A method for determining the deployment time of bridge monitoring sensors

Abstract

The present invention provides a method for judging the deployment duration of bridge monitoring sensors, which belongs to the technical field of bridge structure monitoring. The method for judging the deployment duration of bridge monitoring sensors calculates the node positions of changes in monitoring data required for the bridge based on measured bridge parameters; establishes a monitoring data set using the node positions of changes in monitoring data required for the bridge and the monitoring sensor data corresponding to each node; extracts the node positions of changes in monitoring data required for the bridge and the monitoring sensor data corresponding to each node in the monitoring data set to establish a neural network model; trains the neural network model, and divides the node positions of changes in monitoring data required for the bridge into permanent monitoring nodes and periodic monitoring nodes; the method screens the node positions required to monitor the bridge through a neural network, selects permanent monitoring nodes and periodic monitoring nodes, and solves the problem of judging the deployment duration of bridge sensors.

Description

Technical Field

[0001] The present invention belongs to the technical field of bridge structure monitoring, and in particular relates to a method for determining the deployment duration of bridge monitoring sensors. Background Art

[0002] The Chinese invention patent with publication number CN107247853B (application number: CN201710521762.X) provides a method for arranging measurement points for dynamic load testing of bridges. This invention combines finite element analysis technology with the theory of structural dynamic measurement point selection to quickly and accurately analyze the target mode of the target bridge and obtain the measurement point layout positions distributed along the longitudinal or planar distribution of the bridge. This method is simple and efficient, and the picked-up signals are more representative, thereby improving the ability to identify the lateral target modes of the bridge. The Chinese invention patent with publication number CN107066748B (application number: CN201710273398.X) provides a method for deploying wireless sensors for wind-induced vibration monitoring of bridges. This invention is based on the finite element model of the bridge and the standard wind spectrum to simulate the wind field, and adopts the time domain calculation method of the bridge wind-induced vibration response to realize the deployment of wireless sensors for wind-induced vibration monitoring of bridges.

[0003] In the above two inventions, only the layout positions of the bridge monitoring points are calculated, and the node positions that need to be monitored for data changes in real time cannot be selected, and the layout duration of the sensors at each monitoring point cannot be determined. Summary of the Invention

[0004] In view of this, the present invention provides a method for determining the deployment duration of bridge monitoring sensors, which can select the node locations where real-time monitoring data changes are required, determine the deployment duration of sensors at each monitoring point, and effectively solve the above problems.

[0005] The present invention is achieved in that:

[0006] The present invention provides a method for determining the deployment duration of bridge monitoring sensors, which includes the following steps:

[0007] S01: Based on the measured bridge parameters, the finite element algorithm is used to calculate the node locations where the bridge monitoring data changes;

[0008] S02: placing monitoring sensors at node locations where the bridge requires monitoring data changes, and recording monitoring sensor data;

[0009] S03: establishing a monitoring data set using the node positions of the bridge's required monitoring data changes and the monitoring sensor data corresponding to each node;

[0010] S04: extracting the node positions of the bridge required monitoring data changes from the monitoring data set and the monitoring sensor data corresponding to each node to establish a neural network model;

[0011] S05: training the neural network model, and dividing the node positions of the bridge requiring monitoring data changes into permanent monitoring nodes and periodic monitoring nodes;

[0012] S06: retain the monitoring sensor of the permanent monitoring node and remove the monitoring sensor of the periodic monitoring node;

[0013] S07: Perform periodic data measurement on the periodic monitoring node.

[0014] On the basis of the above technical solution, the method for determining the deployment duration of a bridge monitoring sensor of the present invention can also be improved as follows:

[0015] The specific operation steps in S01 include:

[0016] Step 1: Establish a finite element model of the target structure on the finite element platform;

[0017] Step 2: Extract the target modal vibration matrix U based on the finite element model. The target modal vibration matrix U is composed of node vibration data. The target modal vibration matrix U is:

[0018]

[0019] Where: m is the target mode number, n is the number of model degrees of freedom, that is, the number of nodes, u ij is the j-order vibration mode data of node i, where 1≤i≤n, 1≤j≤m;

[0020] Step 3: Remove nodes in the target modal shape matrix that are not suitable for arranging measurement points and whose vibration signals are less than the first set threshold. For the target modal shape matrix U, the rows corresponding to the removed nodes are removed, and the remaining nodes form the preliminary degrees of freedom.

[0021] Step 4: Calculate the MAC matrix for any combination of two nodes in the initial degree of freedom. The MAC matrix is:

[0022]

[0023] Where: a ij For the matrix A = U T ×U ij element, where U T is the transposed matrix of U;

[0024] The MAC matrix (modal assurance criterion) is a tool for evaluating the intersection angle of modal vector space and is an existing technology.

[0025] Step 5: Select the Z group of nodes with the smallest maximum off-diagonal element value, Z ≥ 1;

[0026] Step 6: Select the best two nodes from the Z group of nodes to measure the degree of freedom:

[0027] ① When Z is 1, the group of nodes enters the measurement degree of freedom as the optimal two measurement points and proceeds to step 7;

[0028] ② When Z is greater than or equal to 2, multiply the mode shape data of the two nodes in each group of Z nodes, and the group of nodes with the largest product value is used as the optimal two measurement points to measure the degree of freedom, or arbitrarily select a group of nodes from two or more groups of nodes with the largest and identical product values ​​as the optimal two measurement points to measure the degree of freedom, and then go to step 7;

[0029] Step 7: Select one node from the remaining nodes of the initially selected degree of freedom one by one to enter the measurement degree of freedom, and calculate the MAC matrix of the measurement degree of freedom to obtain the maximum off-diagonal element value. If the maximum off-diagonal element value is less than or equal to the second set threshold, select the newly added node to enter the measurement degree of freedom and proceed to step 8;

[0030] Otherwise, select the node with the smallest maximum off-diagonal element value to enter the measurement degree of freedom, and repeat this step until the maximum off-diagonal element value is less than or equal to the second set threshold. Select the newly added node to enter the measurement degree of freedom, and go to step 8;

[0031] Step 8: Get the calculation results.

[0032] Furthermore, the first set threshold in the third step is the vibration signal value of the 0.5Ath node arranged in ascending order of the vibration signal values ​​of the nodes, where A is the total number of nodes.

[0033] Furthermore, in the seventh step, the second threshold is set to 0.25.

[0034] The specific operation steps in S04 include:

[0035] Step 1: Select the node locations of the bridge required monitoring data changes in the monitoring data set and the monitoring sensor data corresponding to each node to establish a neural network model;

[0036] Step 2: Design the network architecture and initialization, specifically: build a convolutional neural network, which contains 1 input layer, 3 convolution layers, 3 Relu nonlinear activation layers, 3 pooling layers, 1 fully connected layer and 1 output layer.

[0037] The specific operation steps of training the neural network model in S05 include:

[0038] The frequency of change of monitoring sensor data corresponding to the node is obtained through convolutional neural network training, which specifically includes two stages: forward propagation training and back propagation network training. The forward propagation network training is to input the node position of the monitoring data change required for the bridge and the feature image of the monitoring sensor data corresponding to each node, and convert the obtained feature image into a one-dimensional vector, input it into the fully connected layer, and the analyzer obtains the recognition result, that is, the output vector. Each value of the output vector represents the frequency of change of monitoring sensor data corresponding to the node of the input data block; the back propagation network training is: when the output result of the forward propagation network training does not match the expected output, the stochastic gradient descent optimization algorithm is used to perform back network training to update the parameters of the convolution layer.

[0039] Among them, the permanent monitoring node in S05 is the node whose corresponding monitoring sensor data has a high frequency of change, that is, the node that needs real-time monitoring; the periodic monitoring node is the node whose corresponding monitoring sensor data has a low frequency of change, that is, the node that does not need real-time monitoring. The frequency of change of the monitoring sensor data corresponding to the node is judged by the neural network model.

[0040] The bridge parameters measured in S01 at least include the span, width, clearance, arch rise, length and cross-sectional dimensions of the bridge components, thickness of the bridge deck pavement, and vibration mode data of each node.

[0041] The monitoring data required for the bridge in S01 at least includes stress data at bridge node positions and transverse and vertical vibration mode data.

[0042] The periodic data measurement in S07 is performed once every 24 hours.

[0043] Compared with the prior art, the beneficial effects of the method for judging the deployment duration of bridge monitoring sensors provided by the present invention are: the present invention relates to the field of bridge structure monitoring technology, and in particular to a sensor deployment location that calculates the data changes of the monitoring nodes required for the bridge through a finite element algorithm; and uses a neural network model to analyze the frequency of changes in monitoring point data, and screens out sensor deployment locations that require real-time monitoring of data changes, sets permanent monitoring points for sensor deployment locations that require real-time monitoring of data changes, and sets periodic monitoring points for sensor deployment locations that do not require real-time monitoring of data changes; by setting periodic monitoring points, the present invention eliminates the need for real-time monitoring of all monitoring nodes, thereby saving a large amount of resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments of the present invention. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0045] Figure 1 Schematic diagram of the steps of the present invention. DETAILED DESCRIPTION

[0046] To make the purpose, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0047] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention as claimed, but rather merely represents selected embodiments of the present invention. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without creative effort are also within the scope of protection of the present invention.

[0048] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, it does not require further definition or explanation in subsequent drawings.

[0049] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "length", "width", "thickness", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", "clockwise", "counterclockwise" and the like to indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as limiting the present invention.

[0050] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of the technical features being referred to. Thus, a feature identified as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, "plurality" means two or more, unless otherwise specifically defined.

[0051] like Figure 1 FIG. 1 is a flow chart of a method for determining the deployment duration of bridge monitoring sensors provided by the present invention, comprising the following steps:

[0052] S01: Based on the measured bridge parameters, the finite element algorithm is used to calculate the node locations where the bridge monitoring data changes;

[0053] S02: Place monitoring sensors at the nodes where the bridge requires monitoring data changes, and record the monitoring sensor data;

[0054] S03: establishing a monitoring data set using the node locations of the bridge's required monitoring data changes and the monitoring sensor data corresponding to each node;

[0055] S04: extracting the node locations of the bridges’ required monitoring data changes from the monitoring data set and the monitoring sensor data corresponding to each node to establish a neural network model;

[0056] S05: Train the neural network model and divide the node locations where the bridge needs to monitor data changes into permanent monitoring nodes and periodic monitoring nodes;

[0057] S06: Keep the monitoring sensors of the permanent monitoring nodes and remove the monitoring sensors of the periodic nodes;

[0058] S07: Perform periodic data measurement on the periodic monitoring node.

[0059] Among them, in the above technical solution, the permanent monitoring node in S05 is a node whose corresponding monitoring sensor data has a high frequency of change and requires real-time monitoring, and the periodic monitoring node is a node whose corresponding monitoring sensor data has a low frequency of change and does not require real-time monitoring. Among them, the frequency of change of the monitoring sensor data corresponding to the node is judged by the neural network model.

[0060] Among them, in the above technical solution, the specific operation steps in S01 include:

[0061] Step 1: Establish a finite element model of the target structure on the finite element platform:

[0062] In this example, the target structure is a simply supported prefabricated small box bridge with a calculated span of 29.14m and a width of 18.75m. Due to the bridge's large width-to-length ratio, the target modes are the first three orders, requiring investigation of the bridge's lateral deflection modes. A planar finite element model was constructed using Midas / Civil.

[0063] Step 2: Extract the target modal vibration matrix U based on the finite element model.

[0064] The first and second steps combine finite element analysis technology to obtain the target modal vibration matrix of the target structure by establishing a finite element analysis model of the target structure on the computer;

[0065] The target modal vibration matrix U is composed of node vibration shape data. The target modal vibration matrix U is:

[0066]

[0067] Where m is the target modal number, n is the model's degrees of freedom (i.e., the number of nodes), and uij is the jth-order mode shape data for node i, where 1 ≤ i ≤ n and 1 ≤ j ≤ m. A computer program automatically calculates and analyzes dynamic performance, extracting the first three-order mode shape matrices.

[0068] Step 3: Remove nodes in the target modal vibration matrix that are not suitable for arranging measurement points and whose vibration signals are less than the first set threshold. For the target modal vibration matrix U, the rows corresponding to the removed nodes are removed, and the remaining nodes form the preliminary degrees of freedom.

[0069] Step 4: Calculate the MAC matrix for any combination of two nodes in the initial degree of freedom. The MAC matrix is:

[0070]

[0071] Where: a ij For the matrix A = U T ×U ij element, where U T is the transposed matrix of U.

[0072] Step 5: Select Z groups of nodes with the smallest maximum off-diagonal element values, where Z ≥ 1. In this embodiment, there are 1569 nodes. After removing nodes in the target modal shape matrix that are not suitable for measurement point placement and whose vibration signals are less than the first set threshold, 706 nodes remain. The MAC matrix of any two nodes within this range is calculated and the maximum off-diagonal element values ​​are compared. Three groups of nodes with the smallest maximum off-diagonal element values ​​are obtained. The maximum off-diagonal element value of each of these three groups is 0.6. Each of these three groups of nodes consists of two nodes: node 205# and node 1165#; node 453# and node 1367#; and node 489# and node 1068#.

[0073] Step 6: Select the two best nodes from the three groups of nodes to measure the degrees of freedom. The mode shape data for each node are: 0.61 for node 205; 0.62 for node 1165; 0.52 for node 453; 0.77 for node 1367; 0.49 for node 489; and 0.37 for node 1068. Multiplying the mode shape data for each group of two nodes yields: 0.61 × 0.62 = 0.3782; 0.52 × 0.77 = 0.4004; and 0.49 × 0.37 = 0.1813. The second group of nodes has the largest product value, so the two nodes in this second group are selected as the best two measurement points for measuring the degrees of freedom.

[0074] Step 7: Using the step-by-step accumulation method, select one node from the remaining nodes of the initially selected degree of freedom one by one to enter the measurement degree of freedom, and calculate the MAC matrix of the measurement degree of freedom to obtain the maximum off-diagonal element value. If the maximum off-diagonal element value is less than or equal to the second set threshold, select the newly added node to enter the measurement degree of freedom, and proceed to step 8;

[0075] Otherwise, select the node with the smallest maximum off-diagonal element value to enter the measurement degree of freedom, then calculate the MAC matrix of the measurement degree of freedom to obtain the maximum off-diagonal element value. If the maximum off-diagonal element value is less than or equal to the second set threshold, select the newly added node to enter the measurement degree of freedom and proceed to step 8. Otherwise, repeat this step until the maximum off-diagonal element value is less than or equal to the second set threshold, select the newly added node to enter the measurement degree of freedom and proceed to step 8.

[0076] Using a step-by-step accumulation method, select one node from the remaining nodes of the initially selected degree of freedom one by one to enter the measurement degree of freedom, and calculate the MAC matrix of the measurement degree of freedom to obtain the maximum off-diagonal element value. If the maximum off-diagonal element value is less than or equal to the second set threshold, select the newly added node to enter the measurement degree of freedom, and proceed to step 8;

[0077] In this embodiment, the second set threshold is 0.25, and nodes 156#, 157#, and 158# are selected from the 706 nodes in turn to enter the measurement freedom degree that already includes 453# and 1367#, and the MAC matrix of the measurement freedom degree is calculated, and the maximum off-diagonal element value is examined. It is found that after all nodes are calculated, no group's maximum off-diagonal element value is less than the second set threshold. At this time, the node with the smallest maximum off-diagonal element value, that is, node 759#, is left to enter the measurement freedom degree, and the nodes that have not been selected are continuously selected one by one to enter the measurement freedom degree and calculate the MAC matrix and examine the maximum off-diagonal element value. It is found that when the calculation reaches node 947#, the maximum off-diagonal element value of the node is less than the second set threshold.

[0078] Step 8: Obtain the calculation results and set the measurement degrees of freedom to nodes 453#, 1367#, 759#, and 947#. When sensors are placed at these four nodes, a set of measurement points that meets the requirements is obtained.

[0079] In other embodiments, after completing step 7, if all remaining nodes of the initially selected degrees of freedom are selected and still cannot satisfy the requirement that the maximum off-diagonal element value is less than or equal to the second set threshold, then it means that the first set threshold is too large or the second set threshold is too small. In this case, either of the following two methods can be used:

[0080] ② Reduce the first set threshold and go to the third step;

[0081] ② Increase the second set threshold and go to step 7.

[0082] Among them, in the above technical solution, the specific operation steps in S04 include:

[0083] Step 1: Select the node locations of the bridges in the monitoring data set where the monitoring data changes and the monitoring sensor data corresponding to each node to establish a neural network model;

[0084] Step 2: Design the network architecture and initialization, specifically: build a convolutional neural network, which contains 1 input layer, 3 convolution layers, 3 Relu nonlinear activation layers, 3 pooling layers, 1 fully connected layer and 1 output layer.

[0085] Input layer. It consists of the node locations of the bridge’s required monitoring data changes and the monitoring sensor data corresponding to each node;

[0086] Output layer. The output layer is a node that represents the frequency of changes in the monitoring sensor data corresponding to the sensor deployment node.

[0087] Convolutional layer. The number of convolution kernels in the three convolutional layers is 4, 8, and 16 respectively, and the size is 3×3. In order to make the feature extraction show a nonlinear relationship, each convolutional layer is followed by an activation function. The ReLU activation function is used, and its function expression is as follows:

[0088]

[0089] Pooling layer. Choose max pooling with a size of 2×2 and a stride of 1.

[0090] Fully connected layer. The features extracted by the convolutional and pooling layers are converted into one-dimensional vectors and fully connected to the four nodes of the output layer to achieve classification.

[0091] Among them, in the above technical solution, the specific operation steps of training the neural network model in S05 include:

[0092] The frequency of change of monitoring sensor data corresponding to the node is obtained through convolutional neural network training, which specifically includes two stages: forward propagation training and back propagation network training. Forward propagation network training is to extract the feature image of the node position where the monitoring data of the bridge required to be monitored changes and the monitoring sensor data corresponding to each node, and convert the obtained feature image into a one-dimensional vector, input it into the fully connected layer, and the analyzer obtains the recognition result, that is, the output vector. Each value of the output vector represents the frequency of change of monitoring sensor data corresponding to the node of the input data block; back propagation network training is: when the output result of the forward propagation network training does not match the expected output, the stochastic gradient descent optimization algorithm is used to perform back network training to update the parameters of the convolution layer.

[0093] Forward propagation network training. The input bridge node locations where the monitoring data changes are required and the feature images of the monitoring sensor data corresponding to each node are extracted. The obtained feature images are converted into one-dimensional vectors and input into the fully connected layer. The analyzer obtains the recognition result, namely the output vector. Each value of the output vector represents the frequency of change of the monitoring sensor data corresponding to the node in the input data block.

[0094] Backward propagation network training. When the output of the forward propagation network training does not match the expected output, the stochastic gradient descent optimization algorithm is used to perform reverse network training and update the parameters of the convolutional layer. When calculating the error between the actual output of the convolutional neural network and the expected output, the mean square error loss function is used:

[0095]

[0096] Among them, y i and o i are the output and expected output of the convolutional neural network respectively, and c is the output category. Since there are 2 expected outputs, c = 2.

[0097] Through forward network and back-propagation network training, a bridge monitoring sensor deployment location screening model is formed.

[0098] Among them, in the above technical solution, the bridge parameters measured in S01 include at least the span, width, clearance, arch height, length and cross-sectional dimensions of the bridge components, bridge deck pavement thickness, and vibration mode data of each node.

[0099] Among them, in the above technical solution, the monitoring data required for the bridge in S01 at least includes stress data of bridge node positions and transverse and vertical vibration mode data.

[0100] Furthermore, in the above technical solution, the first set threshold in the third step is the vibration signal value of the 0.5Ath node arranged from small to large, where A is the total number of nodes.

[0101] Among them, in the above technical solution, the periodic data measurement in S07 is measured once every 24 hours.

[0102] Furthermore, in the above technical solution, the second threshold value in the seventh step is set to 0.25.

[0103] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims.

Claims

1. A method for determining the deployment duration of bridge monitoring sensors, characterized in that: The following steps are involved: S01: Based on the measured bridge parameters, the finite element algorithm is used to calculate the node locations where the bridge monitoring data changes; S02: placing monitoring sensors at node locations where the bridge requires monitoring data changes, and recording monitoring sensor data; S03: establishing a monitoring data set using the node positions of the bridge's required monitoring data changes and the monitoring sensor data corresponding to each node; S04: extracting the node positions of the bridge required monitoring data changes from the monitoring data set and the monitoring sensor data corresponding to each node to establish a neural network model; S05: training the neural network model, and dividing the node positions of the bridge requiring monitoring data changes into permanent monitoring nodes and periodic monitoring nodes; S06: retain the monitoring sensor of the permanent monitoring node and remove the monitoring sensor of the periodic monitoring node; S07: performing periodic data measurement on the periodic monitoring node; The specific operation steps of training the neural network model in S05 include: The frequency of change of the monitoring sensor data corresponding to the node is obtained by convolutional neural network training, which specifically includes two stages: forward propagation training and back propagation network training. The forward propagation network training is to input the node position of the monitoring data change required for the bridge and the feature image of the monitoring sensor data corresponding to each node, and convert the obtained feature image into a one-dimensional vector, input it into a fully connected layer, and obtain the recognition result, i.e., the output vector, by the analyzer. Each value of the output vector represents the frequency of change of the monitoring sensor data corresponding to the node of the input data block; the back propagation network training is: when the output result of the forward propagation network training does not match the expected output, the stochastic gradient descent optimization algorithm is used to perform back network training to update the parameters of the convolution layer; the permanent monitoring node in S05 is the node whose corresponding monitoring sensor data has a high frequency of change, that is, the node that needs real-time monitoring, and the periodic monitoring node is the node whose corresponding monitoring sensor data has a low frequency of change, that is, the node that does not need real-time monitoring, wherein the high or low frequency of change of the monitoring sensor data corresponding to the node is judged by the neural network model.

2. The method according to claim 1, characterized in that The specific operation steps in S01 include: Step 1: Establish a finite element model of the target structure on the finite element platform; Step 2: Extract the target modal vibration matrix U based on the finite element model. The target modal vibration matrix U is composed of node vibration data. The target modal vibration matrix U is: ； Where: m is the target mode number, n is the number of model degrees of freedom, that is, the number of nodes, is the j-order vibration mode data of node i, where 1≤i≤n, 1≤j≤m; Step 3: Remove nodes in the target modal shape matrix that are not suitable for arranging measurement points and whose vibration signals are less than the first set threshold. For the target modal shape matrix U, the rows corresponding to the removed nodes are removed, and the remaining nodes form the preliminary degrees of freedom. Step 4: Calculate the MAC matrix for any combination of two nodes in the initial degree of freedom. The MAC matrix is: ； Where: For matrix A= ×U's ij element, where is the transposed matrix of U; Step 5: Select the Z group of nodes with the smallest maximum off-diagonal element value, Z ≥ 1; Step 6: Select the best two nodes from the Z group of nodes to measure the degree of freedom: ① When Z is 1, the group of nodes enters the measurement degree of freedom as the optimal two measurement points and proceeds to step 7; ② When Z is greater than or equal to 2, multiply the mode shape data of the two nodes in each group of Z nodes, and the group of nodes with the largest product value is used as the optimal two measurement points to measure the degree of freedom, or arbitrarily select a group of nodes from two or more groups of nodes with the largest and identical product values ​​as the optimal two measurement points to measure the degree of freedom, and then go to step 7; Step 7: Select one node from the remaining nodes of the initially selected degree of freedom one by one to enter the measurement degree of freedom, and calculate the MAC matrix of the measurement degree of freedom to obtain the maximum off-diagonal element value. If the maximum off-diagonal element value is less than or equal to the second set threshold, select the newly added node to enter the measurement degree of freedom and proceed to step 8; Otherwise, select the node with the smallest maximum off-diagonal element value to enter the measurement degree of freedom, and repeat this step until the maximum off-diagonal element value is less than or equal to the second set threshold. Select the newly added node to enter the measurement degree of freedom, and go to step 8; Step 8: Get the calculation results.

3. The method according to claim 2, characterized in that In the third step, the first set threshold is the vibration signal value of the 0.5Ath node arranged from small to large, where A is the total number of nodes.

4. The method according to claim 2, characterized in that The second set threshold in the seventh step is 0.

25.

5. The method according to claim 1, wherein The specific operation steps in S04 include: Step 1: Select the node locations of the bridge required monitoring data changes in the monitoring data set and the monitoring sensor data corresponding to each node to establish a neural network model; Step 2: Design the network architecture and initialization, specifically: build a convolutional neural network, which contains 1 input layer, 3 convolution layers, 3 Relu nonlinear activation layers, 3 pooling layers, 1 fully connected layer and 1 output layer.

6. The method according to claim 1, characterized in that The bridge parameters measured in S01 at least include the span, width, clearance, arch rise, length and cross-sectional dimensions of the bridge components, thickness of the bridge deck pavement, and vibration mode data of each node.

7. The method according to claim 1, characterized in that The monitoring data required for the bridge in S01 at least includes stress data at bridge node positions and transverse and vertical vibration mode data.

8. The method according to claim 1, characterized in that The periodic data measurement in S07 is performed once every 24 hours.

Images