A bridge deck icing early warning system and method based on multi-source dynamic data fusion

Abstract

This invention relates to the field of road monitoring, specifically to a bridge deck icing early warning system and method based on multi-source dynamic data fusion. The system includes: a road surface monitoring module, a meteorological monitoring module, an icing prediction module, a road surface load module, and a dispatch and early warning module. The road surface monitoring module is used to set up monitoring instruments and weather stations. The meteorological monitoring module is used to collect environmental information and test rainwater salinity. The icing prediction module is used to predict future meteorological data. The road surface load module is used to determine the degree of damage to the road surface caused by icing. The dispatch and early warning module issues risk level and time window warnings. This invention can quantify the conditions for icing, enabling early detection and handling of icing problems, achieving automatic monitoring and response to icing conditions, significantly improving the accuracy and real-time performance of icing time prediction, providing core technical support for intelligent traffic management, and helping to ensure traffic safety, extend road service life, and improve bridge deck traffic efficiency.

Description

Technical Field

[0001] This invention relates to the field of road monitoring, specifically to a bridge deck icing early warning system and method based on multi-source dynamic data fusion. Background Technology

[0002] Road icing refers to the phenomenon where, in cold weather, moisture or snow on the road surface freezes in the low temperature, forming a thin layer of ice or icicle, making the road surface slippery. Due to their rapid heat dissipation and lack of geothermal conduction, bridge decks have a higher risk of icing than ordinary road surfaces, causing vehicles to slip when driving on them in winter. Additionally, icing can seep into cracks in the bridge surface, damaging the road material and causing road surface cracking.

[0003] Icing warning is a crucial tool for monitoring road conditions. By predicting the risk of icing on bridge surfaces, it allows for proactive traffic control, road maintenance, and other protective measures to mitigate the risks associated with icing. Traditional icing warning systems rely primarily on single parameters such as temperature and humidity, and employ linear regression or static threshold judgment models. These methods generally suffer from low prediction accuracy and high latency, failing to allow sufficient time for emergency response and thus hindering the intelligentization of icing prevention and control.

[0004] Existing technologies predict icing conditions through data fusion analysis and the construction of high-precision prediction models. However, the traditional Sigmoid activation function suffers from the gradient vanishing problem, which reduces the accuracy of predictions. At the same time, it lacks quantitative evaluation standards for the maintenance status of road surfaces. Summary of the Invention

[0005] The purpose of this invention is to provide a bridge deck icing early warning system and method based on multi-source dynamic data fusion, so as to solve the problems mentioned in the background art.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: a bridge deck icing early warning system based on multi-source dynamic data fusion, comprising: a road surface monitoring module, a meteorological monitoring module, an icing prediction module, a road surface load module, and a scheduling early warning module;

[0007] The road surface monitoring module is used to install a remote sensing road surface condition monitor and a multi-functional weather station on the bridge surface. The road surface condition monitor is set at a position higher than the bridge surface and is equipped with a stereo camera and image recognition function, which can identify the appearance of cracks and water accumulation on the bridge surface. The multi-functional weather station is set on one side of the road and has weather forecasting function, precipitation sampling function and gateway communication function. The weather station data fusion terminal receives meteorological data sent by the meteorological department and merges it with local meteorological data.

[0008] The meteorological monitoring module is used to collect temperature, humidity, precipitation and wind speed information from the environment according to conventional meteorological monitoring methods, collect bridge surface precipitation using precipitation collection devices in the meteorological station, perform salinity tests on the collected rainwater, and obtain rainwater volume, density and salinity data to calculate the freezing point and freezing volume of precipitation.

[0009] The icing prediction module is used to store historical meteorological data in the database. For bridge surface meteorological data, the Leaky ReLU function based on historical meteorological changes is used as the activation function. An improved BP neural network is used to learn historical meteorological data, determine the weights of the activation function, construct hidden layer network nodes from the activation function, and output the prediction model after feedback testing. The prediction model is used to predict meteorological data in the next cycle to obtain predicted values ​​of temperature, precipitation and salinity.

[0010] The road load module is used to calculate the volume based on the width, depth, and horizontal curvature of the road cracks, calculate the water depth, freezing point, and ice volume of the bridge deck based on predicted meteorological data, and calculate the average load of the cracks when the temperature drops below the freezing point of the water, compare it with the maximum load of the road material, and determine the degree of damage to the road surface caused by ice by combining the indoor test results.

[0011] The dispatch and early warning module is used to transmit early warning data to the traffic management center by relying on 5G or edge computing gateways, issue risk level and time window warnings based on the prediction results, and link with the intelligent dispatch system, integrate with the traffic signal control system or vehicle guidance system to carry out traffic control in advance, and automatically call road maintenance facilities to realize automatic dispatch of snow and ice removal.

[0012] Furthermore, the road surface monitoring module includes: a road surface remote sensing unit and an image recognition unit;

[0013] The road surface remote sensing unit is used to capture the shape of bridge surface cracks and water accumulation using a stereo camera, and to remotely sense the road surface condition.

[0014] The image recognition unit is used to analyze the captured images, obtain the width, depth, and horizontal curvature of the road surface cracks, and determine the extent of water coverage on the road surface.

[0015] Furthermore, the meteorological monitoring module includes: a meteorological sampling unit and a sample testing unit;

[0016] The meteorological sampling unit is used to determine the temperature, humidity, wind force and direction, and precipitation data of the bridge deck using thermometers, hygrometers, wind vanes, and precipitation collectors.

[0017] The sample testing unit is used to test the collected precipitation, obtain the volume, density and salinity of the precipitation, and calculate the freezing characteristics of the precipitation.

[0018] Furthermore, the icing prediction module includes: a data storage unit, a neural network unit, and a model prediction unit;

[0019] The data storage unit is used to integrate cloud-based meteorological data with local meteorological data and to store the integrated meteorological data;

[0020] The neural network unit is used to train an improved BP neural network using a fused meteorological database to determine the weights of the LeakyReLU function;

[0021] The model prediction unit is used to construct a precipitation prediction model and determine the meteorological data of the bridge surface area in the next cycle.

[0022] Furthermore, the road load module includes: a water accumulation status unit, a load calculation unit, and a damage assessment unit;

[0023] The water accumulation status unit is used to calculate the water accumulation volume in each bridge deck crack based on the shape parameters of the crack;

[0024] The load calculation unit is used to calculate the average load of ice on the cracks based on the predicted salinity, temperature and volume of the accumulated water.

[0025] The damage assessment unit is used to conduct a four-level damage risk assessment of the road surface based on the indoor test results of the road surface materials.

[0026] Furthermore, the scheduling and early warning module includes: an edge gateway unit and a risk release unit;

[0027] The edge gateway unit is used to achieve low-latency data transmission using a gateway, and to synchronize local test data to the processing center;

[0028] The risk release unit is used to release corresponding risk reports based on the calculated water depth, icing depth, and road damage, and to call the management center for action.

[0029] A bridge deck icing early warning method based on multi-source dynamic data fusion includes the following steps:

[0030] Step S1. Install a remote sensing road surface condition monitoring instrument on the bridge surface, use a stereo camera to photograph the cracks on the bridge surface, obtain the shape and depth of the cracks, set up a multi-functional weather station on one side of the road, collect local weather data of the bridge surface, and integrate cloud weather data with local weather data to obtain fused weather data.

[0031] Step S2. Collect precipitation from the bridge surface using the precipitation collection device in the weather station, perform a salinity test on the collected precipitation, determine the volume, density and salinity of the precipitation, and calculate the freezing point and freezing volume of the precipitation based on the volume, density and salinity.

[0032] Step S3. Train an improved BP neural network using a fused meteorological database. Use the Leaky ReLU function based on historical meteorological changes as the activation function. After determining the weights of the activation function, construct hidden network nodes using the activation function to obtain a meteorological prediction model and predict the meteorological data of the bridge surface in the next cycle.

[0033] Step S4. Determine the road icing state based on the predicted precipitation and freezing point. When the road is in an icing state, calculate the volume of water in each bridge deck crack based on the shape parameters and depth of the cracks. Calculate the average load of the icing on the cracks based on the salinity, temperature and volume of the water. Compare the results with the indoor test results of the road surface materials to assess the bridge deck damage risk level.

[0034] Step S5. Relying on 5G or edge computing gateways, transmit the early warning data to the traffic control center. When the road is in an ice-free state, issue a risk level and time window warning to remind the traffic control center to carry out traffic control and road maintenance.

[0035] Step S6. Use the trained improved BP neural network to predict the freezing time, output the prediction result and accuracy of the freezing time, and send different levels of risk warnings according to the freezing time.

[0036] Furthermore, step S1 includes:

[0037] Step S11. Set the remote sensing road surface condition monitoring instrument at a position higher than the bridge surface. The remote sensing road surface condition monitoring instrument has a stereo camera and an image processing terminal. The viewing angle of the stereo camera covers the entire bridge surface to be inspected and acquires the shape and depth data of each crack on the bridge surface. The shape data includes the width and horizontal curvature of the road surface crack.

[0038] Step S12. Set up a multi-functional weather station on one side of the bridge deck road, and use the thermometer, hygrometer, wind vane and precipitation collector in the weather station to determine the temperature and humidity, wind force and direction and precipitation data of the bridge deck.

[0039] Step S13. Set up a data fusion terminal in the meteorological station to receive meteorological data sent by the meteorological department, fuse it with the collected local meteorological data, and obtain fused meteorological data after calibrating the local meteorological data.

[0040] Furthermore, step S2 includes:

[0041] Step S21. Place a precipitation collection device with a fixed opening area in the open at the weather station. Collect precipitation each time it occurs. Determine the amount of precipitation based on the opening area, the duration of precipitation, and the volume of collected precipitation. Feedback the precipitation data to the weather station.

[0042] Step S22. Weigh and measure the salinity of the rainwater collected by the precipitation collection device to determine the volume, density and salinity of the precipitation. The salinity measurement methods include: conductivity method, refraction method, density method and optical sensor method.

[0043] Step S23. Calculate the freezing point and freezing volume of precipitation using the formula for the volume change of brine freezing:

[0044]

[0045] Where Tf represents the freezing point of precipitation, Kf represents the freezing point depression constant of water, S represents salinity, Ms represents the molar mass of salt, V represents the volume of ice, mw represents the mass of water, and p represents the density of ice.

[0046] Furthermore, step S3 includes:

[0047] Step S31. Train an improved BP neural network using a fused meteorological database. Construct hidden layer nodes by using preceding meteorological data as the input layer and subsequent meteorological data as the output layer. Use the Leaky ReLU function as the activation function of the neural network. The Leaky ReLU function has the following expression:

[0048]

[0049] Where x represents the input and a represents the scaling constant, which is obtained from the feedback data during the training process;

[0050] Step S32. Construct hidden layer network nodes using activation functions to obtain the weather prediction model. The input layer has n neurons, where n represents the number of weather data points in the training database. The hidden layer has n1 neurons, and the output layer has m neurons. The output of each neuron in each layer is:

[0051]

[0052] Where xj and yk represent the outputs of the j-th hidden layer neuron and the k-th output layer neuron, respectively; wij represents the weights from the i-th input layer neuron to the j-th hidden layer; wjk represents the weights from the j-th hidden layer neuron to the k-th output layer neuron, respectively; xi represents the input sample of the i-th input layer neuron; and θj and θk represent the thresholds of the j-th hidden layer neuron and the k-th output layer neuron, respectively.

[0053] Step S33. Using the trained prediction model, predict the meteorological data of the bridge deck in the next cycle.

[0054] Furthermore, step S4 includes:

[0055] Step S41. Use the prediction model to obtain the temperature and precipitation freezing point for the next cycle. When the temperature is lower than the precipitation freezing point, issue an early warning of freezing status. Calculate the volume of the crack based on the shape parameters and depth of the crack. When the precipitation in the crack exceeds the volume of the crack, calculate the average load of freezing on the crack: W=(V·p·g+h·E·A) / L, where W represents the average load, g represents the gravitational acceleration, h represents the thermal expansion coefficient of ice, E represents the elastic modulus of ice, A is the cross-sectional area of ​​the crack, and L is the inner surface area of ​​the crack.

[0056] Step S42. Conduct indoor load tests on the road surface materials, compare the calculated average load with the test results, assess the damage to the road surface, and conduct a four-level damage risk assessment of the road surface.

[0057] Furthermore, step S5 includes:

[0058] Step S51. Relying on 5G or edge computing gateways, achieve low-latency data transmission, synchronize local test data to the traffic management center, and issue risk level and time window warnings based on the prediction results;

[0059] Step S52. Link the intelligent dispatch system, integrate with the traffic signal control system or vehicle guidance system, conduct traffic control in advance, and automatically call road maintenance facilities to realize automatic dispatch of snow and ice removal.

[0060] Furthermore, the process of the neural network predicting the freezing time in step S6 includes:

[0061] A dynamic weight fusion strategy is adopted to weight and fuse the prediction results of BP neural network and random forest. The model architecture is divided into three stages: parallel prediction, weight calculation and result fusion. In the parallel prediction stage, the two base models independently process the input features. In the weight calculation stage, the model weights are dynamically allocated based on the validation set error. In the result fusion stage, the final predicted value is weighted and output.

[0062] Example 1: The BP neural network has the following architecture: Input layer: 7 nodes, Hidden layer 1: 32 nodes, activation function is Leaky ReLU, Dropout layer: dropout rate 0.2, used to prevent overfitting, Hidden layer 2: 16 nodes, activation function is Leaky ReLU, Output layer: 1 node, used for linear activation;

[0063] In this embodiment, the following Leaky ReLU function is used:

[0064]

[0065] The mean squared error (MSE) was used as the loss function, and the training parameters were as follows: optimizer: Adam (learning rate = 0.001), batch size: 32, training epochs: 200, validation set ratio: 15%;

[0066] Multiple regression decision trees are constructed using Bootstrap sampling, and the final output is the mean of the predictions from all trees. The training parameters for the random forest are as follows: maximum tree depth: 8, minimum number of samples for node splitting: 5, feature selection strategy: d = 7, random seed: 42.

[0067] The final result is output using a dynamic weighted ensemble algorithm.

[0068] Compared with the prior art, the beneficial effects achieved by the present invention are:

[0069] 1. This invention, by setting up a remote-sensing road surface monitoring instrument and a weather station on the bridge deck, acquires road surface crack data from the stereo camera of the road surface monitoring instrument and meteorological data from the weather station, calculates the current water depth on the road surface, detects the salt concentration of the water, calculates the freezing point and the volume of ice, and can quantify the conditions for ice formation, so as to detect and deal with ice formation problems at an early stage, and realize automatic monitoring and response to ice formation conditions.

[0070] 2. This invention utilizes an improved BP neural network to construct a prediction model to predict future meteorological data. When the water depth exceeds the maximum value of road surface cracks and the temperature reaches freezing point, an icing warning is issued. This breaks through the limitations of traditional methods that rely on single meteorological data. The warning system can issue risk level and time window warnings based on the predicted probability and time of icing, significantly improving the accuracy and real-time performance of icing time prediction, and providing core technical support for intelligent traffic management.

[0071] 3. This invention determines the degree of damage to the road surface caused by ice based on road surface crack data and meteorological data, and automatically schedules snow and ice removal to reduce the incidence of traffic accidents in snowy weather, reduce resource waste caused by manual inspection and overuse of de-icing agents, and at the same time support seamless integration with traffic signal control and vehicle guidance systems, which helps to ensure traffic safety, extend the service life of roads and improve the traffic efficiency of bridges. Attached Figure Description

[0072] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0073] Figure 1 This is a schematic diagram of the structure of a bridge deck icing early warning system based on multi-source dynamic data fusion according to the present invention;

[0074] Figure 2This is a schematic diagram illustrating the steps of a bridge deck icing early warning method based on multi-source dynamic data fusion according to the present invention. Detailed Implementation

[0075] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0076] Please see Figure 1 The present invention provides a technical solution: a bridge deck icing early warning system based on multi-source dynamic data fusion, comprising: a road surface monitoring module, a meteorological monitoring module, an icing prediction module, a road surface load module, and a scheduling early warning module;

[0077] The road surface monitoring module is used to install a remote sensing road surface condition monitor and a multi-functional weather station on the bridge surface. The road surface condition monitor is set at a position higher than the bridge surface and is equipped with a stereo camera and image recognition function, which can identify the appearance of cracks and water accumulation on the bridge surface. The multi-functional weather station is set on one side of the road and has weather forecasting function, precipitation sampling function and gateway communication function. The weather station data fusion terminal receives meteorological data sent by the meteorological department and merges it with local meteorological data.

[0078] The road surface monitoring module includes: a road surface remote sensing unit and an image recognition unit;

[0079] The road surface remote sensing unit is used to capture the shape of bridge surface cracks and water accumulation using a stereo camera, and to remotely sense the road surface condition.

[0080] The image recognition unit is used to analyze the captured images, obtain the width, depth, and horizontal curvature of the road surface cracks, and determine the extent of water coverage on the road surface.

[0081] The meteorological monitoring module is used to collect temperature, humidity, precipitation and wind speed information from the environment according to conventional meteorological monitoring methods, collect bridge surface precipitation using precipitation collection devices in the meteorological station, perform salinity tests on the collected rainwater, and obtain rainwater volume, density and salinity data to calculate the freezing point and freezing volume of precipitation.

[0082] The meteorological monitoring module includes: a meteorological sampling unit and a sample testing unit;

[0083] The meteorological sampling unit is used to determine the temperature, humidity, wind force and direction, and precipitation data of the bridge deck using thermometers, hygrometers, wind vanes, and precipitation collectors.

[0084] The sample testing unit is used to test the collected precipitation, obtain the volume, density and salinity of the precipitation, and calculate the freezing characteristics of the precipitation.

[0085] The icing prediction module is used to store historical meteorological data in the database. For bridge surface meteorological data, the Leaky ReLU function based on historical meteorological changes is used as the activation function. An improved BP neural network is used to learn historical meteorological data, determine the weights of the activation function, construct hidden layer network nodes from the activation function, and output the prediction model after feedback testing. The prediction model is used to predict meteorological data in the next cycle to obtain predicted values ​​of temperature, precipitation and salinity.

[0086] The freezing prediction module includes: a data storage unit, a neural network unit, and a model prediction unit;

[0087] The data storage unit is used to integrate cloud-based meteorological data with local meteorological data and to store the integrated meteorological data;

[0088] The neural network unit is used to train an improved BP neural network using a fused meteorological database to determine the weights of the LeakyReLU function;

[0089] The model prediction unit is used to construct a precipitation prediction model and determine the meteorological data of the bridge surface area in the next cycle.

[0090] The road load module is used to calculate the volume based on the width, depth, and horizontal curvature of the road cracks, calculate the water depth, freezing point, and ice volume of the bridge deck based on predicted meteorological data, and calculate the average load of the cracks when the temperature drops below the freezing point of the water, compare it with the maximum load of the road material, and determine the degree of damage to the road surface caused by ice by combining the indoor test results.

[0091] The road load module includes: a water accumulation status unit, a load calculation unit, and a damage assessment unit;

[0092] The water accumulation status unit is used to calculate the water accumulation volume in each bridge deck crack based on the shape parameters of the crack;

[0093] The load calculation unit is used to calculate the average load of ice on the cracks based on the predicted salinity, temperature and volume of the accumulated water.

[0094] The damage assessment unit is used to conduct a four-level damage risk assessment of the road surface based on the indoor test results of the road surface materials.

[0095] The dispatch and early warning module is used to transmit early warning data to the traffic management center by relying on 5G or edge computing gateways, issue risk level and time window warnings based on the prediction results, and link with the intelligent dispatch system, integrate with the traffic signal control system or vehicle guidance system to carry out traffic control in advance, and automatically call road maintenance facilities to realize automatic dispatch of snow and ice removal.

[0096] The scheduling and early warning module includes: an edge gateway unit and a risk release unit;

[0097] The edge gateway unit is used to achieve low-latency data transmission using a gateway, and to synchronize local test data to the processing center;

[0098] The risk release unit is used to release corresponding risk reports based on the calculated water depth, icing depth, and road damage, and to call the management center for action.

[0099] like Figure 2 As shown, a bridge deck icing early warning method based on multi-source dynamic data fusion includes the following steps:

[0100] Step S1. Install a remote sensing road surface condition monitoring instrument on the bridge surface, use a stereo camera to photograph the cracks on the bridge surface, obtain the shape and depth of the cracks, set up a multi-functional weather station on one side of the road, collect local weather data of the bridge surface, and integrate cloud weather data with local weather data to obtain fused weather data.

[0101] Step S1 includes:

[0102] Step S11. Set the remote sensing road surface condition monitoring instrument at a position higher than the bridge surface. The remote sensing road surface condition monitoring instrument has a stereo camera and an image processing terminal. The viewing angle of the stereo camera covers the entire bridge surface to be inspected and acquires the shape and depth data of each crack on the bridge surface. The shape data includes the width and horizontal curvature of the road surface crack.

[0103] Step S12. Set up a multi-functional weather station on one side of the bridge deck road, and use the thermometer, hygrometer, wind vane and precipitation collector in the weather station to determine the temperature and humidity, wind force and direction and precipitation data of the bridge deck.

[0104] Step S13. Set up a data fusion terminal in the meteorological station to receive meteorological data sent by the meteorological department, fuse it with the collected local meteorological data, and obtain fused meteorological data after calibrating the local meteorological data.

[0105] Step S2. Collect precipitation from the bridge surface using the precipitation collection device in the weather station, perform a salinity test on the collected precipitation, determine the volume, density and salinity of the precipitation, and calculate the freezing point and freezing volume of the precipitation based on the volume, density and salinity.

[0106] Step S2 includes:

[0107] Step S21. Place a precipitation collection device with a fixed opening area in the open at the weather station. Collect precipitation each time it occurs. Determine the amount of precipitation based on the opening area, the duration of precipitation, and the volume of collected precipitation. Feedback the precipitation data to the weather station.

[0108] Step S22. Weigh and measure the salinity of the rainwater collected by the precipitation collection device to determine the volume, density and salinity of the precipitation. The salinity measurement methods include: conductivity method, refraction method, density method and optical sensor method.

[0109] Step S23. Calculate the freezing point and freezing volume of precipitation using the formula for the volume change of brine freezing:

[0110]

[0111] Where Tf represents the freezing point of precipitation, Kf represents the freezing point depression constant of water, S represents salinity, Ms represents the molar mass of salt, V represents the volume of ice, mw represents the mass of water, and p represents the density of ice.

[0112] Step S3. Train an improved BP neural network using a fused meteorological database. Use the Leaky ReLU function based on historical meteorological changes as the activation function. After determining the weights of the activation function, construct hidden network nodes using the activation function to obtain a meteorological prediction model and predict the meteorological data of the bridge surface in the next cycle.

[0113] Step S3 includes:

[0114] Step S31. Train an improved BP neural network using a fused meteorological database. Construct hidden layer nodes by using preceding meteorological data as the input layer and subsequent meteorological data as the output layer. Use the Leaky ReLU function as the activation function of the neural network. The Leaky ReLU function has the following expression:

[0115]

[0116] Where x represents the input and a represents the scaling constant, which is obtained from the feedback data during the training process;

[0117] Step S32. Construct hidden layer network nodes using activation functions to obtain the weather prediction model. The input layer has n neurons, where n represents the number of weather data points in the training database. The hidden layer has n1 neurons, and the output layer has m neurons. The output of each neuron in each layer is:

[0118]

[0119] Where xj and yk represent the outputs of the j-th hidden layer neuron and the k-th output layer neuron, respectively; wij represents the weights from the i-th input layer neuron to the j-th hidden layer; wjk represents the weights from the j-th hidden layer neuron to the k-th output layer neuron, respectively; xi represents the input sample of the i-th input layer neuron; and θj and θk represent the thresholds of the j-th hidden layer neuron and the k-th output layer neuron, respectively.

[0120] Step S33. Using the trained prediction model, predict the meteorological data of the bridge deck in the next cycle.

[0121] Step S4. Determine the road icing state based on the predicted precipitation and freezing point. When the road is in an icing state, calculate the volume of water in each bridge deck crack based on the shape parameters and depth of the cracks. Calculate the average load of the icing on the cracks based on the salinity, temperature and volume of the water. Compare the results with the indoor test results of the road surface materials to assess the bridge deck damage risk level.

[0122] Step S4 includes:

[0123] Step S41. Use the prediction model to obtain the temperature and precipitation freezing point for the next cycle. When the temperature is lower than the precipitation freezing point, issue an early warning of freezing status. Calculate the volume of the crack based on the shape parameters and depth of the crack. When the precipitation in the crack exceeds the volume of the crack, calculate the average load of freezing on the crack: W=(V·p·g+h·E·A) / L, where W represents the average load, g represents the gravitational acceleration, h represents the thermal expansion coefficient of ice, E represents the elastic modulus of ice, A is the cross-sectional area of ​​the crack, and L is the inner surface area of ​​the crack.

[0124] Step S42. Conduct indoor load tests on the road surface materials, compare the calculated average load with the test results, assess the damage to the road surface, and conduct a four-level damage risk assessment of the road surface.

[0125] Step S5. Relying on 5G or edge computing gateways, transmit the early warning data to the traffic control center. When the road is in an icy state, issue a risk level and time window warning to remind the traffic control center to carry out traffic control and road maintenance.

[0126] Step S5 includes:

[0127] Step S51. Relying on 5G or edge computing gateways, achieve low-latency data transmission, synchronize local test data to the traffic management center, and issue risk level and time window warnings based on the prediction results;

[0128] Step S52. Link the intelligent dispatch system, integrate with the traffic signal control system or vehicle guidance system, conduct traffic control in advance, and automatically call road maintenance facilities to realize automatic dispatch of snow and ice removal.

[0129] Example 2: The prediction model indicates that the future precipitation on the bridge deck will be 10m. 3 / min, the salinity ratio of precipitation is 0.1, the molar mass of salt is 58 g / mol, and the density of ice is 900 kg / m³. 3 If the freezing point depression constant Kf = 1.86℃·kg / mol, then the freezing point is -3°C. If the predicted temperature for the next cycle is -5°C, then the road is determined to be in an icy state, and the data is synchronized to the traffic control center.

[0130] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0131] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A bridge deck icing early warning method based on multi-source dynamic data fusion, characterized in that, The method includes the following steps: Step S1. Install a remote sensing road surface condition monitoring instrument on the bridge surface, use a stereo camera to photograph the cracks on the bridge surface, obtain the shape and depth of the cracks, set up a multi-functional weather station on one side of the road, collect local weather data of the bridge surface, and integrate cloud weather data with local weather data to obtain fused weather data. Step S2. Collect precipitation from the bridge surface using the precipitation collection device in the weather station, perform a salinity test on the collected precipitation, determine the volume, density and salinity of the precipitation, and calculate the freezing point and freezing volume of the precipitation based on the volume, density and salinity. Step S3. Train an improved BP neural network using a fused meteorological database. Use the LeakyReLU function based on historical meteorological changes as the activation function. After determining the weights of the activation function, construct hidden network nodes using the activation function to obtain a meteorological prediction model and predict the meteorological data of the bridge surface in the next cycle. Step S4. Determine the road icing state based on the predicted precipitation and freezing point. When the road is in an icing state, calculate the volume of water in each bridge deck crack based on the shape parameters and depth of the cracks. Calculate the average load of the icing on the cracks based on the salinity, temperature and volume of the water. Compare the results with the indoor test results of the road surface materials to assess the bridge deck damage risk level. Step S5. Relying on 5G or edge computing gateways, transmit the early warning data to the traffic control center. When the road is in an icy state, issue a risk level and time window warning to remind the traffic control center to carry out traffic control and road maintenance.

2. The bridge deck icing early warning method based on multi-source dynamic data fusion according to claim 1, characterized in that: Step S1 includes: Step S11. Set the remote sensing road surface condition monitoring instrument at a position higher than the bridge surface. The remote sensing road surface condition monitoring instrument has a stereo camera and an image processing terminal. The viewing angle of the stereo camera covers the entire bridge surface to be inspected and acquires the shape and depth data of each crack on the bridge surface. The shape data includes the width and horizontal curvature of the road surface crack. Step S12. Set up a multi-functional weather station on one side of the bridge deck road, and use the thermometer, hygrometer, wind vane and precipitation collector in the weather station to determine the temperature and humidity, wind force and direction and precipitation data of the bridge deck. Step S13. Set up a data fusion terminal in the meteorological station to receive meteorological data sent by the meteorological department, fuse it with the collected local meteorological data, and obtain fused meteorological data after calibrating the local meteorological data.

3. The bridge deck icing early warning method based on multi-source dynamic data fusion according to claim 2, characterized in that: Step S2 includes: Step S21. Place a precipitation collection device with a fixed opening area in the open at the weather station. Collect precipitation each time it occurs. Determine the amount of precipitation based on the opening area, the duration of precipitation, and the volume of collected precipitation. Feedback the precipitation data to the weather station. Step S22. Weigh and measure the salinity of the rainwater collected by the precipitation collection device to determine the volume, density and salinity of the precipitation. The salinity measurement methods include: conductivity method, refraction method, density method and optical sensor method. Step S23. Calculate the freezing point and freezing volume of precipitation using the formula for the volume change of brine freezing: Where Tf represents the freezing point of precipitation, Kf represents the freezing point depression constant of water, S represents salinity, Ms represents the molar mass of salt, V represents the volume of ice, mw represents the mass of water, and p represents the density of ice.

4. The bridge deck icing early warning method based on multi-source dynamic data fusion according to claim 3, characterized in that: Step S3 includes: Step S31. Train an improved BP neural network using a fused meteorological database. Construct hidden layer nodes by using preceding meteorological data as the input layer and subsequent meteorological data as the output layer. Use the Leaky ReLU function as the activation function of the neural network. The Leaky ReLU function has the following expression: Where x represents the input and a represents the scaling constant, which is obtained from the feedback data during the training process; Step S32. Construct hidden layer network nodes using activation functions to obtain the weather prediction model. The input layer has n neurons, where n represents the number of weather data points in the training database. The hidden layer has n1 neurons, and the output layer has m neurons. The output of each neuron in each layer is: Where xj and yk represent the outputs of the j-th hidden layer neuron and the k-th output layer neuron, respectively; wij represents the weights from the i-th input layer neuron to the j-th hidden layer; wjk represents the weights from the j-th hidden layer neuron to the k-th output layer neuron, respectively; xi represents the input sample of the i-th input layer neuron; and θj and θk represent the thresholds of the j-th hidden layer neuron and the k-th output layer neuron, respectively. Step S33. Using the trained prediction model, predict the meteorological data of the bridge deck in the next cycle.

5. The bridge deck icing early warning method based on multi-source dynamic data fusion according to claim 4, characterized in that: Step S4 includes: Step S41. Use the prediction model to obtain the temperature and precipitation freezing point for the next cycle. When the temperature is lower than the precipitation freezing point, issue an early warning of freezing status. Calculate the volume of the crack based on the shape parameters and depth of the crack. When the precipitation in the crack exceeds the volume of the crack, calculate the average load of freezing on the crack: W=(V·p·g+h·E·A) / L, where W represents the average load, g represents the gravitational acceleration, h represents the thermal expansion coefficient of ice, E represents the elastic modulus of ice, A is the cross-sectional area of ​​the crack, and L is the inner surface area of ​​the crack. Step S42. Conduct indoor load tests on the road surface materials, compare the calculated average load with the test results, assess the damage to the road surface, and conduct a four-level damage risk assessment of the road surface. Step S5 includes: Step S51. Relying on 5G or edge computing gateways, achieve low-latency data transmission, synchronize local test data to the traffic management center, and issue risk level and time window warnings based on the prediction results; Step S52. Link the intelligent dispatch system, integrate with the traffic signal control system or vehicle guidance system, conduct traffic control in advance, and automatically call road maintenance facilities to realize automatic dispatch of snow and ice removal.

6. A bridge deck icing early warning system based on multi-source dynamic data fusion, characterized in that, The system includes the following modules: road surface monitoring module, meteorological monitoring module, icing prediction module, road surface load module, and dispatch and early warning module; The road surface monitoring module is used to install a remote sensing road surface condition monitor and a multi-functional weather station on the bridge surface. The road surface condition monitor is set at a position higher than the bridge surface and is equipped with a stereo camera and image recognition function, which can identify the appearance of cracks and water accumulation on the bridge surface. The multi-functional weather station is set on one side of the road and has weather forecasting function, precipitation sampling function and gateway communication function. The weather station data fusion terminal receives meteorological data sent by the meteorological department and merges it with local meteorological data. The meteorological monitoring module is used to collect temperature, humidity, precipitation and wind speed information from the environment according to conventional meteorological monitoring methods, collect bridge surface precipitation using precipitation collection devices in the meteorological station, perform salinity tests on the collected rainwater, and obtain rainwater volume, density and salinity data to calculate the freezing point and freezing volume of precipitation. The icing prediction module is used to store historical meteorological data in the database. For bridge surface meteorological data, the Leaky ReLU function based on historical meteorological changes is used as the activation function. An improved BP neural network is used to learn historical meteorological data, determine the weights of the activation function, construct hidden layer network nodes from the activation function, and output the prediction model after feedback testing. The prediction model is used to predict meteorological data in the next cycle to obtain predicted values ​​of temperature, precipitation and salinity. The road load module is used to calculate the volume based on the width, depth, and horizontal curvature of the road cracks, calculate the water depth, freezing point, and ice volume of the bridge deck based on predicted meteorological data, and calculate the average load of the cracks when the temperature drops below the freezing point of the water, compare it with the maximum load of the road material, and determine the degree of damage to the road surface caused by ice by combining the indoor test results. The dispatch and early warning module is used to transmit early warning data to the traffic management center by relying on 5G or edge computing gateways, issue risk level and time window warnings based on the prediction results, and link with the intelligent dispatch system, integrate with the traffic signal control system or vehicle guidance system to carry out traffic control in advance, and automatically call road maintenance facilities to realize automatic dispatch of snow and ice removal.

7. A bridge deck icing early warning system based on multi-source dynamic data fusion according to claim 6, characterized in that: The road surface monitoring module includes: a road surface remote sensing unit and an image recognition unit; The road surface remote sensing unit is used to capture the shape of bridge surface cracks and water accumulation using a stereo camera, and to remotely sense the road surface condition. The image recognition unit is used to analyze the captured images, obtain the width, depth, and horizontal curvature of the road surface cracks, and determine the coverage of the road surface by the accumulated water. The meteorological monitoring module includes: a meteorological sampling unit and a sample testing unit; The meteorological sampling unit is used to determine the temperature, humidity, wind force and direction, and precipitation data of the bridge deck using thermometers, hygrometers, wind vanes, and precipitation collectors. The sample testing unit is used to test the collected precipitation, obtain the volume, density and salinity of the precipitation, and calculate the freezing characteristics of the precipitation.

8. A bridge deck icing early warning system based on multi-source dynamic data fusion according to claim 7, characterized in that: The freezing prediction module includes: a data storage unit, a neural network unit, and a model prediction unit; The data storage unit is used to integrate cloud-based meteorological data with local meteorological data and to store the integrated meteorological data; The neural network unit is used to train an improved BP neural network using a fused meteorological database to determine the weights of the Leaky ReLU function; The model prediction unit is used to construct a precipitation prediction model and determine the meteorological data of the bridge surface area in the next cycle.

9. A bridge deck icing early warning system based on multi-source dynamic data fusion according to claim 8, characterized in that: The road load module includes: a water accumulation status unit, a load calculation unit, and a damage assessment unit; The water accumulation status unit is used to calculate the water accumulation volume in each bridge deck crack based on the shape parameters of the crack; The load calculation unit is used to calculate the average load of ice on the cracks based on the predicted salinity, temperature and volume of the accumulated water. The damage assessment unit is used to conduct a four-level damage risk assessment of the road surface based on the indoor test results of the road surface materials.

10. A bridge deck icing early warning system based on multi-source dynamic data fusion according to claim 9, characterized in that: The scheduling and early warning module includes: an edge gateway unit and a risk release unit; The edge gateway unit is used to achieve low-latency data transmission using a gateway, and to synchronize local test data to the processing center; The risk release unit is used to release corresponding risk reports based on the calculated water depth, icing depth, and road damage, and to call the management center for action.

Images