A method and system for optical time-domain road fusion perception monitoring

Abstract

The present invention relates to the field of road monitoring technology, and in particular to an optical time-domain road fusion perception monitoring method and system, the method comprising the following steps: collecting temperature data, vibration data and strain data of a target road section based on a sensing optical fiber; dividing the road surface grid and determining the repair area of ​​each grid node, then determining the node strain according to the strain data, and then establishing a road data set in combination with the grid node position, temperature data and vibration data to construct a vibration prediction model; performing simulated repairs on each repair area respectively, and using a vibration prediction model to obtain vibration prediction data, and then using a load prediction model to calculate the road load index after repairing each repair area, so as to formulate a road repair plan. The present invention integrates three types of data, temperature, vibration and strain, to monitor the road, and can provide a road repair plan in a timely manner, thereby improving the real-time and accuracy of road monitoring and helping to reduce safety hazards in road operation.

Description

Technical Field

[0001] The present invention relates to the technical field of road monitoring, and in particular to an optical time-domain road fusion perception monitoring method and system. Background Art

[0002] With the continuous increase in urban traffic volume and the increasing vehicle load, the problem of road damage has become increasingly prominent, and higher requirements have been placed on road monitoring and repair. Traditional road monitoring methods mainly rely on manual inspections or simple sensor monitoring. These methods have problems such as limited monitoring range, low monitoring accuracy, and inability to monitor in real time. They are difficult to meet the real-time, accurate, and comprehensive requirements of modern urban traffic for road monitoring.

[0003] In recent years, with the continuous development of fiber optic sensing technology, optical time domain reflectometry has received extensive attention in the field of road monitoring due to its distributed measurement, long-distance monitoring, and high sensitivity. However, how to apply optical time domain reflectometry to road monitoring and combine it with the actual use of the road to achieve accurate perception and prediction of road conditions is an urgent problem to be solved in the current road monitoring field.

[0004] In addition, the formulation of road repair plans also faces many challenges. Traditional road repair plans are often based on empirical judgment or simple data statistics, lacking scientificity and accuracy. At the same time, road repair usually consumes a lot of manpower, material resources and financial resources. How to formulate the best repair plan with limited resources to maximize the service life of the road and improve the road's traffic capacity is an important topic in the current road repair field. Summary of the invention

[0005] In view of the defects in the prior art, the present invention provides an optical time-domain road fusion perception monitoring method and system.

[0006] In order to achieve the above-mentioned purpose, in the first aspect, the present invention provides an optical time-domain road fusion perception monitoring method, the method comprising the following steps: laying a sensing optical fiber on the target road section, and collecting the temperature data, strain data and vibration data of the target road section in real time based on the sensing optical fiber; dividing the road surface grid and determining the repair area of ​​each grid node, and then determining the node temperature, node strain and road vibration data according to the temperature data, the strain number and the vibration data; establishing a road data set according to the node temperature, grid node position, the node strain and the road vibration data, and then constructing a vibration prediction model; respectively simulating the repair of the repair area corresponding to each grid node to update the strain data, and using the vibration prediction model to obtain the vibration prediction data; according to the vibration prediction data, using the load prediction model to calculate the road load index after repairing each repair area, and then formulating a road repair plan. The present invention integrates the three data of temperature, vibration and strain to monitor the road, and can give a road repair plan in time, which improves the real-time and accuracy of road monitoring and helps to reduce the safety hazards of road operation.

[0007] Optionally, laying a sensing optical fiber on the target road section and collecting temperature data, strain data and vibration data of the target road section in real time based on the sensing optical fiber comprises the following steps:

[0008] Laying the sensing optical fiber on the target road section;

[0009] Sending an optical pulse signal to the sensing optical fiber, and receiving a Brillouin scattered light signal reflected by the sensing optical fiber to obtain a Brillouin gain spectrum;

[0010] According to the Brillouin gain spectrum, a Brillouin frequency shift extraction network is used to extract the Brillouin frequency shift, thereby obtaining the temperature data, the vibration data and the strain data.

[0011] Furthermore, by laying the sensing optical fiber on the target road section to obtain the Brillouin gain spectrum, and then by extracting the Brillouin frequency shift to obtain temperature data, vibration data and strain data, real-time and accurate monitoring of road temperature, vibration and strain can be achieved, providing a reliable data basis for subsequent data analysis and model building.

[0012] Optionally, the Brillouin frequency shift extraction network includes a first convolution unit, a second convolution unit and a fully connected unit, and the convolution kernel of the first convolution unit and the convolution kernel of the second convolution unit have different sizes.

[0013] Optionally, the Brillouin frequency shift extraction network performs the following steps when running:

[0014] Extracting a first feature of the Brillouin gain spectrum using a first convolution unit, and extracting a second feature of the Brillouin gain spectrum using the second convolution unit;

[0015] Matching the first feature and the second feature, and then performing nonlinear transformation on the first feature and the second feature through a convolution layer activation function to obtain feature fusion data;

[0016] The feature fusion data is input into the fully connected unit, and the Brillouin frequency shift is output.

[0017] Furthermore, the first convolution unit and the second convolution unit are used to obtain information features under different receptive fields, and feature fusion is performed to extract more accurate information features, and then the Brillouin frequency shift is accurately extracted through the fully connected unit, providing a reliable data foundation for subsequent data analysis and model construction.

[0018] Optionally, dividing the road surface grid and determining the repair area of ​​each grid node, and then determining the node temperature, node strain and road vibration data according to the temperature data, the strain number and the vibration data comprises the following steps:

[0019] Using a plurality of horizontal lines and vertical lines to divide the target road section into a plurality of square grids of the same size to form the road surface grid, wherein the intersection of the horizontal lines and the vertical lines is the grid node;

[0020] Divide the corresponding repair area with the grid node as the center, the repair area has the same shape and size as the road surface grid, and the midpoint of each side in the repair area is located on the grid line;

[0021] Determine the position of the strain point corresponding to each of the strain data at the same time on the road surface grid;

[0022] If the grid node has an adjacent strain point, the average of the temperature data of the strain points adjacent to the grid node is used as the node temperature of the grid node;

[0023] If the grid node does not have the adjacent strain point, the node temperature of the grid node is the ambient temperature;

[0024] If the strain point coincides with a certain grid node, the strain data of the strain point is used as the node strain of the grid node;

[0025] If the strain point is located on a grid line between a pair of adjacent grid nodes, the strain data of the strain point is distributed to the pair of adjacent grid nodes using a strain distribution model;

[0026] If the strain point is located in a certain road surface grid, the strain data of the strain point is distributed to four grid nodes adjacent to the strain point using the strain distribution model;

[0027] If there is no strain point around the grid node, the node strain of the grid node is zero;

[0028] The average value of all the vibration data at the same time is calculated as the road vibration data.

[0029] Furthermore, the present invention divides the road surface into grids and determines the node temperature, node strain and road vibration data according to the temperature data, vibration data and strain data to improve the availability of strain data, temperature data and vibration data, thereby facilitating subsequent model training and data prediction.

[0030] Optionally, the strain allocation model satisfies the following relationship:

[0031]

[0032] in, is the strain value assigned from strain point i to mesh node j, is the strain data of strain point i, is the distance between strain point i and grid node j, is the number of mesh nodes to which strain values ​​need to be assigned from strain point i, , m is the total number of the grid nodes.

[0033] Furthermore, the strain distribution model is simple to calculate, which is beneficial to improving the efficiency of road monitoring and can reduce the quality requirements of road monitoring on hardware equipment.

[0034] Optionally, establishing a road data set according to the node temperature, the grid node position, the node strain and the road vibration data, and then constructing a vibration prediction model comprises the following steps:

[0035] establishing the road data set according to the node temperature, the grid node position, the node strain and the road vibration data;

[0036] An initial vibration prediction model is constructed using a long short-term memory recurrent neural network, and the road data set is used to complete training and verification of the initial vibration prediction model to obtain the vibration prediction model.

[0037] Furthermore, the long short-term memory recurrent neural network can capture the long-term dependencies in time series data and is suitable for the prediction of road vibration data. This method can improve the accuracy and reliability of vibration prediction and provide a more scientific basis for the formulation of subsequent road repair plans.

[0038] Optionally, the step of calculating the road bearing index after repairing each of the repair areas using a bearing prediction model based on the vibration prediction data, and then formulating a road repair plan comprises the following steps:

[0039] Calculating the road bearing index after repairing each of the repair areas using the bearing prediction model according to the vibration prediction data;

[0040] Arrange the road load indexes in descending order, and then determine the repair order of each repair area to obtain the road repair plan;

[0041] Complete the repair of the target road section according to the road repair plan.

[0042] Furthermore, using the load prediction model to calculate the road load index based on the vibration prediction data will help to formulate the optimal repair plan under limited resources, thereby maximizing the service life of the road and improving the road's traffic capacity, thereby reducing safety hazards in road operation.

[0043] Optionally, the load prediction model satisfies the following relationship:

[0044]

[0045] Wherein, P is the road load index, is the equivalent elastic modulus of the road design base top, is the thickness of the road subbase layer, is the fundamental frequency, Extracted from the vibration prediction data.

[0046] Furthermore, the load-bearing prediction model can calculate the road load-bearing index based on parameters such as the equivalent rebound modulus of the road design base, the thickness of the road subbase layer and the fundamental frequency. The calculation is simple and has high reliability, which helps to quickly and accurately evaluate the traffic capacity of roads after repairs in different repair areas, and provide a reliable technical basis for formulating road repair plans.

[0047] In a second aspect, the present invention provides an optical time-domain road fusion perception monitoring system, the optical time-domain road fusion perception monitoring system comprising: a data acquisition device, a data output device, a processor and a storage, the storage comprising a computer-readable storage medium, the computer-readable storage medium storing a computer program, the computer program comprising program instructions, the program instructions when executed by the processor enable the processor to implement an optical time-domain road fusion perception monitoring method provided by the present invention. The system can stably run the method, thereby improving the detection efficiency of the road and improving the practicality of the method.

[0048] Beneficial effects:

[0049] First, the method obtains the Brillouin gain spectrum by laying the sensing optical fiber on the target road section, and constructs a Brillouin frequency shift extraction network to extract the Brillouin frequency shift to obtain temperature data, vibration data and strain data, thereby realizing real-time and accurate monitoring of road temperature, vibration and strain. Secondly, the method divides the road surface into grids and determines the node temperature, node strain and road vibration data according to the temperature data, vibration data and strain data to improve the availability of strain data, temperature data and vibration data, and then constructs a vibration prediction model based on a long short-term memory recurrent neural network. Finally, the method predicts the vibration prediction data after the simulated repair of the repair area according to the vibration prediction model, and then calculates the corresponding road bearing index based on the bearing prediction model, which is used to formulate a road repair plan, which is conducive to maximizing the service life of the road and improving the traffic capacity of the road under limited resources, and reducing the safety hazards of road operation. In addition, the system can run the method stably, thereby improving the detection efficiency of the road and improving the practicality of the method. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0051] Figure 1 A schematic diagram of a process of an optical time-domain road fusion perception monitoring method according to an embodiment of the present invention;

[0052] Figure 2 A schematic diagram of a repair area according to an embodiment of the present invention;

[0053] Figure 3 A schematic diagram of the framework of an optical time-domain road fusion perception monitoring system according to an embodiment of the present invention. DETAILED DESCRIPTION

[0054] The specific embodiments of the present invention will be described in detail below. It should be noted that the embodiments described herein are only for illustration and are not intended to limit the present invention. In the following description, a large number of specific details are set forth in order to provide a thorough understanding of the present invention. However, it is obvious to those of ordinary skill in the art that these specific details do not need to be adopted to implement the present invention. In other examples, in order to avoid confusing the present invention, known circuits, software or methods are not specifically described.

[0055] Throughout the specification, references to "one embodiment," "an embodiment," "an example," or "an example" mean that a particular feature, structure, or characteristic described in conjunction with the embodiment or example is included in at least one embodiment of the present invention. Therefore, the phrases "in one embodiment," "in an embodiment," "an example," or "an example" appearing in various places throughout the specification do not necessarily all refer to the same embodiment or example. In addition, particular features, structures, or characteristics may be combined in one or more embodiments or examples in any suitable combination and / or subcombination. In addition, it should be understood by those of ordinary skill in the art that the figures provided herein are for illustrative purposes and that the figures are not necessarily drawn to scale.

[0056] It should be noted in advance that, in an optional embodiment, except for independent explanations, the same symbols or letters appearing in all formulas have the same meanings and values.

[0057] In an alternative embodiment, see Figure 1 The present invention provides an optical time-domain road fusion perception monitoring method, the method comprising the following steps:

[0058] S1. Laying a sensing optical fiber on a target road section, and collecting temperature data, strain data and vibration data of the target road section in real time based on the sensing optical fiber.

[0059] Wherein, step S1 specifically includes the following steps:

[0060] S11, laying the sensing optical fiber on the target road section.

[0061] Specifically, in this embodiment, the laying route and layout of the sensor optical fiber are planned according to the length, width, traffic flow and monitoring requirements of the target road section to ensure that the optical fiber can fully cover the area that needs to be monitored while avoiding interference with traffic.

[0062] Furthermore, the sensor optical fiber is laid on the target road section according to the planned route and layout, and then the laid sensor optical fiber is preliminarily tested to ensure that the optical fiber can transmit the optical pulse signal normally and can receive the reflected Brillouin scattered light signal.

[0063] S12, sending an optical pulse signal to the sensing optical fiber, and receiving a Brillouin scattered light signal reflected by the sensing optical fiber to obtain a Brillouin gain spectrum.

[0064] Specifically, in this embodiment, a laser is used to send a series of short pulse laser signals to the sensing optical fiber to generate a Brillouin scattered light signal, and a photodetector is used to receive the Brillouin scattered light signal reflected by the sensing optical fiber. The received Brillouin scattered light signal is amplified, filtered, demodulated, and the processed Brillouin scattered light signal is converted into a digital signal for further analysis and calculation to obtain a Brillouin gain spectrum. This step is a prior art and is only briefly described here for simplicity.

[0065] S13. According to the Brillouin gain spectrum, a Brillouin frequency shift extraction network is used to extract the Brillouin frequency shift, thereby obtaining the temperature data, the vibration data and the strain data.

[0066] Specifically, in this embodiment, it is first necessary to construct a Brillouin frequency shift extraction network, and complete the training and verification of the Brillouin frequency shift extraction network. The Brillouin frequency shift extraction network includes a first convolution unit, a second convolution unit and a fully connected unit, and the convolution kernel of the first convolution unit and the convolution kernel of the second convolution unit are of different sizes. Among them, the first convolution unit includes a first convolution sequence, a second convolution sequence and a third convolution sequence, and the second convolution unit includes a fourth convolution sequence and a fifth convolution sequence. The first convolution sequence, the second convolution sequence and the fourth convolution sequence all include a convolution layer, a batch normalization layer, a pooling layer and an activation function, while the third convolution sequence and the fifth convolution sequence only include a convolution layer and a batch normalization layer. The convolution layers of the first convolution sequence to the fifth convolution sequence all have 16 convolution kernels, but the sizes of the convolution kernels are 1×3, 1×4, 1×2, 1×7 and 1×3 respectively, and the step size of the pooling layer is 2. The fully connected unit includes an input layer, three hidden layers and one output layer. The number of neurons in the input layer, hidden layer and output layer can be specifically set and adjusted according to the size of the Brillouin gain spectrum, and the activation function of the hidden layer uses a Gaussian function.

[0067] Furthermore, the Brillouin gain spectrum passes through the first convolution sequence, the second convolution sequence, and the third convolution sequence in the first convolution unit to obtain the first feature; the Brillouin gain spectrum passes through the fourth convolution sequence and the fifth convolution sequence in the second convolution unit to obtain the second feature. Next, the first feature is bilinearly interpolated to match the size of the second feature, and then the convolution layer activation function is used to perform a nonlinear transformation on the first feature and the second feature to obtain feature fusion data. The convolution layer activation function is specifically the ReLU activation function, and the feature fusion data satisfies the following relationship:

[0068]

[0069] in, For feature fusion data, is the first feature after bilinear interpolation, The second feature.

[0070] After the above description, it is easy to know that the Brillouin frequency shift extraction network performs the following steps when running:

[0071] 1. Use a first convolution unit to extract a first feature of the Brillouin gain spectrum, and use the second convolution unit to extract a second feature of the Brillouin gain spectrum.

[0072] 2. Match the first feature and the second feature, and then perform nonlinear transformation on the first feature and the second feature through a convolutional layer activation function to obtain feature fusion data.

[0073] 3. Input the feature fusion data into the fully connected unit and output the Brillouin frequency shift.

[0074] This embodiment obtains information features under different receptive fields through the first convolution unit and the second convolution unit, and performs feature fusion to extract more accurate information features, and then further realizes accurate extraction of Brillouin frequency shift through the fully connected unit, providing a reliable data foundation for subsequent data analysis and model construction.

[0075] Furthermore, after the Brillouin frequency shift extraction network is constructed, it is necessary to construct a corresponding data set to complete the training and verification of the constructed Brillouin frequency shift extraction network. This embodiment generates multiple experimental Brillouin gain spectra by experimental simulation to construct a data set for training and verifying the Brillouin frequency shift extraction network, specifically: setting the Brillouin frequency shift range, and setting multiple Brillouin frequency shifts at equal frequency intervals within the Brillouin frequency shift range; setting the signal-to-noise ratio range, and setting multiple signal-to-noise ratios at equal intervals within the signal-to-noise ratio range; setting the line width range, and setting multiple line widths at equal intervals within the line width range; randomly combining the Brillouin frequency shift, signal-to-noise ratio and line width, simulating the Lorentz curve for each combination of Brillouin frequency shift, signal-to-noise ratio and line width, and then obtaining multiple experimental Brillouin gain spectra.

[0076] After the Brillouin frequency shift is extracted using the Brillouin frequency shift extraction network, temperature data, vibration data and strain data can be further obtained based on the Brillouin frequency shift. The specific acquisition method is the existing technology and will not be described in detail here. Among them, the obtained vibration data includes vibration amplitude, vibration frequency and phase information.

[0077] The present invention obtains the Brillouin gain spectrum by laying the sensing optical fiber on the target road section, and then obtains temperature data, vibration data and strain data by extracting the Brillouin frequency shift, thereby realizing real-time and accurate monitoring of road temperature, vibration and strain, and providing a reliable data basis for subsequent data analysis and model construction.

[0078] S2, dividing the road surface into grids and determining the repair area of ​​each grid node, and then determining the node temperature, node strain and road vibration data according to the temperature data, the strain number and the vibration data.

[0079] In this embodiment, the road surface is divided into grids and node temperatures, node strains and road vibration data are determined based on temperature data, vibration data and strain data to improve the availability of strain data, temperature data and vibration data, which is convenient for subsequent model training and data prediction. Step S2 specifically includes the following steps:

[0080] S21. Use a plurality of horizontal lines and vertical lines to divide the target road section into a plurality of square grids of the same size to form the road surface grid, wherein the intersections of the horizontal lines and the vertical lines are the grid nodes.

[0081] Specifically, in this embodiment, any horizontal line and the vertical line are perpendicular to each other, and the distance between any two adjacent horizontal lines is the same as the distance between any two adjacent vertical lines. Since the target road section is divided into road surface grids using multiple horizontal lines and vertical lines, the horizontal lines and vertical lines are also called grid lines.

[0082] S22. Divide the corresponding repair area with the grid node as the center, the repair area has the same shape and size as the road surface grid, and the midpoint of each edge in the repair area is located on the grid line.

[0083] Specifically, in this embodiment, see Figure 2 , Figure 2 The midpoint O represents a grid node, the solid line box represents the road surface grid, and the dotted line box represents the repair area of ​​the grid node O. For the grid node O, the midpoints of the four sides of its corresponding repair area are all located on the grid lines. Since the repair area has the same size and shape as the road surface grid, the midpoints of the four sides of the repair area are located on the midpoints of the grid lines between the grid node O and the four adjacent grid nodes.

[0084] S23, determining the position of the strain point corresponding to each strain data at the same time on the road surface grid.

[0085] Specifically, in this embodiment, the strain point is the location where the strain is detected on the target road section, and one strain point corresponds to one strain data at one time point. A plane rectangular coordinate system is established in situ with any grid node on the target road section as the coordinate, and the positions of the grid nodes and strain points can be represented using the coordinates on the plane rectangular coordinate system.

[0086] S24: If there is an adjacent strain point to the grid node, the average of the temperature data of the strain points adjacent to the grid node is used as the node temperature of the grid node.

[0087] Specifically, in this embodiment, the strain point adjacent to the grid node referred to in this step refers to a strain point whose line segment defined by the grid node does not exceed the road surface grid where the grid node is located. For ease of understanding, the following will be explained by example:

[0088] For any four adjacent pavement grids, there will be a grid node that is shared by these four pavement grids. This grid node is recorded as node W. Then, any strain point existing in these four pavement grids or on the edges of these four pavement grids will be considered as a strain point adjacent to node W.

[0089] S25. If the grid node does not have any adjacent strain point, the node temperature of the grid node is the ambient temperature.

[0090] S26. If the strain point coincides with a certain grid node, the strain data of the strain point is used as the node strain of the grid node.

[0091] S27: If the strain point is located on a grid line between a pair of adjacent grid nodes, use a strain allocation model to allocate strain data of the strain point to the pair of adjacent grid nodes.

[0092] Specifically, in this embodiment, if a strain point is located on a grid line between a pair of adjacent grid nodes, it is considered that the strain point is only adjacent to the two adjacent grid nodes, and the strain data of the strain point is distributed to the pair of adjacent grid nodes using the strain distribution model. The strain distribution model satisfies the following relationship:

[0093]

[0094] in, is the strain value assigned from strain point i to mesh node j, is the strain data at strain point i, is the distance between strain point i and grid node j, is the number of mesh nodes to which strain values ​​need to be assigned from strain point i, , m is the total number of grid nodes. It is not difficult to know that the maximum value of j in this step is 2. The strain distribution model is simple to calculate, which is conducive to improving the efficiency of road monitoring and can reduce the requirements of road monitoring on computing resources, thereby reducing the quality requirements of road monitoring on hardware equipment.

[0095] Furthermore, since a certain grid node may have multiple adjacent strain points, the node strain of the grid node becomes the sum of the strain values ​​allocated to the grid node from the multiple strain points.

[0096] S28. If the strain point is located in a certain road surface grid, the strain data of the strain point is distributed to four grid nodes adjacent to the strain point using the strain distribution model.

[0097] Specifically, in this embodiment, if the strain point is located in a road surface grid, it is considered that the strain point is adjacent to four grid nodes on the road surface grid, and the strain data of the strain point is distributed to the four grid nodes on the road surface grid using the strain distribution model. It is not difficult to know that the maximum value of j is 4 when the strain distribution model is used in this step.

[0098] S29. If there is no strain point around the grid node, the node strain of the grid node is zero.

[0099] Specifically, in this embodiment, if a grid node is not adjacent to any strain point, the node strain at the grid node is recorded as 0.

[0100] S210: Calculate the average of all the vibration data at the same time as the road vibration data.

[0101] Specifically, in this embodiment, the road vibration data includes road vibration amplitude, road vibration frequency and road vibration phase information. The mean of all vibration amplitudes, the mean of vibration frequencies and the mean of phase information at the same time are calculated respectively, and the calculated mean values ​​are used as the road vibration amplitude, road vibration frequency and road vibration phase information at that time.

[0102] S3. Establish a road data set according to the node temperature, grid node position, node strain and road vibration data, and then construct a vibration prediction model.

[0103] Wherein, step S3 specifically includes the following steps:

[0104] S31. Establishing the road data set according to the node temperature, the grid node position, the node strain and the road vibration data.

[0105] Specifically, in this embodiment, a time step is set, and then multiple input time series and output time series are constructed using the node strains, grid node positions, temperature data, and vibration data of all grid nodes at different time points on the target road section according to the set time step. In order to improve the accuracy of the road load index calculated subsequently, the time step should be set as small as possible, such as 0.05s.

[0106] More specifically, the input time series G can be expressed as , is the node strain of mesh node j at the kth time point, is the horizontal coordinate of the grid node j at the kth time point, is the ordinate of the grid node j at the kth time point, is the node temperature of grid node j at the kth time point, , N is the number of time points in the input time series. The output time series H can be expressed as ,in, is the road vibration amplitude at the kth time point, is the road vibration frequency at the kth time point, is the road vibration phase information at the kth time point.

[0107] Furthermore, if a sufficient number of data sets are not collected on the target road section, the road data set can be supplemented with the node strain, grid node position, temperature data and vibration data of all grid nodes at different time points on other sections, but the number of grid nodes should be consistent.

[0108] S32. Use a long short-term memory recurrent neural network to build an initial vibration prediction model, and use the road data set to complete training and verification of the initial vibration prediction model to obtain the vibration prediction model.

[0109] Specifically, in this embodiment, the input time series corresponding to each grid node is used as input, and the output time series is used as output, and a multi-input single-output long short-term memory recurrent neural network is used to construct an initial vibration prediction model. Then, the road data set is divided into a training set and a validation set in a ratio of 8:2, and the training set and the validation set are used to complete the training and validation of the initial vibration prediction model, respectively, to obtain a vibration prediction model.

[0110] Furthermore, the long short-term memory recurrent neural network can capture the long-term dependencies in time series data and is suitable for the prediction of road vibration data. This method can improve the accuracy and reliability of vibration prediction and provide a more scientific basis for the formulation of subsequent road repair plans.

[0111] S4. Performing simulated repair on the repair area corresponding to each of the grid nodes to update strain data, and simultaneously using the vibration prediction model to obtain vibration prediction data.

[0112] Specifically, in this embodiment, for the node temperature and node strain obtained in real time, the repair area corresponding to any grid node is simulated and repaired, that is, the node strain at the grid node is set to zero, while the node strains of other grid nodes are kept unchanged, and then the node strain of the grid node is updated. After updating the node strain of a certain grid node, the vibration prediction model is used to obtain the vibration prediction data of the target road section after this repair. The repair areas corresponding to each grid node are simulated and repaired respectively, and then the vibration prediction data of the target road section after repairing different repair areas can be obtained.

[0113] S5. Calculate the road bearing index after repairing each of the repair areas using a bearing prediction model based on the vibration prediction data, and then formulate a road repair plan.

[0114] Wherein, step S5 specifically includes the following steps:

[0115] S51. Calculate the road bearing index after repairing each of the repair areas using the bearing prediction model according to the vibration prediction data.

[0116] Specifically, in this embodiment, the load prediction model satisfies the following relationship:

[0117]

[0118] Among them, P is the road load index, is the equivalent elastic modulus of the road design base top, is the thickness of the road subbase layer, is the base frequency. It can be obtained by querying the design plan of the target road section. It is extracted from the vibration prediction data.

[0119] More specifically, the greater the road bearing index, the greater the road bearing capacity. The predicted vibration prediction data is a series of discrete data. In order to extract the fundamental frequency, the obtained vibration prediction data can be linearly interpolated and fitted to obtain the spectrum curve of road vibration, and then the fundamental frequency can be extracted from the obtained spectrum curve. The process of extracting the fundamental frequency according to the vibration prediction data is a prior art means and will not be described in detail here.

[0120] Furthermore, the load prediction model provided in this embodiment can calculate the road load index based on parameters such as the equivalent modulus of resilience of the road design base, the thickness of the road subbase layer and the fundamental frequency. The calculation is simple and has high reliability, which helps to quickly and accurately evaluate the road capacity after repairs in different repair areas, and provides a reliable technical basis for formulating road repair plans. In addition, using the load prediction model to calculate the road load index based on vibration prediction data can also help formulate the optimal repair plan under limited resources, thereby maximizing the service life of the road and improving the road's capacity, and reducing safety hazards in road operation.

[0121] S52: Arrange the road bearing indexes in descending order, and then determine the repair order of each repair area to obtain the road repair plan.

[0122] Specifically, in this embodiment, the larger the road bearing index is, the better the road bearing capacity is restored after the repair area is repaired. Therefore, the repair order of each repair area can be determined according to the size of the road bearing index, and the formulated road repair plan is to repair each repair area in turn according to the repair order.

[0123] S53: Complete the repair of the target road section according to the road repair plan.

[0124] Specifically, in this embodiment, completing the repair of the target road section according to the road repair plan is beneficial to maximally extending the service life of the road and improving the road's traffic capacity under limited resources, and reducing safety hazards in road operation.

[0125] It should be noted that, in some cases, the actions described in the specification can be performed in a different order and still achieve the desired results. In this embodiment, the order of steps given is only to make the embodiment appear clearer and easier to explain, rather than to limit it.

[0126] In an alternative embodiment, see Figure 3 The present invention also provides an optical time-domain road fusion perception monitoring system, which includes: a data acquisition device A1, a data output device A2, a processor A3 and a storage A4, wherein the storage A4 includes a computer-readable storage medium, wherein a computer program is stored in the computer-readable storage medium, wherein the computer program includes program instructions, and when the program instructions are executed by the processor A3, the processor A3 implements the contents described in steps S1 to S5.

[0127] Specifically, in this embodiment, the data acquisition device A1 includes a laser and a photodetector, the laser is used to send a series of short pulse laser signals to the sensing optical fiber to generate Brillouin scattered light signals, and the photodetector is used to receive the Brillouin scattered light signals reflected by the sensing optical fiber. The data output device A2 includes a digital display screen for displaying real-time temperature data, vibration data and strain data, and can output a road repair plan.

[0128] The system can stably run the method, thereby improving the efficiency of road detection and improving the practicality of the method.

[0129] In summary, firstly, the method obtains the Brillouin gain spectrum by laying the sensing optical fiber in the target road section, and constructs a Brillouin frequency shift extraction network to extract the Brillouin frequency shift to obtain temperature data, vibration data and strain data, thereby realizing real-time and accurate monitoring of road temperature, vibration and strain. Secondly, the method of the present invention divides the road surface into grids and determines the node temperature, node strain and road vibration data according to the temperature data, vibration data and strain data to improve the availability of strain data, temperature data and vibration data, and then constructs a vibration prediction model based on a long short-term memory recurrent neural network. Finally, the method predicts the vibration prediction data after the simulated repair of the repair area according to the vibration prediction model, and then calculates the corresponding road bearing index based on the bearing prediction model, which is used to formulate a road repair plan, which is conducive to maximizing the service life of the road and improving the traffic capacity of the road under limited resources, and reducing the safety hazards of road operation. In addition, the system can stably run the method, thereby improving the detection efficiency of the road and improving the practicality of the method.

[0130] It can capture the long-term dependencies in time series data and is suitable for the prediction of road vibration data. This method can improve the accuracy and reliability of vibration prediction and provide a more scientific basis for the formulation of subsequent road repair plans.

[0131] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein by equivalents. These modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be included in the scope of the specification of the present invention.

Claims

1. An optical time-domain road fusion perception monitoring method, characterized in that: The steps include: Laying a sensing optical fiber on a target road section, and collecting temperature data, strain data, and vibration data of the target road section in real time based on the sensing optical fiber; Using a plurality of horizontal lines and vertical lines to divide the target road section into a plurality of square grids of the same size to form a road surface grid, wherein the intersections of the horizontal lines and the vertical lines are grid nodes; Divide the corresponding repair area with the grid node as the center, the repair area has the same shape and size as the road surface grid, and the midpoint of each side in the repair area is located on the grid line; Determine the position of the strain point corresponding to each of the strain data at the same time on the road surface grid; If the grid node has an adjacent strain point, the average of the temperature data of the strain points adjacent to the grid node is used as the node temperature of the grid node; If the grid node does not have the adjacent strain point, the node temperature of the grid node is the ambient temperature; If the strain point coincides with a certain grid node, the strain data of the strain point is used as the node strain of the grid node; If the strain point is located on a grid line between a pair of adjacent grid nodes, the strain data of the strain point is distributed to the pair of adjacent grid nodes using a strain distribution model; If the strain point is located in a certain road surface grid, it is considered that the strain point is adjacent to four grid nodes on the road surface grid, and the strain data of the strain point is distributed to the four grid nodes adjacent to the strain point using a strain distribution model; If there is no strain point around the grid node, the node strain of the grid node is zero; Calculating the average of all the vibration data at the same time as the road vibration data; establishing a road data set based on the node temperatures, the grid node positions, the node strains, and the road vibration data; Using a long short-term memory recurrent neural network to construct an initial vibration prediction model, and using the road data set to complete training and verification of the initial vibration prediction model to obtain a vibration prediction model; Simulating repairs for the repair areas corresponding to the respective grid nodes to update strain data, and acquiring vibration prediction data using the vibration prediction model; Calculating the road bearing index after repairing each of the repair areas using a bearing prediction model according to the vibration prediction data; The load prediction model satisfies the following relationship: ， Wherein, P is the road load index, is the equivalent elastic modulus of the road design base top, is the thickness of the road subbase layer, is the fundamental frequency, Extracted from the vibration prediction data; Arrange the road load indexes in descending order, and then determine the repair order of each repair area to obtain a road repair plan; Complete the repair of the target road section according to the road repair plan; When acquiring the node temperature, the strain point adjacent to the grid node refers to a strain point whose line segment defined by the grid node does not exceed the road surface grid where the grid node is located.

2. The optical time-domain road fusion perception monitoring method according to claim 1 is characterized in that: The step of laying a sensing optical fiber on the target road section and collecting temperature data, strain data and vibration data of the target road section in real time based on the sensing optical fiber comprises the following steps: Laying the sensing optical fiber on the target road section; Sending an optical pulse signal to the sensing optical fiber, and receiving a Brillouin scattered light signal reflected by the sensing optical fiber to obtain a Brillouin gain spectrum; According to the Brillouin gain spectrum, a Brillouin frequency shift extraction network is used to extract the Brillouin frequency shift, thereby obtaining the temperature data, the vibration data and the strain data.

3. The optical time-domain road fusion perception monitoring method according to claim 2 is characterized by: The Brillouin frequency shift extraction network includes a first convolution unit, a second convolution unit and a fully connected unit, and the convolution kernel of the first convolution unit and the convolution kernel of the second convolution unit have different sizes.

4. The optical time-domain road fusion perception monitoring method according to claim 3 is characterized in that: The Brillouin frequency shift extraction network performs the following steps when running: Extracting a first feature of the Brillouin gain spectrum using a first convolution unit, and extracting a second feature of the Brillouin gain spectrum using the second convolution unit; Matching the first feature and the second feature, and then performing nonlinear transformation on the first feature and the second feature through a convolution layer activation function to obtain feature fusion data; The feature fusion data is input into the fully connected unit, and the Brillouin frequency shift is output.

5. The optical time-domain road fusion perception monitoring method according to claim 1 is characterized in that: The strain distribution model satisfies the following relationship: ， in, is the strain value assigned from strain point i to mesh node j, is the strain data of strain point i, is the distance between strain point i and grid node j, is the number of mesh nodes to which strain values ​​need to be assigned from strain point i, , m is the total number of the grid nodes.

6. An optical time-domain road fusion perception monitoring system, characterized in that: The optical time-domain road fusion perception monitoring system includes: a data acquisition device, a data output device, a processor and a storage device, the storage device includes a computer-readable storage medium, the computer-readable storage medium stores a computer program, the computer program includes program instructions, and when the program instructions are executed by the processor, the processor implements the optical time-domain road fusion perception monitoring method as described in any one of claims 1-5.

Images