and fusion module 402 is used to clean the multi-modal bridge real-time data, extract features from the cleaned data, and fuse the extracted feature bridge data to obtain fused bridge real-time data; Specifically, this embodiment also includes the following submodules: The detection submodule is used to detect the abnormal values ​​of the bridge strain data in the multi-modal bridge real-time data by using the linear interpolation method, reasonably estimate and replace according to the trend of the adjacent normal data points, and judge the abnormal values ​​of the vehicle acceleration data by the vehicle model and the number of axles combined with the average speed range of the vehicle in the historical data; remove the abnormal values ​​of the multi-modal bridge real-time data to obtain normal multi-modal bridge real-time data; The filling submodule is used to fill the missing values ​​in the normal multimodal bridge real-time data based on the ARIMA time series model to obtain complete multimodal bridge real-time data. The model building submodule is used for the AR autoregressive part of the ARIMA time series model. Assume that the time series is , the autoregressive order is , then the expression of the autoregressive part is: ,in is the autoregressive coefficient; The difference part of the ARIMA time series model, the difference order is , by performing The first-order difference is converted into a stationary time series. The calculation formula of the first-order difference is , the second-order difference is to differentiate the sequence after the first-order difference again, that is, ; The sliding average part of the ARIMA time series model has a sliding average order of , whose expression is ,in is a white noise sequence, is the sliding mean coefficient; Combining the AR autoregressive part, the difference part and the moving average part of the ARIMA time series model, the ARIMA ( , , ) The calculation formula of the model is: +

[0076] in, The mean is 0 and the variance is White noise sequence.

[0077] Specifically, this embodiment also includes the following submodules: The acquisition submodule is used to obtain the complete multimodal bridge real-time data after data cleaning, extract the daily variation characteristics of temperature in the complete multimodal bridge real-time data, including the daily maximum temperature, daily minimum temperature, daily average temperature and temperature change rate; the mean, variance, maximum and minimum values ​​of relative humidity, and calculate the correlation coefficient between humidity and temperature; the maximum value of wind speed, average wind speed, standard deviation of wind speed, main direction of wind direction and distribution frequency of wind direction, and obtain the characteristic data of environmental data; The extraction submodule is used to extract the mean, variance, maximum and minimum values ​​of strain data in the complete multimodal bridge real-time data; extract the mean, variance, kurtosis and skewness of acceleration in the time domain characteristics of acceleration data; extract the peak, peak-to-peak value and root mean square value of the time domain characteristics of displacement data to obtain the characteristic data of structural response data; The analysis submodule is used to calculate the daily flow, hourly flow and peak-hour flow of traffic, and analyze the periodic characteristics of traffic flow to obtain traffic load characteristic data; The recognition submodule is used to obtain the image data of the key structural parts of the bridge in the complete multimodal bridge real-time data, use the YOLOv5 image detection algorithm to recognize the image data, obtain the crack data and aging data of the key structural parts of the bridge, and obtain the characteristic data of the key structural parts of the bridge; The normalization submodule is used to normalize the environmental data characteristic data, structural response data characteristic data, traffic load characteristic data and bridge key structural component characteristic data to obtain the fused bridge real-time data. Among them, the integrated bridge real-time data includes at least environmental data and characteristic data , structural response data characteristic data , traffic load characteristic data and characteristic data of key structural parts of bridges .

[0078] An assessment model establishment module 403 is used to establish a 2D-CNN two-dimensional convolutional neural network model, add an Attention mechanism to the 2D-CNN two-dimensional convolutional neural network model, and use an improved IWOA whale optimization algorithm to optimize the model parameters of the 2D-CNN two-dimensional convolutional neural network model after adding the attention mechanism, so as to obtain an IWOA-Attention-2D-CNN bridge safety assessment model; Specifically, this embodiment also includes the following submodules: Establish a submodule for establishing a 2D-CNN two-dimensional convolutional neural network model. The input layer receives the preprocessed two-dimensional data matrix ,matrix The dimension is , each element of the input Indicated in In the time step The normalized value of the fused data; The first convolutional layer uses the convolution kernel , the size is , the number of convolution kernels is , the step length is , filled with , for the input two-dimensional data matrix The convolution formula is:

[0079] in, is the output of the first convolutional layer exist The value of the position, is the first convolution kernel exist The weight of the position, is the bias term; using RELU activation function , nonlinearly change the output of the first convolutional layer, that is , and get the output after activation ; Add a submodule to add a second convolutional layer using a product kernel , the size is , the number of convolution kernels is , the step length is , filled with , the formula is:

[0080] in, is the output of the second convolutional layer exist The value of the position, is the first convolution kernel exist The weight of the position, is the bias term; using RELU activation function , make nonlinear changes to the output of the second convolutional layer, that is , and get the output after activation ; Set the submodule to set the pooling window size to , the step length is , for the output of the activated convolutional layer , the pooling operation formula is:

[0081] After multiple convolutional and pooling layers, the feature map is flattened into a one-dimensional vector and output through a fully connected layer; The attention mechanism submodule is used to add the Attention mechanism to the 2D-CNN two-dimensional convolutional neural network model. Extract Query, Key and Value, and use different weight matrices through linear transformation , , , the dimensions are , , ,in Is the dimension of the input feature, calculate the Query 、Key and Value :

[0082]

[0083]

[0084] The calculation submodule is used to calculate the similarity through the dot product of Query and Key, and then obtain the Attention weight through Softmax normalization. and Elements and , Attention weight The calculation formula is as follows:

[0085] in, is the dimension of Key, represents the number of time windows, represents the number of data points in a time window, represents the exponential function, Indicates the first The elements at the positions, Indicates the first The elements at the positions are weighted and summed using the calculated Attention weights to get the output of the Attention mechanism.

[0086] Specifically, this embodiment also includes the following submodules: The optimization submodule is used to optimize the model parameters of the 2D-CNN two-dimensional convolutional neural network model after adding the attention mechanism using the improved IWOA whale optimization algorithm; Add a submodule to set the mean square error function as the objective function of the 2D-CNN two-dimensional convolutional neural network model after adding the attention mechanism ,set up is a bridge safety indicator. is the model prediction result. samples, the objective function The expression is:

[0087] Set up submodules for using the convolution kernel weights, biases, and fully connected layer weights of the 2D-CNN two-dimensional convolutional neural network model as individual positions in the whale optimization algorithm , for each individual Assign it to the corresponding parameters of the 2D-CNN two-dimensional convolutional neural network model, use the training data for forward propagation to obtain the prediction results, and calculate each individual Fitness ; The update submodule is used to update parameters and update individual positions according to the whale position update formula. , find the current optimal individual , Represents the number of iterations. When the number of iterations is 120, the optimal individual is assigned to the 2D-CNN two-dimensional convolutional neural network model.

[0088] Specifically, this embodiment also includes the following submodules: The optimization submodule is used to optimize the optimization efficiency and solution accuracy of the initialization population of the WOA whale optimization algorithm in the search space by using the Tent chaotic map; The classification submodule is used to classify the WOA whale optimization algorithm population using the K-means++ clustering algorithm, and randomly select a whale from the whale population and use its position coordinates as the initial cluster center. ; Iterator module, used to calculate the remaining whales in the population and The Euclidean distance , No. The probability of the first whale being selected as the next cluster center is: , select the next cluster center by roulette wheel method ; Repeat the iteration to get Cluster centers; Get the submodule, which is used to calculate the individual whales in the population except the cluster center and Clustering of cluster centers, dividing the remaining whale individuals into the center point with the smallest distance In the subsets where the center points of each subset are located, the iteration is repeated until the result does not change, and the improved IWOA whale optimization algorithm is obtained.

[0089] The bridge safety assessment module 404 is used to input the fused bridge real-time data into the IWOA-Attention-2D-CNN bridge safety assessment model for identification, and obtain the real-time bridge safety status; Specifically, the integrated bridge real-time data includes at least environmental data and characteristic data. , structural response data characteristic data , traffic load characteristic data and characteristic data of key structural parts of bridges . Two-dimensional data matrix It is a fusion of real-time bridge data. Arranged in a two-dimensional matrix according to certain rules In the above, the different data are concatenated into a vector, and then these vectors are arranged into a matrix, for example, the concatenated vectors of multiple time steps are stacked to form a matrix ,matrix The dimension is expressed as ,in is the time step, is the length of the fused data in each time step.

[0090] The input layer of the IWOA-Attention-2D-CNN bridge safety assessment model receives the preprocessed two-dimensional data matrix , the dimension is . Each element of the input Indicated in In the time step The normalized value of the fused data.

[0091] The bridge status display module 405 is used to input the real-time bridge safety status into the BIM bridge safety monitoring system for block color display and to distinguish colors according to different safety levels.

[0092] Specifically, in this embodiment, the system converts the real-time bridge safety status into an easy-to-understand form, such as a color-coded map block display. Green represents normal operation, yellow indicates that attention is needed, and red means that there is a potential danger and an immediate response is required. The visual interface is intuitive and practical, greatly improving communication efficiency.

[0093] The beneficial effect is that the real-time vehicle weight data in the bridge is obtained by the sensor module in the BIM bridge safety monitoring system, and the real-time vehicle weight data is judged. If the real-time vehicle weight data is greater than the bridge load-bearing threshold, the multimodal bridge real-time data is obtained; the multimodal bridge real-time data is cleaned, and the features are extracted from the cleaned data, and the extracted feature bridge data is fused to obtain the fused bridge real-time data; a 2D-CNN two-dimensional convolutional neural network model is established, and the Attention mechanism is added to the 2D-CNN two-dimensional convolutional neural network model, and the model parameters of the 2D-CNN two-dimensional convolutional neural network model after adding the attention mechanism are optimized by using the improved IWOA whale optimization algorithm to obtain the IWOA-Attention-2D-CNN bridge safety assessment model; the fused bridge real-time data is input into the IWOA-Attention-2D-CNN bridge safety assessment model for identification to obtain the real-time bridge safety status; the real-time bridge safety status is input into the BIM bridge safety monitoring system for block color display, and the colors are distinguished according to different safety levels. Make an immediate judgment on overload conditions. When a bridge load threshold is exceeded, the system automatically collects multimodal real-time bridge data, including structural stress, displacement and other parameters. This enables managers to respond quickly to abnormal conditions and take necessary restrictions or maintenance measures, thereby effectively preventing safety accidents. Feature fusion can provide more comprehensive information on the status of the bridge, which helps to improve the accuracy of subsequent analysis. It not only enhances the availability of data, but also reduces the waste of computing resources caused by redundant information. It improves the accuracy of the model, speeds up model convergence and shortens training time. Different colors are used to distinguish and display according to different safety levels. It can help managers quickly understand the current status of the bridge and make scientific and reasonable maintenance decisions. It also facilitates public supervision and improves transparency.

[0094] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments. The above embodiments and descriptions are only preferred examples of the present invention and are not intended to limit the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. A bridge safety assessment method based on a BIM bridge safety monitoring system, characterized in that: The bridge safety assessment method based on the BIM bridge safety monitoring system comprises the following steps: Based on the sensor module in the BIM bridge safety monitoring system, real-time vehicle weight data in the bridge is obtained, and the real-time vehicle weight data is judged. If the real-time vehicle weight data is greater than the bridge load-bearing threshold, multi-modal bridge real-time data is obtained; Performing data cleaning on the multimodal bridge real-time data, performing feature extraction on the cleaned data, and fusing the extracted feature bridge data to obtain fused bridge real-time data; A 2D-CNN two-dimensional convolutional neural network model is established, an Attention mechanism is added to the 2D-CNN two-dimensional convolutional neural network model, and the model parameters of the 2D-CNN two-dimensional convolutional neural network model after adding the attention mechanism are optimized using an improved IWOA whale optimization algorithm to obtain an IWOA-Attention-2D-CNN bridge safety assessment model; The fused bridge real-time data is input into the IWOA-Attention-2D-CNN bridge safety assessment model for identification to obtain the real-time bridge safety status; The real-time bridge safety status is input into the BIM bridge safety monitoring system for block color display, and the colors are differentiated according to different safety levels.

2. A bridge safety assessment method based on a BIM bridge safety monitoring system as claimed in claim 1, characterized in that: The sensor module in the BIM-based bridge safety monitoring system obtains real-time vehicle weight data in the bridge, and judges the real-time vehicle weight data. If the real-time vehicle weight data is greater than the bridge load-bearing threshold, multi-modal bridge real-time data is obtained, including: Acquire real-time vehicle weight data in the bridge based on the sensor module in the BIM bridge safety monitoring system, wherein the sensor module includes at least an axle weight sensor, a strain sensor, a displacement sensor, an image sensor, an acceleration sensor, a piezoelectric sensor, and a temperature and humidity sensor; Assumptions , ,..., is real-time vehicle weight data from bridge sensors of independent samples, each independent sample represents data within a period of time, then: ;in, yes The maximum value among the sample values, yes The minimum value among the sample values, then: ;in, represents the probability symbol, Indicates the maximum value of the sample Less than or equal to a value The probability of Indicates the minimum value of the sample Less than or equal to a value probability; is the distribution function of the population, describing the random variable Value less than or equal to The probability of is the overall distribution function of Power, in In, it means when there is When there are independent and identically distributed random variables, their maximum value is less than or equal to The probability is equal to the population distribution function of Power, Indicates that the random variable is greater than The probability of express At least one of the sample values ​​is less than or equal to probability; The statistical threshold value is calculated by the formula The distribution is estimated based on the Fisher-Tippett extreme value type theorem. The threshold extreme value of follows the generalized Pareto distribution, and the calculation formula is as follows: ;in, represents the shape parameter, when When , it is represented as Pareto type I distribution. When When , it is represented as Pareto type III distribution. represents the standard deviation of the Pareto distribution, Indicates the maximum load-bearing capacity of the bridge; and according to Determining the real-time vehicle weight data; If the real-time vehicle weight data is greater than the bridge load-bearing threshold, the multimodal bridge real-time data in the bridge is obtained through the sensor module in the BIM bridge safety monitoring system. The multimodal bridge real-time data at least includes bridge strain data, bridge displacement data, vehicle acceleration data, bridge temperature and humidity data, bridge deck image data and bridge structure image data.

3. A bridge safety assessment method based on a BIM bridge safety monitoring system as claimed in claim 1, characterized in that: The data cleaning of the multi-modal bridge real-time data includes: The outliers of the bridge strain data in the multimodal bridge real-time data are detected by using a linear interpolation method, and reasonable estimation and replacement are performed according to the trend of adjacent normal data points. The outliers of the vehicle acceleration data are determined by combining the vehicle model and the number of axles with the average speed range of the vehicle in the historical data; the outliers of the multimodal bridge real-time data are removed to obtain normal multimodal bridge real-time data; Based on the ARIMA time series model, the missing values ​​in the normal multimodal bridge real-time data are filled to obtain complete multimodal bridge real-time data. The AR autoregressive part of the ARIMA time series model assumes that the time series is , the autoregressive order is , then the expression of the autoregressive part is: ,in is the autoregressive coefficient; The difference part of the ARIMA time series model, the difference order is , by performing The first-order difference is converted into a stationary time series. The calculation formula of the first-order difference is , the second-order difference is to differentiate the sequence after the first-order difference again, that is, ; The sliding average part of the ARIMA time series model has a sliding average order of , whose expression is ,in is a white noise sequence, is the sliding mean coefficient; Combining the AR autoregressive part, the difference part and the moving average part of the ARIMA time series model, the ARIMA ( , , ) The calculation formula of the model is: + ;in, The mean is 0 and the variance is White noise sequence.

4. A bridge safety assessment method based on a BIM bridge safety monitoring system as claimed in claim 1, characterized in that: The cleaned data is subjected to feature extraction, and the extracted feature bridge data is fused to obtain fused bridge real-time data, including: Acquire the complete multimodal bridge real-time data after data cleaning, extract the daily variation characteristics of temperature in the complete multimodal bridge real-time data, including the daily maximum temperature, daily minimum temperature, daily average temperature and temperature change rate; the mean, variance, maximum and minimum values ​​of relative humidity, and calculate the correlation coefficient between humidity and temperature; the maximum value of wind speed, average wind speed, standard deviation of wind speed, main direction of wind direction and distribution frequency of wind direction, and obtain environmental data characteristic data; Extract the mean, variance, maximum and minimum values ​​of the strain data in the complete multimodal bridge real-time data; extract the mean, variance, kurtosis and skewness of the acceleration in the time domain characteristics of the acceleration data; extract the peak value, peak-to-peak value and root mean square value of the time domain characteristics of the displacement data to obtain the structural response data characteristic data; Calculate the daily flow, hourly flow and peak-hour flow of traffic, and analyze the periodic characteristics of traffic flow to obtain traffic load characteristic data; Obtain image data of key structural parts of the bridge in the complete multimodal bridge real-time data, identify the image data using the YOLOv5 image detection algorithm, obtain crack data and aging data of the key structural parts of the bridge, and obtain characteristic data of the key structural parts of the bridge; After normalizing the environmental data characteristic data, the structural response data characteristic data, the traffic load characteristic data and the bridge key structural component characteristic data, the fused bridge real-time data is obtained, and the fused bridge real-time data at least includes the environmental data characteristic data , structural response data characteristic data , traffic load characteristic data and characteristic data of key structural parts of bridges .

5. The bridge safety assessment method based on the BIM bridge safety monitoring system according to claim 1, characterized in that: The step of establishing a 2D-CNN two-dimensional convolutional neural network model and adding an Attention mechanism to the 2D-CNN two-dimensional convolutional neural network model includes: Establish a 2D-CNN two-dimensional convolutional neural network model, and the input layer receives the preprocessed two-dimensional data matrix ,matrix The dimension is , each element of the input Indicated in In the time step The normalized value of the fused data; The first convolutional layer uses the convolution kernel , the size is , the number of convolution kernels is , the step length is , filled with , for the input two-dimensional data matrix The convolution formula is: ;in, is the output of the first convolutional layer exist The value of the position, is the first convolution kernel exist The weight of the position, is the bias term; using RELU activation function , make nonlinear changes to the output of the first convolutional layer, that is , and get the output after activation ; Add a second convolutional layer using the product kernel , the size is , the number of convolution kernels is , the step length is , filled with , the formula is: ;in, is the output of the second convolutional layer exist The value of the position, is the first convolution kernel exist The weight of the position, is the bias term; using RELU activation function , nonlinearly change the output of the second convolutional layer, that is , and get the output after activation ; Assume the pooling window size is , the step length is , for the output of the activated convolutional layer , the pooling operation formula is: ; After multiple convolutional layers and pooling layers, the feature map is flattened into a one-dimensional vector and output through a fully connected layer; Add the Attention mechanism to the 2D-CNN two-dimensional convolutional neural network model, assuming that the output of the pooling layer Extract Query, Key and Value, and use different weight matrices through linear transformation , , , the dimensions are , , ,in Is the dimension of the input feature, calculate the Query 、Key and Value : ; ; ; Calculate the similarity by the dot product of Query and Key, and then get the Attention weight through Softmax normalization. and Elements and , Attention weight The calculation formula is as follows: ;in, is the dimension of Key, represents the number of time windows, represents the number of data points in a time window, represents the exponential function, Indicates the first The elements at the positions, Indicates the first The elements at the positions are weighted and summed using the calculated Attention weights to get the output of the Attention mechanism.

6. A bridge safety assessment method based on a BIM bridge safety monitoring system as claimed in claim 1, characterized in that: The improved IWOA whale optimization algorithm is used to optimize the model parameters of the 2D-CNN two-dimensional convolutional neural network model after adding the attention mechanism, and the IWOA-Attention-2D-CNN bridge safety assessment model is obtained, including: The improved IWOA whale optimization algorithm is used to optimize the model parameters of the 2D-CNN two-dimensional convolutional neural network model after adding the attention mechanism; The mean square error function is set as the objective function of the 2D-CNN two-dimensional convolutional neural network model after adding the attention mechanism ,set up is a bridge safety indicator. is the model prediction result. samples, the objective function The expression is: ; The convolution kernel weights, biases, and fully connected layer weights of the 2D-CNN two-dimensional convolutional neural network model are used as individual positions in the whale optimization algorithm , for each individual Assign it to the corresponding parameters of the 2D-CNN two-dimensional convolutional neural network model, use the training data for forward propagation to obtain the prediction results, and calculate each individual Fitness ; Update parameters and update individual positions according to the whale position update formula , find the current optimal individual , Represents the number of iterations. When the number of iterations is 120, the optimal individual is assigned to the 2D-CNN two-dimensional convolutional neural network model.

7. A bridge safety assessment method based on a BIM bridge safety monitoring system as claimed in claim 6, characterized in that: The improved IWOA whale optimization algorithm includes: The Tent chaotic map is used to optimize the optimization efficiency and solution accuracy of the initial population of the WOA whale optimization algorithm in the search space; The WOA whale optimization algorithm population is classified using the K-means++ clustering algorithm, and a whale is randomly selected from the whale population and its position coordinates are used as the initial cluster center. ; Calculate the remaining whales in the population and The Euclidean distance , No. The probability of the first whale being selected as the next cluster center is: , select the next cluster center by roulette wheel method ; Repeat the iteration to get Cluster centers; Calculate the individual whales in the population except the cluster center and Clustering of cluster centers, dividing the remaining whale individuals into the center point with the smallest distance In the subsets where the center points of each subset are located, the iteration is repeated until the result does not change, and the improved IWOA whale optimization algorithm is obtained.

8. A bridge safety assessment system based on a BIM bridge safety monitoring system, characterized in that: The bridge safety assessment system based on the BIM bridge safety monitoring system includes a data analysis and judgment module, a data cleaning and fusion module, an assessment model establishment module, a bridge safety assessment module, and a bridge status display module, wherein: A data analysis and judgment module is used to obtain real-time vehicle weight data in the bridge based on the sensor module in the BIM bridge safety monitoring system, and to judge the real-time vehicle weight data. If the real-time vehicle weight data is greater than the bridge load-bearing threshold, multi-modal bridge real-time data is obtained; A data cleaning and fusion module is used to clean the multi-modal bridge real-time data, extract features from the cleaned data, and fuse the extracted feature bridge data to obtain fused bridge real-time data; An evaluation model establishment module is used to establish a 2D-CNN two-dimensional convolutional neural network model, add an Attention mechanism to the 2D-CNN two-dimensional convolutional neural network model, and use an improved IWOA whale optimization algorithm to optimize the model parameters of the 2D-CNN two-dimensional convolutional neural network model after adding the attention mechanism, so as to obtain an IWOA-Attention-2D-CNN bridge safety evaluation model; A bridge safety assessment module, used for inputting the fused bridge real-time data into the IWOA-Attention-2D-CNN bridge safety assessment model for identification, and obtaining a real-time bridge safety status; The bridge status display module is used to input the real-time bridge safety status into the BIM bridge safety monitoring system for block color display and to distinguish colors according to different safety levels.

9. A bridge safety assessment system based on a BIM bridge safety monitoring system as claimed in claim 8, characterized in that: The data analysis and judgment module includes the following submodules: An acquisition submodule is used to acquire real-time vehicle weight data in the bridge based on a sensor module in the BIM bridge safety monitoring system, wherein the sensor module includes at least an axle weight sensor, a strain sensor, a displacement sensor, an image sensor, an acceleration sensor, a piezoelectric sensor, and a temperature and humidity sensor; Computational submodule, used to assume , ,..., is real-time vehicle weight data from bridge sensors of independent samples, each independent sample represents data within a period of time, then: ;in, yes The maximum value among the sample values, yes The minimum value among the sample values, then: ;in, represents the probability symbol, Indicates the maximum value of the sample Less than or equal to a value The probability of Indicates the minimum value of the sample Less than or equal to a value probability; is the distribution function of the population, describing the random variable Value less than or equal to The probability of is the overall distribution function of Power, in In, it means when there is When there are independent and identically distributed random variables, their maximum value is less than or equal to The probability is equal to the population distribution function of Power, Indicates that the random variable is greater than The probability of express At least one of the sample values ​​is less than or equal to The probability of The estimation submodule is used to calculate the threshold value through the statistics of the formula The distribution is estimated based on the Fisher-Tippett extreme value type theorem. The threshold extreme value of follows the generalized Pareto distribution, and the calculation formula is as follows: ;in, represents the shape parameter, when When , it is represented as Pareto type I distribution. When When , it is represented as Pareto type III distribution. represents the standard deviation of the Pareto distribution, Indicates the maximum load-bearing capacity of the bridge; and according to Determining the real-time vehicle weight data; The judgment submodule is used to judge if the real-time vehicle weight data is greater than the bridge load-bearing threshold, and then obtain the multimodal bridge real-time data in the bridge through the sensor module in the BIM bridge safety monitoring system, and the multimodal bridge real-time data at least includes bridge strain data, bridge displacement data, vehicle acceleration data, bridge temperature and humidity data, bridge deck image data and bridge structure image data.

10. A bridge safety assessment system based on a BIM bridge safety monitoring system as claimed in claim 8, characterized in that: The data cleaning and fusion module includes the following submodules: The acquisition submodule is used to acquire the complete multimodal bridge real-time data after data cleaning, extract the daily variation characteristics of temperature in the complete multimodal bridge real-time data, including the daily maximum temperature, daily minimum temperature, daily average temperature and temperature change rate; the mean, variance, maximum and minimum values ​​of relative humidity, and calculate the correlation coefficient between humidity and temperature; the maximum value of wind speed, average wind speed, standard deviation of wind speed, main direction of wind direction and distribution frequency of wind direction, and obtain environmental data characteristic data; The extraction submodule is used to extract the mean, variance, maximum and minimum values ​​of the strain data in the complete multimodal bridge real-time data; extract the mean, variance, kurtosis and skewness of the acceleration in the time domain characteristics of the acceleration data; extract the peak value, peak-to-peak value and root mean square value of the time domain characteristics of the displacement data to obtain the structural response data characteristic data; The analysis submodule is used to calculate the daily flow, hourly flow and peak-hour flow of traffic, and analyze the periodic characteristics of traffic flow to obtain traffic load characteristic data; An identification submodule is used to obtain image data of key structural parts of the bridge in the complete multimodal bridge real-time data, identify the image data using the YOLOv5 image detection algorithm, obtain crack data and aging data of the key structural parts of the bridge, and obtain characteristic data of the key structural parts of the bridge; The normalization submodule is used to normalize the environmental data characteristic data, the structural response data characteristic data, the traffic load characteristic data and the bridge key structural component characteristic data to obtain the fused bridge real-time data, wherein the fused bridge real-time data at least includes the environmental data characteristic data. , structural response data characteristic data , traffic load characteristic data and characteristic data of key structural parts of bridges .

Images