Method and system for constructing and diagnosing three-dimensional space-time diagram of truss bridge fracture fault

Abstract

The invention discloses a truss bridge fracture fault three-dimensional space-time diagram construction and diagnosis method and system, and relates to the technical field of structure health monitoring, and the method comprises the steps: collecting truss bridge data, and carrying out the preprocessing; performing feature extraction and correlation calculation; updating a distance coefficient between the nodes, and generating a three-dimensional space-time diagram based on a point set, an edge set and an adjacent set in the three-dimensional diagram structure; establishing a three-dimensional physical model of the bridge, combining the acquired space-time signals with a three-dimensional structure, and constructing multi-dimensional fusion features among nodes; the reliability and accuracy of the algorithm are verified by simulating a truss bridge in a healthy state. According to the invention, a truss bridge fracture diagnosis algorithm based on a three-dimensional space-time diagram signal is introduced, and cross edge weight standard deviation characteristics and a multi-dimensional data fusion technology are combined, so that the problems of inaccurate correlation analysis between nodes and untimely fault diagnosis in existing structure health monitoring are solved; the effects of efficiently reflecting the health state of the truss bridge in real time, accurately positioning the structural damage position and realizing early risk early warning are achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of structural health monitoring, and in particular to a method and system for constructing and diagnosing a three-dimensional spatio-temporal diagram of a truss bridge fracture fault. Background Technique

[0002] Due to advantages such as high strength, high stability, economy, and wide application range, truss bridges have become an ideal design choice for long-span bridges in civil engineering. However, in the face of climate change and high-intensity use, as well as the importance and high cost of bridges in the national economy and people's lives, it is necessary to monitor the health status of bridges in order to maintain and repair them when needed. Structural health monitoring is a process of real-time sensing, identifying, and evaluating the safety and performance evolution of a structure by mimicking the self-perception and self-diagnosis capabilities of humans. The structural health monitoring of bridges is a process of collecting and analyzing data such as temperature, load, and vibration on bridge structures using contact or non-contact sensors, and judging the structural health status based on the calculation results to further predict the remaining life of the building. The structural health monitoring technology of truss bridges can reduce the maintenance cost after a fault occurs through fault warning, effectively extend the structural life, and improve safety. The existing truss bridge fault diagnosis algorithms are facing the continuous development of the timeliness and interpretability of signal processing technology diagnosis.

[0003] With the development of intelligent sensors, communication technologies, and intelligent algorithms, it has become possible to continuously monitor the dynamic responses of civil structures and buildings. For the data science and engineering technologies for processing and calculating a large amount of data in the structural health monitoring process in other fields, a large number of studies have been carried out and applied to the dynamic monitoring of civil engineering structures, including data acquisition algorithms based on compressive sampling, abnormal data diagnosis methods using deep learning algorithms, crack identification methods using computer vision technology, and bridge state evaluation methods using machine learning algorithms. The structural health monitoring in civil engineering is divided into two main sub-fields: vibration-based SHM and image-based SHM. With the development of material science and technology and the continuous improvement of modern design processes, the popularity of large-scale structures such as steel structures in construction, and the increasing scale of building complexes and the size and complexity of individual structures, it is difficult to conduct overall monitoring and analysis of the entire building complex or combined structure using the existing traditional single-point sampling analysis. Therefore, it is necessary to introduce new technical theories to establish a system-level structural health monitoring system.

[0004] Graph signal processing, as an emerging technology, has demonstrated its powerful potential and advantages in multiple fields. Graph signal processing can handle complex network structures and spatio-temporal data, providing more comprehensive and accurate monitoring results. The research progress of graph signal processing technology in the fields of signal sampling, filtering, and graph learning is remarkable, especially in the application scenarios of sensor network data, image processing, etc. The graph Fourier transform effectively establishes the connection between the signal node domain and the graph domain. On this basis, graph signal processing extends the ideas of transformation, filtering, etc. in classical signal processing, forming a graph signal processing system with the main research contents of sampling and reconstruction of signals on graphs, graph filter design, graph frequency domain analysis, sparse signal graph representation, and stationary signal processing on graphs. Currently, graph signal processing has been successfully applied in fields such as social computing, information retrieval, computer vision, mechanical fault diagnosis, bioinformatics, knowledge graphs, physics, graph-domain machine learning, and clustering analysis. Some people have proposed low-pass graph filters, revealing that graph topologies can be efficiently represented by graph sampling methods and applied in fields such as missing data recovery and graph data denoising. Some people have proposed a reconstruction model for band-limited graph signals based on graph spectrum domain shifting for the problem of reconstructing band-limited graph signals. This model models the identity invariance property of graph band-limited components as a least squares problem, improving the iterative efficiency and reconstruction accuracy. Some people have also studied graph filtering methods when only the graph structure is known. Based on the Fourier-Galerkin translation operator, they have proposed a graph filter design method that relies only on the probabilistic structure of the graph and verified the effectiveness of the method on different graph random models. As a complex data structure type, a graph can represent numerous structural features containing rich information and can well characterize the interaction relationships between various complex data feature sets. Graph signal processing can analyze multi-dimensional data separately and establish correlations on graphs, with higher accuracy compared to existing one-dimensional data processing algorithms and stronger data correlations compared to existing multi-dimensional data processing algorithms. Graph signal processing has established a basic theoretical system similar to classical signal processing techniques, making its application in civil engineering possible. Analyzing and extracting features from graph signals of complex structure monitoring can more accurately obtain spatial relationships and temporal dynamic changes, and can accurately reflect the change law of the entire system state. Bridges, as a type of large infrastructure, have continuity, systematicness, and structural stability, and their structural nodes and time nodes have strong correlations, making them very suitable for applying graph signal processing technology. Summary of the Invention

[0005] In view of the above existing problems, the present invention is proposed.

[0006] Therefore, the technical problem solved by the present invention is: how to accurately identify the dynamic changes between truss bridge nodes and effectively warn of fracture faults based on the construction and analysis of three-dimensional spatio-temporal graph signals.

[0007] To solve the above technical problems, the present invention provides the following technical solutions: A method for constructing and diagnosing a three-dimensional spatio-temporal diagram of a truss bridge fracture fault, which includes the following steps,

[0008] Arrange sensors to collect truss bridge data and perform preprocessing;

[0009] Extract features from the preprocessed data and perform correlation calculations;

[0010] Update the distance coefficient between nodes according to the correlation calculation results, and generate a three-dimensional spatio-temporal diagram based on the point set, edge set, and neighbor set in the three-dimensional graph structure;

[0011] Establish a three-dimensional physical model of the bridge, combine the collected spatio-temporal signals with the three-dimensional structure, and construct a multi-dimensional fusion feature between nodes;

[0012] Verify the reliability and accuracy of the algorithm under different conditions by simulating truss bridges in different health states.

[0013] As a preferred solution of a method for constructing and diagnosing a three-dimensional spatio-temporal diagram of a truss bridge fracture fault according to the present invention, wherein: arranging sensors to collect truss bridge data is to arrange multi-channel acceleration sensors on each node of the truss bridge to collect vibration signals.

[0014] As a preferred solution of a method for constructing and diagnosing a three-dimensional spatio-temporal diagram of a truss bridge fracture fault according to the present invention, wherein: the preprocessing is to perform Hilbert transform on the original signal, obtain the upper and lower peak envelopes of the signal, align the signals by the sliding window method, intercept the effective signals and reduce the dimension of the signals, retaining the main signals of the key nodes and the secondary signals of the adjacent nodes;

[0015] Perform Hilbert transform on several channel signals of signal S to obtain upper and lower signals respectively, and divide the vibration signals into positive vibration signals and negative vibration signals according to positive and negative values. The expression is:

[0016]

[0017] where i is the channel number of signal S, and are the upper and lower envelope signals of the vibration signal respectively;

[0018] and The Hilbert transform of is:

[0019]

[0020] where, are respectively and The value of the Hilbert transform at time t, where P.V. represents the "principal value". The integral used in the Hilbert transform is a singular integral and is solved in the form of the Cauchy principal value. τ is the variable in the integral and represents past or future time;

[0021] Construct the analytic signal of the original signal, and the expression is:

[0022]

[0023] where are the replicated analytic signals of the up and down vibration signals respectively, and j is the imaginary unit;

[0024] Solve the modulus value of the analytic signal as the envelope signal. The lower envelope signal is the negative value of the modulus, and the expression is:

[0025]

[0026] where are the upper envelope signal and the lower envelope signal respectively;

[0027] Use a sliding window to remove invalid empty-sampled signals. Determine the moment when the vehicle gets on the bridge by identifying the change in signal energy, intercept the corresponding effective signal segment, extract signals from the starting point using two windows with a width of ΔL, and the distance between the two windows is T 2 -ΔL, and calculate the difference between the upper and lower envelope peaks of the signals in the two windows and E i (t 1 ), and the expression is:

[0028]

[0029] where t 1 is the starting point of the front moving window, and fs is the sampling frequency;

[0030] When t 1 is the moment when the front wheel of the vehicle drives onto the bridge, E i (t 1 ) reaches the maximum value. When the signal in the moving window contains empty-sampled time signals, the energy value of the signal is less than the maximum value of E i (t 1 ). Therefore, the t i (t 1 ) corresponding to the maximum value of E 1 is the starting point of the vehicle getting on the bridge, that is, the starting point of the effective signal ΔT, and the expression is:

[0031]

[0032] As a preferred solution of the three-dimensional spatio-temporal diagram construction and diagnosis method for the fracture fault of a truss bridge according to the present invention, wherein: the feature extraction and correlation calculation of the preprocessed data include,

[0033] Using the standard deviation of cross-edge weights as a quantization index for signal correlation between nodes, the index combines signal waveform features and energy features to evaluate the signal transmission between adjacent nodes;

[0034] Combining the standard deviation of cross-edge weights with time-frequency domain features to extract the signal correlation between nodes and establish the fusion features of multi-dimensional data;

[0035] When two nodes P A and P B are in the same time period, the standard deviation of cross-edge weights between the two nodes is defined as the standard deviation of the edge weights of the signals on the corresponding t a-1 , t a and t a+1 time windows when the vehicle passes through the current node and adjacent nodes, and the expression is:

[0036]

[0037] wherein, and are respectively the upper envelope and the lower envelope of the signals collected by the sensors installed on P A ; are respectively the upper envelope and the lower envelope of the signals collected by the sensors installed on P B ;

[0038] The standard deviation coefficient of cross-edge weights between nodes P A and P B in the same time interval, and the expression is:

[0039]

[0040] ρ + σ + τ = 1, σ > 0.5

[0041] wherein, ρ, σ, and τ are weight coefficients;

[0042] When two nodes P A and P B are in adjacent time periods, the standard deviation of cross-edge weights between the two nodes is defined as the standard deviation of the edge weights of the signals on the corresponding t a-1 , t a and t a+1 time windows when the vehicle passes through the current node and adjacent nodes, and the expression is:

[0043]

[0044] Nodes P in adjacent time intervals A and P B The standard deviation coefficient of the cross-edge weight between them is expressed as:

[0045]

[0046] ρ + σ + τ = 1, σ > 0.5

[0047] where ρ, σ, and τ are weight coefficients.

[0048] As a preferred solution of the three-dimensional spatio-temporal graph construction and diagnosis method for the fracture fault of a truss bridge according to the present invention, wherein: updating the distance coefficient between nodes according to the correlation calculation result, and generating a three-dimensional spatio-temporal graph based on the point set, edge set, and neighbor set in the three-dimensional graph structure includes

[0049] Using the root mean square, Pearson correlation coefficient, and phase correlation for weighted fusion to generate a comprehensive feature coefficient Φ to quantify the association strength between node pairs, and the expression is:

[0050]

[0051] where Q A-B is the feature matrix, n is the number of repeated experiments, and the coefficient matrix C Τ ∈R 1*5 ;

[0052] Updating the distance coefficient between nodes according to the cross-edge weight standard deviation calculation result, and generating a three-dimensional spatio-temporal graph based on the point set, edge set, and neighbor set in the three-dimensional graph structure to reflect the dynamic relationship between the nodes of the truss bridge in the time and space dimensions, and the expression is:

[0053]

[0054] where E (3) and Φ d are the initial edge set and the distance coefficient matrix respectively, is the updated edge set;

[0055] Identifying and warning of potential fracture risks of the truss bridge through the update of node coordinates and the change of distances between nodes.

[0056] As a preferred solution of the three-dimensional spatio-temporal graph construction and diagnosis method for the fracture fault of a truss bridge according to the present invention, wherein: establishing a three-dimensional physical model of the bridge, combining the collected spatio-temporal signals with the three-dimensional structure, and constructing the multi-dimensional fusion features between nodes includes

[0057] A three-dimensional physical model of a bridge is established using an experimental device. The collected spatio-temporal signals are combined with the three-dimensional structure to construct multi-dimensional fusion features between nodes. For repeated experiments under different experimental parameters, data dimensionality reduction and multi-dimensional data fusion are performed to generate correlation features between nodes.

[0058] According to the changes in weights and distances between nodes, the three-dimensional spatio-temporal graph model is updated in real time.

[0059] As a preferred solution of a method for constructing and diagnosing a three-dimensional spatio-temporal graph of a truss bridge fracture fault according to the present invention, wherein: verifying the reliability and accuracy of the algorithm in different situations by simulating truss bridges in different health states includes,

[0060] Verifying the reliability and accuracy of the algorithm in different situations by simulating truss bridges in different health states in a laboratory environment;

[0061] The three-dimensional spatio-temporal graph signal reflects the dynamic changes and structural damage between nodes;

[0062] Using the node correlation features fused by the standard deviation of cross-edge weights and time-frequency features, a graph signal model between nodes is established, and the experimental data is corrected and optimized.

[0063] Another object of the present invention is to provide a system for constructing and diagnosing a three-dimensional spatio-temporal graph of a truss bridge fracture fault, which can accurately generate a three-dimensional spatio-temporal graph model reflecting the health state of the bridge by collecting and preprocessing vibration signals in real time and combining the dynamic analysis of the three-dimensional graph structure of the truss nodes. Through model verification and multi-dimensional fusion feature extraction, the fracture fault of the truss bridge can be effectively diagnosed, solving the problems of insufficient dynamic response, incomplete data analysis, and untimely fault warning in the existing system for large-scale bridge monitoring.

[0064] To solve the above technical problems, the present invention provides the following technical solution: A system for constructing and diagnosing a three-dimensional spatio-temporal graph of a truss bridge fracture fault, including: a data collection and preprocessing module, a feature extraction and correlation analysis module, a three-dimensional spatio-temporal graph generation module, and a model verification and fault diagnosis module;

[0065] The data collection and preprocessing module collects vibration signals by arranging multi-channel acceleration sensors on bridge nodes and performs preprocessing on the original data through Hilbert transform and signal envelope calculation;

[0066] The feature extraction and correlation analysis module extracts features from the preprocessed data and performs signal correlation calculation to quantify the signal transmission characteristics between adjacent nodes;

[0067] The three-dimensional spatio-temporal graph generation module updates the distance coefficient of the nodes and generates a three-dimensional spatio-temporal graph according to the result of the node correlation analysis;

[0068] The model verification and fault diagnosis module constructs a three-dimensional physical model of the bridge, combines the spatio-temporal signals with the three-dimensional structure, and simulates the truss bridges in different health states for verification.

[0069] A computer device includes a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, the steps of the above-mentioned method for constructing and diagnosing a three-dimensional spatio-temporal diagram of a truss bridge fracture fault are implemented.

[0070] A computer-readable storage medium stores a computer program thereon. When the computer program is executed by a processor, the steps of the above-mentioned method for constructing and diagnosing a three-dimensional spatio-temporal diagram of a truss bridge fracture fault are implemented.

[0071] Advantages of the present invention: By introducing a truss bridge fracture diagnosis algorithm based on three-dimensional spatio-temporal diagram signals, combining the cross-edge weight standard deviation (C-EWSD) feature and multi-dimensional data fusion technology, the present invention successfully solves the problems of inaccurate analysis of the correlation between nodes and untimely fault diagnosis in existing structural health monitoring, achieving the effects of reflecting the health state of the truss bridge in real time and efficiently, accurately locating the structural damage position, and realizing early risk warning. Description of the Drawings

[0072] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts. Among them:

[0073] Figure 1 It is the overall flowchart of a method for constructing and diagnosing a three-dimensional spatio-temporal diagram of a truss bridge fracture fault provided by the first embodiment of the present invention;

[0074] Figure 2 It is the truss bridge experimental device in a method for constructing and diagnosing a three-dimensional spatio-temporal diagram of a truss bridge fracture fault provided by the first embodiment of the present invention, where (a) is the truss bridge model; (b) is the signal acquisition system; (c) is the faulty truss;

[0075] Figure 3 It is the truss fault position and location in a method for constructing and diagnosing a three-dimensional spatio-temporal diagram of a truss bridge fracture fault provided by the first embodiment of the present invention;

[0076] Figure 4 It is the flowchart of the algorithm in a method for constructing and diagnosing a three-dimensional spatio-temporal diagram of a truss bridge fracture fault provided by the first embodiment of the present invention;

[0077] Figure 5 The basis for 3D modeling of the bridge and signal segmentation in a method for constructing and diagnosing a 3D spatio-temporal diagram of a truss bridge fracture fault provided by the first embodiment of the present invention;

[0078] Figure 6 The calculation flow chart of the C-EWSD algorithm in a method for constructing and diagnosing a 3D spatio-temporal diagram of a truss bridge fracture fault provided by the first embodiment of the present invention;

[0079] Figure 7 The 8-channel signal of REC0473 in a method for constructing and diagnosing a 3D spatio-temporal diagram of a truss bridge fracture fault provided by the first embodiment of the present invention;

[0080] Figure 8 The 3D spatio-temporal diagram under normal conditions in a method for constructing and diagnosing a 3D spatio-temporal diagram of a truss bridge fracture fault provided by the first embodiment of the present invention;

[0081] Figure 9 The 3D spatio-temporal diagram in the fracture state of the P8-P14 main truss in a method for constructing and diagnosing a 3D spatio-temporal diagram of a truss bridge fracture fault provided by the first embodiment of the present invention;

[0082] Figure 10 The 3D spatio-temporal diagram in the semi-fracture state of the P8-P14 main truss in a method for constructing and diagnosing a 3D spatio-temporal diagram of a truss bridge fracture fault provided by the first embodiment of the present invention;

[0083] Figure 11 The 3D spatio-temporal diagram of semi-welding of the P7-P13 truss in a method for constructing and diagnosing a 3D spatio-temporal diagram of a truss bridge fracture fault provided by the first embodiment of the present invention; Figure 12 The offset of the nodes on the diagram and the multiple diagram of the edges between nodes in the fracture state of the P8-P14 truss in a method for constructing and diagnosing a 3D spatio-temporal diagram of a truss bridge fracture fault provided by the first embodiment of the present invention; Figure 13 The offset of the nodes on the diagram and the multiple diagram of the edges between nodes in the semi-fracture state of the P8-P14 main truss in a method for constructing and diagnosing a 3D spatio-temporal diagram of a truss bridge fracture fault provided by the first embodiment of the present invention. Detailed implementation manners

[0084] To make the above objects, features, and advantages of the present invention more obvious and understandable, the following provides a detailed description of the specific implementation manners of the present invention with reference to the accompanying drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of them. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0085] Example 1, refer to Figures 1 to 13 This is an embodiment of the present invention, which provides a method for constructing and diagnosing a three-dimensional spatio-temporal diagram of a truss bridge fracture fault, including:

[0086] The experimental device of the present invention refers to Figure 2 Simulate the operation system of a two-span continuous through highway simply supported steel truss bridge. The experimental device consists of three reinforced concrete piers, two steel truss bridges, a vehicle start-stop buffer section, and several load trolleys, and can simulate a complete vehicle-bridge operation system. In the experiment, the vehicle starts from the start-stop buffer section, accelerates to a constant speed, autonomously passes through the bridge at a constant speed, and then decelerates and stops at a designated position, and the signal of the vehicle passing through the bridge deck can be collected. The vehicle has the ability of autonomous operation, which can ensure the simulation of the real vehicle passing situation during the experiment.

[0087] The experimental device of the present invention includes a total of three sections of bridges, two of which are connected by bolts and one is connected by rivets. During the experiment, different faults in the bridge experiment process are simulated by replacing the faulty parts. The experimental bridge in this experiment is a steel truss bridge, which is assembled by I-beams and square steels through connectors. The weight of a single section of the bridge is 60 kg, the length of the bridge deck is 1.73 m, the width of the bridge deck is 0.5 m, the height of the main truss is 0.25 m, the bridge deck is arranged with 5 longitudinal beams, a total of 6 bays, the bridge deck is provided with 2 traffic lanes, and the bridge deck is paved with wood board + rubber surface to simulate the concrete + asphalt bridge deck. The piers are three reinforced concrete column piers, each weighing 60 kg and with a pier height of 0.6 m.

[0088] The vehicle system of the experimental device is a self-designed heavy-duty rear-wheel drive electric flat trolley, which can carry a maximum load of 30 kg. The vehicle has a complete power resource, transmission system, power supply system and control system. The vehicle is driven by a stepping motor and powered by a 24V DC battery, and has the ability of autonomous operation. It can preset the program to control acceleration, deceleration and running time, and can complete vehicle acceleration and deceleration tests, same-direction driving, opposite-direction driving, and passing through the bridge deck simulation with different loads and different speeds on the bridge deck.

[0089] The data acquisition system is the American Crystal Diamond CoCo 80X data acquisition / vibration analyzer and CA-YD186 acceleration sensor. The acquisition device can collect a maximum of 8-channel vibration data, and can achieve the acquisition of up to 8-node vibration data in this experiment. The acceleration sensor can collect the frequency range of 0.5 - 5000 Hz, and the sensitivity reaches 10 mV / m*s-2. Through this data acquisition device, the vibration information of different nodes of the bridge during the whole process of the vehicle passing through the bridge deck can be collected.

[0090] In the present invention, node markings are made on the bridge, such as Figure 3Four health states are set as shown (normal, fracture of the main truss from 8 - 14, fracture of the main truss from 7 - 13, semi - fracture of the main truss from 8 - 14). To verify the above three hypotheses, we set up different experiments with 3 different load weights and 4 speeds under the three health states. Since there are 29 bridge - marked nodes and the data acquisition supports a maximum of 8 channels, in order to analyze the data correlation relationship between nodes, we designed a variety of node combinations, and each combination was repeated 5 times under different speeds and loads. The experimental grouping and parameter design are shown in Table 1. This extensive experimental design ensures a comprehensive analysis of the health state of the bridge under various conditions and provides a reliable data set for verifying the proposed diagnostic algorithm.

[0091] Table 1: Experimental Grouping and Parameter Settings

[0092]

[0093]

[0094] The present invention proposes a spatio - temporal graph construction algorithm based on multi - channel vibration signals to establish the spatio - temporal relationship between bridge nodes. The algorithm flow is as Figure 4 shown. This algorithm establishes the correlation between nodes based on the proposed C - EWSD index combined with time - frequency domain metrics, and updates the distance between nodes on the graph according to the node spatial relationship and correlation coefficient to establish a spatio - temporal graph signal that can provide damage warnings. Specifically, this algorithm fuses the bridge spatial nodes and the vibration signal time series of the vehicle crossing the bridge. First, the bridge spatial relationship is established through 3D modeling and a complete experimental device is built. Then, in the vehicle - bridge joint experiment, the vibration signals of multiple nodes are collected simultaneously to establish the spatio - temporal relationship: in the time dimension, the time series signals of each node when the vehicle crosses the bridge are collected; in the spatial dimension, the vibration signals of each node at the same time point or time interval are obtained. The spatial dimension and the time dimension can enhance the correlation with each other, so as to more accurately establish the correlation between nodes. This is the fusion correlation of time - series vibration signals based on the spatial network. We collect the continuous signals of the vehicles passing through the nodes equipped with sensors on the bridge. Through signal interception, we obtain the data of the vehicles passing through different nodes at different time intervals and establish a strong correlation between nodes from the spatio - temporal perspective.

[0095] In the present invention, a total of 29 nodes of the truss bridge are marked and sensors are installed, as Figure 5 (a) shown. The start and stop of data acquisition are controlled manually, so there is an empty acquisition time at the start and end of the detection signal (i.e., the time difference between the operation of the data acquisition instrument and the running of the vehicle). The original data is aligned and segmented to obtain the data collected by different nodes at different time intervals. As Figure 5(As shown in (b), the bridge truss of the present invention is divided into 7 segments according to the node distribution, and each segment corresponds to different time and space nodes. Taking a set of normal signals S (data record number REC0473) as an example, it contains 8-channel signals with a length of L. After repeated experiments, the running time T of the vehicle when the speed v = 0.1 m / s is obtained.) 2 = 20.8 s. The present invention designs a zero acquisition time calculation algorithm to intercept the effective signals. By identifying the maximum energy difference in the signals when the vehicle drives onto the bridge, the optimal signal interval can be determined. It can remove the signals corresponding to the zero acquisition time, obtain accurate operation signals, and align the starting points of different data sets.)

[0096] (The energy of the signals collected by the sensors changes greatly near the starting point and the ending point of the effective acquisition interval. The empty acquisition time of the starting point is calculated by using the moving average maximum energy difference. Because each section of the bridge is not rigidly connected to its adjacent road surface, when the front wheel of the vehicle drives onto the bridge, the signal energy increases instantaneously, and the energy will decay after the rear wheel of the vehicle passes the bridge. The most direct manifestation of energy on the signal is the peak height of the waveform. First, calculate the upper and lower peak envelopes of the signal.)

[0097] (1) Perform Hilbert transform on the 8-channel signals of signal S. In order to obtain the upper and lower signals respectively, the vibration signals are divided into positive vibration signals and negative vibration signals according to positive and negative values:)

[0098]

[0099] (where i is the channel number of signal S,) and are the upper and lower envelope signals of the vibration signal respectively.)

[0100] and (The Hilbert transforms of

[0101]

[0102] (2) Construct the analytic signal of the original signal S, and this analytic signal is a complex signal:)

[0103]

[0104] (3) Calculate the envelope: The envelope signal is the modulus of the analytic signal, and the lower envelope signal is the negative value of the modulus.)

[0105]

[0106] (where,) and are the upper and lower peak envelope signals of signal S i respectively.)

[0107] Due to vehicle movement, the signal energy value is relatively large after the vehicle starts and before it stops. Therefore, the algorithm proposed in the present invention extracts signals from the starting point using two windows with a width of ΔL, and the distance between the two windows is T 2 -ΔL, and then calculates the difference between the upper and lower envelope peaks of the signals in the two windows and E i (t 1 ):

[0108]

[0109] where t 1 is the starting point of the front moving window, and fs is the sampling frequency.

[0110] When t 1 is the moment when the front wheels of the vehicle drive onto the bridge, E i (t 1 ) reaches its maximum value. When the signal in the moving window contains the signal of the empty sampling time, the energy value of the signal is less than the maximum value of E i (t 1 ). Therefore, the t i (t 1 ) corresponding to the maximum value of E 1 is the starting point of the vehicle driving onto the bridge, that is, the starting point of the effective signal ΔT:

[0111]

[0112] After removing the signal corresponding to the empty sampling time from the signal S, the effective signal S' can be obtained, and the expression is:

[0113] S' = {S(t)|ΔT ≤ t ≤ ΔT + T 2} (7)

[0114] The distance between the vehicle and the sensor will affect the signal. To reduce the dimension, we discard the signals when the vehicle is far from the sensor. As Figure 3 shown, only the signal segments collected when the vehicle passes through this node and adjacent nodes are retained. When the vehicle passes through a certain node, the signal segment collected by the sensor on this node is defined as the main signal, and the signal segments collected by the sensors arranged on adjacent nodes are defined as the secondary signals of this node.

[0115] To extract the spatio-temporal correlation between nodes, the present invention creatively proposes a cross-edge weight standard deviation (C-EWSD) feature extraction algorithm. In the present invention, the vehicle is used as a moving excitation source, the excitation points are the four wheels of the vehicle, and the generated signal features are correlated with the movement path. The cross-edge weight standard deviation between nodes is a feature based on signal waveforms and energy. The node connection method of the truss bridge has a high degree of similarity and there is a physical connection between adjacent nodes, thus having transitivity.

[0116] Therefore, we make two assumptions:

[0117] a. When the vehicle passes through a fixed distance before and after each node, the signal characteristics collected by the sensors arranged at that node are similar;

[0118] b. The signals collected by the sensors on adjacent nodes are highly similar due to the transitivity of physical connection.

[0119] Based on data dimensionality reduction, the signal collected by the vehicle passing through the sensors deployed at the node is defined as the main signal, and the signal segment when the vehicle passes through and collects the adjacent nodes of the sensor is defined as the secondary signal of that node. According to the spatial relationship of the bridge, we define the nodes with spatial connection as associated node pairs, that is, there is an adjacent relationship between these two nodes, such as node 7 and 8, node 8 and 14. We can obtain that all node pairs are in the same time period or adjacent time periods in the time dimension. Therefore, in the present invention, when calculating the association relationship between node P A and P B there are two cases, that is, the two nodes are in the same time period and adjacent time periods. When calculating the node correlation, we intercept the signal according to the specified vehicle driving distance, and use the time required for the vehicle to drive a certain distance before and after passing through the sensor at a fixed speed to intercept the signal in a window.

[0120] (1) Nodes P A and P B are in the same time period

[0121] CH a and CH b are the complete signals collected by the sensors deployed on nodes P A and P B respectively, as shown by the C-EWSD of ch5 and ch6 in Figure 6 . The time interval corresponding to nodes P A and P B is t a . There are a total of 6 data segments for calculating the relationship between node pairs P 8 and P 14 , as shown by the C-EWSD of CH Figure 6 and CH 2 and CH 6 in A When nodes P B and P A and P B are in the same time interval, the C-EWSD between P A and P B is defined as the standard deviation of the edge weight standard deviation (EWSD) of the two-channel signals in the same time interval when the vehicle passes through these two nodes and their adjacent nodes, that is, ta-1 , t a and t a+1 EWSD of the signal over the time interval:

[0122]

[0123] Wherein, and are respectively the upper envelope and the lower envelope of the signal collected by the sensors installed on P A .

[0124] At the nodes P A and P B in the same time interval, the C-EWSD coefficient is defined as:

[0125]

[0126] Where ρ + σ + τ = 1, σ > 0.5, and these three parameters are dynamically adjusted according to the training results.

[0127] (2) Nodes P A and P B are located in adjacent time periods

[0128] When P A and P B are in adjacent time intervals: t a and t b , as shown in the C-EWSD of CH Figure 3 in 5 and CH 6 , there are 4 signal segments in both of the two node channels used to calculate the C-EWSD corresponding to CH 5 and CH 6 .

[0129]

[0130] At the nodes P A and P B in the adjacent time interval, the C-EWSD coefficient is defined as:

[0131]

[0132] Where ρ + σ + τ = 1, σ > 0.5, and these three parameters are dynamically adjusted according to the training results.

[0133] To establish the graph signal relationship of nodes more accurately, we incorporate the time-frequency domain features of signals based on C-ewsd to establish the fusion features of node signals. During the experiment, we conducted repeated experiments and comparative experiments to ensure accuracy. Therefore, it is necessary to perform data dimensionality reduction on the repeated experiment structure and perform multi-dimensional fusion on the same node pair features in the comparative experiment results. The main calculation process is as follows:

[0134] (1) Select C-SDEW and time-frequency features to calculate the correlation weight coefficient between nodes:

[0135] The present invention designs a self-weighted analysis algorithm to measure the correlation information between nodes, selects 5 sensitive features from 13 time-domain indicators, 11 frequency-domain indicators, and 15 statistical-domain indicators, and constructs a feature matrix:

[0136] Q A-B =[W A-B ,W rms ,W per ,W pha ,W psd ,W cos (12)

[0137] Among them, W A-B is the proposed C-EWSD feature, and the five sensitive features are root mean square W rms , Pearson correlation coefficient W per , phase correlation W pha , power spectral density W psd and spectral cosine similarity W cos . Among them, root mean square W rms and power spectral density W psd are inverse indicators, and the other three are positive indicators.

[0138] (2) Feature fusion and multi-dimensional data fusion

[0139] In the experiment of the present invention, 4 kinds of vehicle driving speeds are set: v 1 =400, v 2 =500, v 3 =800, v 4 =900. Due to the differences in experimental parameters, the original feature parameters of the calculated signals vary greatly. First, perform data dimensionality reduction on the repeated experiments under the same experimental conditions to reduce the computational amount. Calculate the fusion feature weight coefficient Φ between nodes according to the feature matrix Q A-B :

[0140]

[0141] Among them, the coefficient matrix C Τ ∈R 1*5 .

[0142] To more accurately describe the fault, the measurement index adopted in the present invention is a comparison index based on energy characteristics, so the characteristics of different parameters can be fused:

[0143] Φ A-B = max(Φ A-B,400 , Φ A-B,500 , Φ A-B,800 , Φ A-B,900 )(14)

[0144]

[0145] where m is the bridge node number.

[0146] (3) Regularization

[0147] Convert the weight into the distance coefficient Φ between nodes d and normalize it:

[0148]

[0149] Combine the signal characteristics obtained under four working conditions into a new feature matrix, compare the signal characteristics under different health states with the healthy state, establish the correlation coefficient between nodes according to the comparison results of the fused features, establish the original graph signal with the C-EWSD feature and the three-dimensional structure, and update the graph signals under different health states according to the comparison results.

[0150] Graph learning expands the traditional single-node analysis into an intelligent diagnosis system for multi-source information fusion. Relying on the "node-edge-neighbor" system to establish a graph signal model, it can be used to establish an intelligent diagnosis system for system-level structural health monitoring. In the "node-edge-neighbor" relationship, "node" represents the sample object, "edge" represents the correlation relationship index between nodes, and "neighbor" represents the relationship of whether nodes are related. Update the node coordinates of the graph according to the weight information and the original three-dimensional coordinates, and the defect location can be effectively judged through the node offset.

[0151] In this study, the definition of a graph is:

[0152]

[0153] where V ∈ R m×4 is the point set, including the node set with m nodes and the three-dimensional coordinates; the diagonal matrix E ∈ R m×m is the edge set, including the association and distance information; A ∈ R m×m is the neighbor set, including the neighborhood information of each node.

[0154] The correlation between C-EWSD feature calculation nodes proposed by the present invention can effectively establish two-dimensional spatio-temporal graph signals for 8 nodes by analyzing 8-channel time signals and three-dimensional spatial structures, and finally obtain the overall node correlation of the entire truss structure through data fusion to establish three-dimensional spatio-temporal graph signals. The main calculation steps are as follows:

[0155] According to the definition of the graph, first establish a point set (used to describe coordinates), an edge set (used to describe the correlation coefficient between nodes), and a neighbor set (used to describe the existence of correlation and spatial distance) of the signal based on the three-dimensional structure. In the three-dimensional graph structure, the point set is variable and is updated by the combined calculation results of the edge set and the neighbor set. The neighbor set is determined by the three-dimensional spatial structure and is fixed. The edge set is obtained through signal analysis calculation.

[0156] (1) Establishment of two-dimensional spatio-temporal graph signals

[0157] According to the 8-channel signals collected in each group of experiments, establish spatio-temporal graph signals between 8 nodes. First, give the definition of the two-dimensional graph signal:

[0158]

[0159] where V (2) ∈R 8×4 ,E (2) ∈R 8×8 ,A (2) ∈R 8×8 . Taking the data set REC0473 as an example, the initial point set V (2) is:

[0160]

[0161] Based on the initial point set V (2) calculate the initial edge set E (2)

[0162]

[0163] where U (2) ∈R 4×8 is the incidence matrix:

[0164]

[0165] Based on the distance coefficient Φ d in 3.3, update to obtain the edge set

[0166]

[0167] Update the edge set It contains the node distance information updated according to the vibration signal, and the graph signal is obtained by updating the node coordinates based on the node distance. The present invention designs an 8-node update algorithm based on a three-dimensional model, and obtains the updated adjacent set through the adjacent set A and the updated edge set Obtain the updated adjacent set The algorithm process is as follows:

[0168]

[0169]

[0170] Finally, a two-dimensional spatio-temporal graph of 8 nodes can be drawn according to the updated nodes, and the offset of the nodes on the graph can intuitively show the change of the relationship between the nodes.

[0171] (2) Establishment of three-dimensional spatio-temporal graph signal

[0172] Based on the signals collected in the experiment of different 8 nodes in (1), after signal dimensionality reduction and fusion, the feature set of 29 nodes marked on the entire truss bridge is obtained, and the edge set of the graph signal is updated through this feature set. First, the definition of the three-dimensional graph signal is given:

[0173]

[0174] Among them, V (3) ∈R 29×4 , E (3) ∈R 29×29 , A (3) ∈R 29×29 .

[0175] Based on the initial point set V (3) Calculate the initial edge set E (3) :

[0176]

[0177] Among them, U (3) ∈R 4×29 is the incidence matrix:

[0178]

[0179] Based on the distance coefficient Φ in 3.3 d Update to obtain the edge set

[0180]

[0181] Combine the edge set and the adjacent set A (3) For the point set Update. Based on Algorithm 1, a three-dimensional graph signal coordinate update algorithm based on the physical model of the bridge is designed:

[0182]

[0183]

[0184] Finally, the updated three-dimensional graph signal coordinates are obtained This algorithm can amplify the structural damage state into the graph signal based on the vibration signal.

[0185] The purpose of bridge structural health monitoring is to provide early warnings or judgments for points or edges that may fail currently or in the future. The judgment of failure is reflected as a significant deviation of points or edges from the normal situation in the graph signal.

[0186] Bridge health monitoring can be divided into the following two situations from the perspectives of points and edges:

[0187] (1) Edge change

[0188] The calculated distance coefficient directly reflects the change in edge length. The change of the edge on the graph corresponds to the change in truss energy transfer. By observing the change in edge length, the change in the health state of the truss between two nodes can be effectively determined.

[0189] (2) Point displacement

[0190] When the starting point of the graph signal is fixed, the vibration of the node will propagate along the truss structure. The health status of the node can be determined according to its displacement from the original position in the graph signal. A significant deviation in the coordinates of the node on the graph indicates a change in the structure of the node or the edge connected to it, thus changing the vibration transmission mode. In a two-dimensional spatio-temporal graph, the number of nodes is relatively small, and the fixed drawing point cannot be associated with the actual fixed point of the bridge. While a three-dimensional spatio-temporal graph can consider all nodes of the bridge according to the physical model of the bridge, draw the graph signal starting from the fixed end, accurately reflect the node displacement, and set the color mapping of the displacement on the graph to reflect the health state of the structure.

[0191] To verify the scientificity of the algorithm proposed in the present invention, we set up four working condition experiments in Table 1 on the truss bridge test bench. Taking the healthy state as the standard, we judged whether there was a faulty truss through comparative experiments of the other three healthy states.

[0192] (1) Healthy state

[0193] In the present invention, 162 repeated experiments were carried out in the healthy state. Among them, a group of 8-node (channel) signals were collected in each experiment, and the 8-channel signals obtained in each experiment (such as REC0473) were as Figure 7As shown. Each use can obtain the correlation coefficients between a group of 8 nodes. After data dimensionality reduction, a C-EWSD feature matrix between 44 node pairs at 4 different speeds is finally obtained, as shown in Table 2.

[0194] Table 2 C-EWSD eigenvalue of node pairs at four speeds

[0195]

[0196]

[0197] Fuse the C-EWSD at 4 speeds with time-frequency features to obtain the correlation coefficients between each node pair. According to the above relationship, draw the graph signal of the entire truss bridge as Figure 8 shown. It can be seen from the graph signal established based on the three-dimensional model of the bridge and the vibration signal that the distance between nodes in the graph relationship is highly correlated with the actual spatial distance, indicating that the overall structure is stable and consistent with the normal state of the bridge.

[0198] On the test bench, replace the P8-P14 truss with a faulty truss and conduct 330 experiments under four different operating parameters. Compare the eigenvalue of 20 groups of the same 8-channel signals in the two healthy states of the healthy state and the P8-P14 truss failure. The sensor installation positions are all (P7, P8, P9, P12, P13, P14, P15, P16). There are 10 node pairs with physical connections among the 8 nodes. The feature comparison results between node pairs after fusion at four speeds show that there are significant differences between the two healthy states of the P8-P14 truss. However, it can be seen from the feature comparison results at different speeds that the faulty positions when v = 500 are different from those at the other three speeds. Therefore, there may be deviations in structural health monitoring using two-dimensional spatio-temporal signal graphs.

[0199] Perform multi-dimensional data fusion through repeated experiments with different node combinations. The repeated node pairs are classified and dimensionally reduced according to different speeds. Finally, 46 node pairs are obtained for drawing a three-dimensional spatio-temporal graph. Table 3 shows the 46 node pairs and their corresponding C-EWSD indicators.

[0200] Table 3 C-EWSD eigenvalue of node pairs at four speeds

[0201]

[0202]

[0203] Compare the node pair features in this state with the node signal features in the healthy state to obtain the change magnification of the edges between associated nodes on the graph. Use the node coordinate update algorithm to obtain the node coordinates in the case of the fracture of the P8 - P14 truss. According to the node coordinates, the offset of the node from its original position can be obtained, and based on the node coordinates on the graph, the 3D spatio - temporal graph of the truss bridge can be obtained. Figure 9 And perform color mapping according to the node offset on the graph.

[0204] It can be seen from the spatio - temporal graph signal that the offset of node 14 is relatively large. It can be speculated that the truss related to node 14 has failed, resulting in a decrease in the correlation between node 14 and other nodes. From the node offset and side - length magnification Figure 12 it can be seen that the change in the correlation between node 14 and node 8 is the largest. Based on this, it is inferred that the P8 - P14 truss has failed.

[0205] (3) Semi - fracture experiment of the P8 - P14 main truss

[0206] Replace the 8 - 14 truss with a semi - fractured faulty truss on the test bench and conduct 330 experiments under four different operating parameters. For the 8 - channel signals obtained in each experiment, compare the eigenvalue of 20 groups of the same 8 - channel signals in the healthy state and the semi - fractured state of the P8 - P14 truss. The sensor installation positions are the same as those in the main truss fracture experiment of P8 - P14. There are a total of 10 node pairs with physical connections among the 8 nodes. The comparison results of the node pair features after fusing the four speeds show that there are significant differences in the two healthy states of the P8 - P14 truss. Therefore, using the two - dimensional spatio - temporal signal graph for structural health monitoring may produce deviations.

[0207] Perform multi - dimensional data fusion through repeated experiments with different node combinations, and reduce the dimension of the repeated node pairs according to different speeds. Finally, 46 node pairs are obtained for drawing the three - dimensional spatio - temporal graph. Table 4 lists the 46 node pairs and their corresponding C - EWSD indicators.

[0208] Table 4 C - EWSD values between node pairs at four speeds

[0209]

[0210]

[0211] Compare the node pair features in the semi - fractured state with the node signal features in the healthy state to obtain the change magnification of the edges between associated nodes on the graph. Use the node coordinate update algorithm to obtain the node coordinates in the semi - fractured state of the P8 - P14 truss. According to the node coordinates on the graph, the offset of the node from its original position can be obtained, as Figure 10 shown. According to the node coordinates, the 3D spatio - temporal graph of the truss bridge can be drawn, and color mapping is performed according to the node offset.

[0212] It can be seen from the spatio-temporal diagram signal that the offset of Node 14 is relatively large. It is speculated that the truss related to Node 14 has failed, resulting in a decrease in the correlation between Node 14 and other nodes. From Figure 13 the changes in the node offset and side length magnification, it can be seen that the correlation change between Node 14 and Node 8 is the largest. Therefore, it is inferred that the P8 - P14 truss has suffered a semi-fracture failure.

[0213] (4) Fracture of the P7 - P13 main truss

[0214] In this experiment, the P7 - P13 truss was replaced with a completely fractured faulty truss on the test bench, and a total of 108 experiments were carried out under four different working conditions. The characteristic values of 20 signals of the same 8 channels were compared in two healthy states: healthy and P7 - P13 truss fractured. The sensor installation positions were set as [P6, P7, P8, P11, P12, P13, P14, P15]. These positions formed 10 physically connected node pairs. After aggregating the C - EWSD characteristics at four different speeds, it was found that there were significant differences between the two healthy states of the truss connected to P13. However, it can be seen that there are obvious differences in the results at four speeds. This indicates that under certain conditions, using a two-dimensional spatio-temporal signal diagram for structural health monitoring may introduce biases.

[0215] Multiple experimental iterations with different node combinations and different speeds were carried out to classify and reduce the dimensions of the repeated node pairs. This multi-dimensional data integration finally determined 44 node pairs suitable for drawing a three-dimensional spatio-temporal diagram, as shown in Table 5, which lists the C - EWSD indicators corresponding to each node pair.

[0216] Table 5 C - EWSD values between node pairs at four speeds

[0217]

[0218]

[0219] By comparing the characteristics of node pairs in the faulty state with those in the healthy state, we obtained the edge change rate between the relevant nodes on the graph. The coordinates of the nodes in the fractured state of the P7 - P13 truss were calculated using the node coordinate update algorithm. Based on these coordinates, the displacement of the nodes relative to their original positions was determined. A three-dimensional spatio-temporal diagram of the truss bridge was generated according to the node coordinates, and color mapping was performed according to the node displacement values, as Figure 11 shown.

[0220] As can be seen from the spatio-temporal graph signal, obvious displacements occurred at nodes P6 and P13, especially the largest offset at node P13. This means that the truss associated with node P13 may have failed, thus reducing the connection strength between node P13 and other nodes. Analysis Figure 13 By analyzing the node displacements and side length change rates, it can be seen that the association change between node P13 and node P7 is the most obvious. Based on this observation, we can infer that the P7-P13 truss may have been damaged.

[0221] Embodiment 2, an embodiment of the present invention, provides a system for constructing and diagnosing a three-dimensional spatio-temporal graph of a truss bridge fracture fault, including: a data collection and preprocessing module, a feature extraction and correlation analysis module, a three-dimensional spatio-temporal graph generation module, and a model verification and fault diagnosis module;

[0222] The data collection and preprocessing module collects vibration signals by arranging multi-channel acceleration sensors on bridge nodes and performs preprocessing on the original data through Hilbert transform and signal envelope calculation;

[0223] The feature extraction and correlation analysis module extracts features from the preprocessed data and performs signal correlation calculation to quantify the signal transmission characteristics between adjacent nodes;

[0224] The three-dimensional spatio-temporal graph generation module updates the distance coefficient of nodes and generates a three-dimensional spatio-temporal graph according to the node correlation analysis results;

[0225] The model verification and fault diagnosis module constructs a three-dimensional physical model of the bridge, combines the spatio-temporal signal with the three-dimensional structure, and simulates truss bridges in different health states for verification.

[0226] If the above functions are implemented in the form of software function units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), magnetic disks, or optical discs that can store program codes.

[0227] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a definable sequence list of executable instructions for implementing logical functions, and can be specifically implemented in any computer-readable medium for use by an instruction execution system, apparatus, or device (such as a computer-based system, a system including a processor, or other systems that can fetch and execute instructions from the instruction execution system, apparatus, or device), or used in conjunction with these instruction execution systems, apparatuses, or devices. For the purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0228] More specific examples (non-exhaustive list) of computer-readable media include the following: an electrical connection part (electronic device) having one or more wirings, a portable computer disk cartridge (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disc read-only memory (CDROM). Additionally, a computer-readable medium can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other media, followed by editing, interpretation, or other suitable processing as necessary, and then stored in a computer memory.

[0229] It should be understood that various parts of the present invention can be implemented by hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, any one or a combination of the following techniques well known in the art can be used: discrete logic circuits having logic gate circuits for implementing logical functions on data signals, application-specific integrated circuits having suitable combinational logic gate circuits, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0230] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered by the scope of the claims of the present invention.

Claims

1. A three-dimensional spatiotemporal graph construction and diagnosis method for truss bridge fracture faults, characterized in that: include: Arrange sensors to collect truss bridge data and perform pre-processing; Perform feature extraction and correlation calculation on the preprocessed data; The distance coefficients between nodes are updated according to the association calculation results, and a three-dimensional space-time graph is generated based on the point set, edge set and neighbor set in the three-dimensional graph structure; Establish a three-dimensional physical model of the bridge, combine the collected spatiotemporal signals with the three-dimensional structure, and construct multi-dimensional fusion features between nodes; By simulating truss bridges in different health states, the reliability and accuracy of the algorithm in different situations are verified.

2. A 3D spatiotemporal graph construction and diagnosis method for truss bridge fracture faults according to claim 1, characterized in that: The method of arranging sensors to collect truss bridge data is to arrange multi-channel acceleration sensors on each node of the truss bridge to collect vibration signals.

3. A 3D spatiotemporal graph construction and diagnosis method for truss bridge fracture faults according to claim 2, characterized in that: The preprocessing is to perform Hilbert transform on the original signal, obtain the upper and lower peak envelopes of the signal, align the signal by sliding window method, intercept the effective signal and reduce the dimension of the signal, and retain the main signal of the key node and the secondary signal of the adjacent node; Perform Heilbert transform on several channel signals of signal S to obtain upper and lower signals respectively, and divide the vibration signal into positive vibration signal and negative vibration signal according to positive and negative values. The expression is: Where i is the channel number of signal S, and They are the upper and lower envelope signals of the vibration signal respectively; and The Heilbert transform is: in, They are and The Heilbert transform value at time t, PV stands for "principal value". The integral used in the Hilbert transform is a singular integral, which is solved using the Cauchy principal value method. It is the variable in the integral and represents the past or future time. Construct the analytical signal of the original signal, the expression is: in, are the replicated analytical signals of the upper and lower vibration signals, respectively, and j is an imaginary unit; The modulus value of the analytical signal is used as the envelope signal. The lower envelope signal is the negative value of the modulus. The expression is: in, They are the constructed upper envelope signal and lower envelope signal respectively; Use a sliding window to remove invalid empty sampling signals, determine the time when the vehicle goes on the bridge by identifying the change in signal energy, intercept the corresponding valid signal segment, use two windows with a width of ΔL to extract the signal from the starting point, the distance between the two windows is T2-ΔL, calculate the difference between the upper and lower envelope peaks of the two window signals and E i (t1), the expression is: Among them, t1 is the starting point of the front moving window, and fs is the sampling frequency; When t1 is the moment when the front wheel of the vehicle drives onto the bridge, E i (t1) maximum value, when the signal in the moving window contains the empty time signal, the energy value of the signal is less than E i (t1) is the maximum value, so E i When (t1) reaches the maximum value, the corresponding t1 is the starting point of the vehicle going on the bridge, that is, the effective signal starting point ΔT, which is expressed as:

4. A method for constructing and diagnosing a three-dimensional spatiotemporal graph of a truss bridge fracture fault according to claim 3, characterized in that: The feature extraction and association calculation of the preprocessed data includes: The standard deviation of cross-edge weights is used as a quantitative indicator of signal association between nodes. The indicator combines signal waveform characteristics and energy characteristics to evaluate the signal transmission between adjacent nodes. The cross-edge weight standard deviation is combined with the time-frequency domain features to extract the signal correlation between nodes and establish the fusion features of multi-dimensional data; When two nodes P A and P B When located in the same time period, the standard deviation of the cross-edge weights of two nodes is defined as the value of the cross-edge weights when the car passes through the current node and the adjacent node, corresponding to t a-1 , t a and t a+1 The standard deviation of the edge weight of the signal in the time window is expressed as: in, and They are installed in P A The upper envelope and the lower envelope of the signal collected by the sensor on the ground; They are installed in P B The upper envelope and the lower envelope of the signal collected by the sensor on the ground; Node P in the same time interval A and P B The standard deviation coefficient of the cross-edge weight between is expressed as: Among them, ρ, σ, τ are weight coefficients; When two nodes P A and P B When the time period is adjacent, the standard deviation of the cross-edge weights of the two nodes is defined as when the car passes through the current node and the adjacent node, that is, corresponding to t a-1 , t a and t a+1 The standard deviation of the edge weight of the signal segment in the time window is expressed as: Node P in adjacent time intervals A and P B The standard deviation coefficient of the cross-edge weight between is expressed as: Among them, ρ, σ, and τ are weight coefficients.

5. A method for constructing and diagnosing a three-dimensional spatiotemporal graph of a truss bridge fracture fault according to claim 4, characterized in that: The updating of the distance coefficients between nodes according to the association calculation results and the generation of a three-dimensional spatiotemporal graph based on the point set, edge set and neighbor set in the three-dimensional graph structure include: The root mean square, Pearson correlation coefficient, and phase correlation are used for weighted fusion to generate a comprehensive characteristic coefficient Φ to quantify the association strength between node pairs. The expression is: Among them, Q A-B is the feature matrix, n is the number of repeated experiments, and the coefficient matrix C Τ ∈R 1*5 ; The distance coefficient between nodes is updated according to the calculation results of the standard deviation of the cross-edge weights, and a three-dimensional space-time graph is generated based on the point set, edge set and neighbor set in the three-dimensional graph structure to reflect the dynamic relationship between the truss bridge nodes in time and space dimensions. The expression is: Among them, E (3) , Φ d are the initial edge set and distance coefficient matrix respectively, is the updated edge set; By updating the node coordinates and changing the distance between nodes, the potential fracture risk of the truss bridge can be identified and warned.

6. A method for constructing and diagnosing a three-dimensional spatiotemporal graph of a truss bridge fracture fault according to claim 5, characterized in that: The three-dimensional physical model of the bridge is established, the collected spatiotemporal signals are combined with the three-dimensional structure, and the multi-dimensional fusion features between nodes are constructed, including: The three-dimensional physical model of the bridge is established using the experimental device. The collected spatiotemporal signals are combined with the three-dimensional structure to construct multi-dimensional fusion features between nodes. For repeated experiments under different experimental parameters, data dimension reduction and multi-dimensional data fusion are performed to generate correlation features between nodes. The 3D spatiotemporal graph signal is updated in real time according to the changes in weights and distances between nodes.

7. A method for constructing and diagnosing a three-dimensional spatiotemporal graph of a truss bridge fracture fault according to claim 6, characterized in that: The reliability and accuracy of the algorithm under different conditions are verified by simulating truss bridges in different health states, including: The reliability and accuracy of the algorithm under different conditions were verified by simulating truss bridges in different health states in a laboratory environment; The 3D spatiotemporal graph signal reflects the dynamic changes and structural damage between nodes; The node association features fused with the cross-edge weight standard deviation and time-frequency features are used to establish a graph signal model between nodes, and the experimental data are corrected and optimized.

8. A system using the method for constructing and diagnosing a three-dimensional spatiotemporal graph of a truss bridge fracture fault according to any one of claims 1 to 7, characterized in that: It includes data collection and preprocessing module, feature extraction and correlation analysis module, 3D space-time graph generation module, model verification and fault diagnosis module; The data collection and preprocessing module collects vibration signals by arranging multi-channel acceleration sensors on bridge nodes, and performs Hilbert transform and signal envelope calculation preprocessing on the original data; The feature extraction and correlation analysis module extracts features from the preprocessed data and performs signal correlation calculation to quantify the transfer characteristics of signals between adjacent nodes; The three-dimensional space-time graph generation module updates the distance coefficient of the nodes and generates a three-dimensional space-time graph according to the result of the association analysis between the nodes; The model verification and fault diagnosis module constructs a three-dimensional physical model of the bridge, combines the spatiotemporal signal with the three-dimensional structure, and simulates truss bridges in different health states for verification.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of a method for constructing and diagnosing a three-dimensional spatiotemporal graph of a truss bridge fracture fault according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for constructing and diagnosing a three-dimensional spatiotemporal graph of a truss bridge fracture fault as described in any one of claims 1 to 7 are implemented.

Images