Structural digital twin method, equipment, storage medium and product

Abstract

The present invention discloses a structural digital twin method, device, storage medium and product, the method comprising obtaining a dynamic response signal of an actual structure, processing the dynamic response signal, and obtaining first response data; obtaining a vibration response video stream of the actual structure, and processing the vibration response video stream, and obtaining second response data; aligning and fusing the first response data with the second response data, and obtaining structural observation data; training and verifying a proxy model using a sample data set; inputting the structural observation data into a dynamic Bayesian network model, and obtaining a structural material parameter vector; inputting the structural material parameter vector into the proxy model, and obtaining predicted response data; judging whether the optimal goodness of fit is achieved according to the predicted response data and the structural observation response data, and realizing structural digital twin according to the optimal structural material parameter vector. The present invention not only improves the efficiency of structural monitoring, but also improves the monitoring accuracy.

Description

Technical Field

[0001] The present invention belongs to the field of structural engineering and structural health monitoring, and in particular, relates to a structural digital twin method, device, storage medium and product based on video images and sensors. Background Art

[0002] With the development of industrial technology and the growth of transportation demand, long-span bridges play an important role in infrastructure and urbanization construction. With the increase in the service life of bridges, the structural performance of in-service bridges has gradually deteriorated, posing serious hidden dangers to the normal use of bridge structures. How to effectively evaluate and predict the status and safety performance of bridges is a key issue currently facing bridge health monitoring and operation and maintenance. Exploring scientific and systematic methods for structural health monitoring, establishing a digital twin model of bridges, and realizing accurate evaluation of structural status can provide strong support for safe operation and maintenance of bridges.

[0003] Traditional structural health monitoring models mostly rely on contact sensors installed on the surface of the structure. Large-scale sensors introduce additional mass to the structure to a certain extent. At the same time, there are great difficulties in the deployment and maintenance of various monitoring equipment, and it is difficult for various monitoring equipment to be interconnected, facing many problems such as high monitoring costs and low detection efficiency. In the actual monitoring work, as the monitoring time increases, the multi-source data obtained is complex and continues to accumulate. The massive amount of monitoring data not only increases the cost of data transmission and storage, but also affects the operating efficiency of the monitoring system. In addition, it has become particularly difficult to extract effective information from low-quality and complex data for rapid and accurate assessment of the state of the bridge structure.

[0004] At present, the structural characteristic parameters and mechanical behavior responses are mainly obtained through on-site static or dynamic tests, and the objective function based on the model is established. The objective function is optimized by the difference between the test data and the model data, and it is continuously iterated to converge. For large and complex structures, the objective function optimization process based on finite element model correction needs to ensure that the finite element data corresponds to the measured data. However, the number of bridge structure measurement points and sensors is limited, which will lead to insufficient calculation accuracy when used for finite element model update and bridge digital twin modeling.

[0005] Most bridge digital twin constructions rely on contact sensors to establish physical model-based objective functions or neural network-based adaptive models. This method is costly and can only obtain data in a specific area. If the measurement points are sparsely arranged, the accuracy of the constructed digital twin will be insufficient. Relying only on sensors to obtain multi-source data, the data are independent of each other and difficult to calibrate, and the twin model in the digital space cannot be dynamically updated. In addition, due to the limitations of sensor accuracy and data loss, data discretization or insufficient key structural information may occur, key status information or damage information may be lost, and the health status of the bridge cannot be accurately assessed. Summary of the invention

[0006] The purpose of the present invention is to provide a structural digital twin method, device, storage medium and product to solve the problems that the traditional digital twin method relies on contact sensors, establishes objective functions based on physical models or adaptive models based on neural networks, has high costs, insufficient digital twin accuracy, and cannot accurately evaluate the health status of the structure.

[0007] The present invention solves the above technical problems through the following technical solutions: a structural digital twin method, comprising:

[0008] Acquire a dynamic response signal of the actual structure, and process the dynamic response signal to obtain first response data; wherein the dynamic response signal is acquired by a sensor arranged on the actual structure;

[0009] Acquire a vibration response video stream of the actual structure, and process the vibration response video stream to obtain second response data;

[0010] Aligning and fusing the first response data with the second response data to obtain structural observation data;

[0011] Constructing a sample data set; wherein each sample in the sample data set includes structural material parameters and corresponding response data;

[0012] Loading the proxy model, and using the sample data set to train and verify the proxy model to obtain a target proxy model;

[0013] Loading a dynamic Bayesian network model, inputting the structural observation data into the dynamic Bayesian network model, and obtaining a structural material parameter vector;

[0014] Inputting the structural material parameter vector into the target proxy model to obtain predicted response data;

[0015] Whether the optimal fit goodness of fit is achieved is determined based on the predicted response data and the structural observed response data. If so, the structural digital twin is realized based on the structural material parameter vector; if not, the steps of obtaining the structural material parameter vector, obtaining the predicted response data and determining whether the optimal fit accuracy is achieved are repeated.

[0016] Further, the vibration response video stream is processed, including:

[0017] A motion amplification algorithm is used to amplify the vibration response video stream;

[0018] The amplified vibration response video stream is framed to obtain a vibration response image;

[0019] Dividing the target area on the vibration response image to obtain a region of interest;

[0020] Performing displacement calculation on the region of interest to obtain structural displacement;

[0021] The structural displacement is subjected to Fourier transformation to obtain the structural vibration frequency.

[0022] Furthermore, the first response data and the second response data both include structural displacement and structural vibration frequency.

[0023] Furthermore, the first response data and the second response data are aligned and fused to obtain structural observation data, which specifically includes:

[0024] Performing temporal and spatial alignment processing on the first response data and the second response data of the same part;

[0025] The first response data and the second response data after alignment are fused to obtain structural observation data.

[0026] Furthermore, the process of constructing the sample data set includes:

[0027] Construct structural finite element models;

[0028] Determine the value range of structural material parameters;

[0029] Sampling the value range of the structural material parameters to obtain different material parameter combinations;

[0030] Utilizing the structural finite element model, calculating responses under different material parameter combinations, and obtaining response data under different structural material parameters;

[0031] The sample data set is constructed according to response data under different structural material parameters.

[0032] Furthermore, the structural material parameters include elastic modulus, density and Poisson's ratio.

[0033] Furthermore, the proxy model uses a Kriging model.

[0034] Based on the same concept, the present invention also provides an electronic device, including a memory, a processor, and a computer program / instruction stored in the memory, wherein the processor executes the computer program / instruction to implement the target state detection method as described above.

[0035] Based on the same concept, the present invention also provides a computer-readable storage medium on which a computer program / instruction is stored. When the computer program / instruction is executed by a processor, the target state detection method as described above is implemented.

[0036] Based on the same concept, the present invention also provides a computer program product, including a computer program / instruction, which implements the target state detection method as described above when executed by a processor.

[0037] Beneficial Effects

[0038] Compared with the prior art, the advantages of the present invention are:

[0039] The present invention combines image acquisition equipment with contact sensors on the surface of structures to achieve synchronous measurement of structural dynamic response, which not only improves the efficiency of structural monitoring, but also improves monitoring accuracy, reduces monitoring costs, and avoids the shortcomings of traditional contact sensor measurement technology for large structures, such as low measurement point density and insufficient monitoring data information, which leads to low accuracy.

[0040] The present invention has flexible on-site arrangement and can realize multi-target synchronous monitoring, thus avoiding the disadvantage of engineering early warning lag caused by traditional monitoring methods relying on offline processing. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the technical solution of the present invention, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only one embodiment of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0042] Figure 1 is a flow chart of a bridge digital twin method in an embodiment of the present invention;

[0043] Figure 2 is a schematic diagram of the arrangement of sensors and image acquisition devices on a bridge in an embodiment of the present invention;

[0044] Figure 3is a schematic diagram of mapping between various coordinate systems in an embodiment of the present invention;

[0045] Figure 4 is the cable acceleration response signal collected by the acceleration sensor in the embodiment of the present invention;

[0046] Figure 5 In the embodiment of the present invention, Figure 4 The corresponding spectrum diagram;

[0047] Figure 6 is a displacement signal obtained by processing the vibration response video stream in an embodiment of the present invention;

[0048] Figure 7 In the embodiment of the present invention, Figure 6 The corresponding spectrum diagram;

[0049] Figure 8 is a schematic diagram of a digital twin model of a bridge in an embodiment of the present invention;

[0050] Fig. 9 is a comparison diagram of the first-order vibration modes of the test model and the digital twin model in the embodiment of the present invention;

[0051] Fig.10 is a comparison diagram of the second-order vibration modes of the test model and the digital twin model in the embodiment of the present invention;

[0052] Fig.11 It is a comparison diagram of the third-order vibration modes of the test model and the digital twin model in the embodiment of the present invention. DETAILED DESCRIPTION

[0053] The following is a clear and complete description of the technical solutions in the present invention 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 creative work are within the scope of protection of the present invention.

[0054] The technical solution of the present application is described in detail with specific embodiments below. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described in detail in some embodiments.

[0055] This embodiment takes a cable-stayed bridge as an example to illustrate the specific implementation process of the structural digital twin method of the present invention. Figure 1 The flow chart of the bridge digital twin method is shown. Figure 1 As shown, the bridge digital twin method includes the following steps:

[0056] Step 1: Obtain the dynamic response signal of the actual bridge, and process the dynamic response signal to obtain the first response data.

[0057] Sensors are arranged at key positions and node positions of the bridge to obtain the actual dynamic response signal of the bridge. In a specific embodiment of the present invention, the sensors include acceleration sensors, displacement sensors and cable tension meters. Displacement sensors are arranged on the main beams of the bridge, and acceleration sensors and cable tension meters are arranged on the cables of the bridge. Figure 2 As shown in the figure, cable tension meters and acceleration sensors are arranged crosswise on the cables S1 to S8, and multiple displacement sensors are arranged at equal intervals at the bottom of the main beam. The dynamic response signals (i.e., acceleration response signals, displacement signals, and cable forces) obtained by each sensor are de-noised and filtered to obtain the first response data of the bridge. That is, the first response data of the bridge includes the displacement of the main beam, the vibration frequency of the cable (the vibration frequency is obtained by Fourier transforming the acceleration response signal), and the cable force.

[0058] Step 2: Obtain a vibration response video stream of an actual bridge, and process the vibration response video stream to obtain second response data.

[0059] In order to obtain the vibration response video stream of the actual bridge under different environmental backgrounds and different vehicle load excitations, a bridge test model is built, and a guide rail is laid on the main beam of the bridge test model. Vehicles with different vehicle weights and speeds are used as external loads of the bridge. The motor drives the vehicle to travel on the guide rail, and the motor speed is changed to control the vehicle speed. The image acquisition device (such as a high-definition camera) is set up at a distance of 1m from the bridge test model. For large bridges, the number of image acquisition devices can be increased to ensure that the image acquisition device can capture video images of the entire bridge structure or multiple regions.

[0060] The pixel resolution of the image acquisition device of this embodiment is 4096×2160 pixels, and the video shooting frame rate is 60Hz. After adjusting the focal length, the image acquisition device is calibrated using a checkerboard calibration plate, and then the image acquisition device is used to collect the vibration response video stream. During calibration, the checkerboard calibration plate is parallel to the plane where the lens is located, and 3 to 5 images of the calibration plate and the bridge are taken. In addition, by driving a counterweight vehicle as an external load excitation, the image acquisition device is used to synchronously collect the vibration response video stream of the bridge under the load, and the vibration response video stream is saved in mp4 format. According to the shooting frame rate, the vibration response video stream is framed and processed to obtain a vibration response image. A frame image in an unexcited state is used as the initial frame, and the resolution and size of subsequent frame images are kept consistent with the initial frame image, and the initial frame is used as the static form of the bridge. In order to avoid the contingency of the test and the influence of environmental noise on the test, multiple load excitation tests are used.

[0061] Image acquisition equipment has lens distortion problems during the imaging process, such as Figure 3 As shown, the world coordinate system, image pixel coordinate system, and camera coordinate system need to be converted using the following formula:

[0062]

[0063] Among them, s represents the scale factor of converting three-dimensional coordinates to two-dimensional coordinates; f x 、f y Respectively represent the focal length in the horizontal and vertical directions; γ represents the tilt factor, which is used to describe the tilt degree on the pixel coordinate axis; c x 、c y represents the pixel coordinates of the principal point (i.e., the pixel coordinates of the intersection of the principal optical axis and the image plane), R represents the rotation matrix, T represents the translation matrix, A represents the intrinsic parameter matrix, and r ij represents the rotation coefficient, t i represents the translation coefficient, X, Y, Z represent the coordinates of point P in the world coordinate system, and u and v represent the coordinates of point P' projected onto the image plane. Figure 3 The f in the figure represents the actual focal length of the image acquisition device.

[0064] Generally, a checkerboard calibration plate is used as a reference object to calibrate the image acquisition device to accurately determine the internal parameters of the image acquisition device. Use the image acquisition device to take about 25 photos of the checkerboard calibration plate from different angles, and then use the Camera Calibrator toolbox in Matlab to calibrate the image acquisition device to obtain the internal and external parameters of the image acquisition device.

[0065] When the monitoring parameter is the displacement of the main beam, the displacement extracted from the vibration response image is at the pixel level. In order to convert the pixel-level displacement to the world coordinate system, it is necessary to calculate the scale factor. The pixel-level displacement in the vibration response image is converted to the displacement in the world coordinate system through the scale factor, so that the displacement of the main beam monitored by the sensor and the displacement monitored by the image acquisition device are at the same scale. The scale factor between the image pixel size and the spatial distance in the world coordinate system is calculated through the conversion relationship between the world coordinate system and the camera coordinate system of the measured object. The checkerboard calibration plate is placed above the bridge and parallel to the plane where the lens is located. By identifying the corner point position on the checkerboard calibration plate, corresponding to the actual size in the world coordinate system, the scale factor a between the image pixel size and the spatial distance in the world coordinate system is obtained after homography transformation. The specific calculation formula is:

[0066]

[0067] Among them, w known ,I konwnThey represent the actual distance between adjacent corner points in the checkerboard calibration plate and the pixel distance between adjacent corner points in the pixel coordinate system after homology transformation.

[0068] In a specific embodiment of the present invention, the vibration response video stream is processed, including:

[0069] Step 2.1: Use motion amplification algorithm to amplify the vibration response video stream.

[0070] Since the excitation effect of vehicle load on the vibration displacement of the entire bridge structure is relatively small, especially the displacement of the main beam, which is difficult to observe with the naked eye, it is necessary to use a motion amplification algorithm to amplify the vibration response video stream. Specifically, the vibration response video stream is first decomposed into different spatial and temporal frequency components through image decomposition technology (such as Laplace pyramid); secondly, a rough vibration frequency range is selected for different parts of the bridge, and then an amplification factor is set for the frequency components within the vibration frequency range, and the frequency components within the vibration frequency range are enhanced using the amplification factor; finally, the amplified frequency components are recombined with other parts of the original vibration response video stream to generate a new video containing amplified motion.

[0071] Step 2.2: According to the shooting frame rate, the amplified vibration response video stream is divided into frames to obtain a vibration response image.

[0072] Step 2.3: Divide the target area on the vibration response image to obtain the region of interest (ROI).

[0073] Step 2.4: Use the straight line tracking algorithm to calculate the displacement of the cable area to obtain the vibration displacement of the cable; use the grayscale centroid method to calculate the displacement of the main beam area to obtain the main beam displacement; and use the template matching algorithm to extract the vibration mode of each part;

[0074] Step 2.5: Perform Fourier transform on the vibration displacement of the cable to obtain the cable vibration frequency; convert the cable vibration frequency into the cable force. That is, the second response data includes the main beam displacement, the cable vibration frequency and the cable force.

[0075] Step 3: Align and fuse the first response data obtained in step 1 with the second response data obtained in step 2 to obtain bridge observation data.

[0076] In a specific embodiment of the present invention, the first response data and the second response data are aligned and fused to obtain the structural observation data, which specifically includes:

[0077] Step 3.1: Performing temporal and spatial alignment processing on the first response data and the second response data of the same part;

[0078] Step 3.2: The aligned first response data and the second response data are fused to obtain bridge observation data.

[0079] The sampling frequency of the sensor is higher than that of the image acquisition device. Therefore, it is necessary to perform temporal alignment processing on the first response data and the second response data of the same part (such as the cable or the main beam). The acquisition point data is increased by upsampling or the acquisition point data is reduced by downsampling, so that the number of acquisition points of the first response data and the second response data of the same part is the same. For the displacement in the first response data and the second response data, the displacement measured by the sensor is at the millimeter level, while the displacement measured by the image acquisition device is at the pixel level. Therefore, it is necessary to use the scale factor in formula (2) to realize the pixel-level displacement conversion to the world coordinate system to realize the spatial alignment of the first response data and the second response data.

[0080] The Kalman filter algorithm is used to fuse the first response data and the second response data after the alignment process. The Kalman filter algorithm is an existing technology. For details, please refer to the literature: "Structural dynamic displacement identification based on visual and vibration monitoring data fusion and its experimental verification" proposed by Xiu Sheng et al., Engineering Mechanics, November 2023. Fusion of the first response data and the second response data can obtain richer bridge response data. Taking vibration frequency as an example, the image acquisition device can accurately extract low-frequency data of the bridge, while high-frequency vibration is difficult to identify. It can be combined with the vibration frequency collected by the acceleration sensor to make the extracted bridge response richer.

[0081] Before the alignment process, the first response data and the second response data can also be cross-checked: taking the vibration frequency as an example, the vibration frequency obtained by the acceleration sensor and the vibration frequency obtained by the image acquisition device have an overlapping part, such as low-frequency information. At this time, the peak error between the two can be calculated. If the peak error between the two exceeds the set error threshold, manual inspection and other means are used for adjustment, thereby achieving mutual verification between the first response data and the second response data.

[0082] The bridge observation data of this embodiment includes cable force, cable vibration frequency and cable force. Figure 4 and Figure 5 The bridge vibration response data obtained based on the acceleration sensor is shown. Figure 6 and Figure 7 The bridge vibration response data obtained based on computer vision is shown in Figures 4 to 7 It can be seen that the bridge vibration response data obtained by using the acceleration sensor is consistent with the bridge vibration response data obtained by using computer vision (i.e., acquired and processed by the image acquisition device). The two can be integrated to realize the digital twin of the bridge.

[0083] Step 4: Build a sample dataset.

[0084] In order to achieve rapid calculation of response data, a proxy model is used to calculate the response data after the bridge material parameters are updated. Before using the proxy model to calculate the response data, the proxy model must be trained first, so a sample data set needs to be constructed. In a specific embodiment of the present invention, the construction process of the sample data set includes:

[0085] Step 4.1: Construct the finite element model of the bridge.

[0086] According to the geometric dimensions of the bridge, the ABAQUS finite element simulation software was used to construct a three-dimensional model of the bridge; the three-dimensional model of the bridge was meshed, and the corresponding loads and boundary conditions were applied to the model according to the actual fixing method and loading method to simulate the working state of the bridge in the actual environment; the different components in the meshed model were defined as a unit set, and appropriate material parameters were set for each component to obtain the finite element model of the bridge.

[0087] In this embodiment, in order to ensure the accuracy of the digital twin model, the grid size is set to 0.1m. The material parameters include elastic modulus, density and Poisson's ratio. In this embodiment, the elastic modulus is set to 2.1×10 5 N / mm 2 , the density is set to 7.9×10 - 6 kg / mm 3 , Poisson’s ratio is set to 0.3.

[0088] Step 4.2: Determine the value range of bridge material parameters.

[0089] Assume that the elastic modulus is 2.1×10 5 N / mm 2 The material density is a normal distribution with a mean of 7.9×10 -6 kg / mm 3 The normal distribution of .

[0090] Step 4.3: Sample the value range of bridge material parameters to obtain different material parameter combinations.

[0091] Latin hypercube sampling is used to sample the elastic modulus and material density respectively. For example, 10 samples are taken from each sample, and then 100 groups of samples can be generated. Then, combined with different Poisson's ratios, multiple groups of samples are obtained (each group of samples includes the elastic modulus, material density and Poisson's ratio).

[0092] Step 4.4: Use the bridge finite element model to calculate the responses under different material parameter combinations and obtain the response data under different structural material parameters.

[0093] Through the ABAQUS-Python secondary development interface, the input parameters of the bridge finite element model are modified according to the different material parameter combinations generated in step 4.3, and the responses under different material parameter combinations are calculated, that is, the response data under different structural material parameters (including main beam displacement, cable vibration frequency and cable force) are obtained.

[0094] Step 4.5: Construct a sample data set according to the response data under different structural material parameters. Each sample in the sample data set includes bridge material parameters and response data under the material parameters.

[0095] Step 5: Load the proxy model, and use the sample data set constructed in step 4 to train and verify the proxy model to obtain the target proxy model.

[0096] A proxy model is a simplified model used to approximate a complex mathematical model or simulation model. It uses a lower computational cost and provides prediction results close to the original complex model. It is mainly used in models that consume a lot of computational time to save resources through approximate calculations. It is widely used in engineering optimization, parameter identification, uncertainty analysis and other fields.

[0097] The proxy model of this embodiment uses the Kriging model, and uses the Kriging model to learn the relationship between the material parameter combination and the response data in the sample data set, verify the calculation accuracy of the Kriging model, and ensure that the Kriging model can well approximate the calculation results of the bridge finite element model. The trained Kriging model (i.e., the target Kriging model) is used to replace the bridge finite element model to calculate the response data. The response data of the bridge structure can be quickly predicted directly through the proxy model, without the need to use finite element software to analyze and calculate the response data, which greatly improves the calculation efficiency of the response data.

[0098] Step 6: Load the dynamic Bayesian network model, input the bridge observation data obtained in step 3 into the dynamic Bayesian network model, and obtain the bridge material parameter vector (i.e., the updated bridge material parameters).

[0099] The dynamic Bayesian network model is a probabilistic graphical model for modeling time series data. It is a time-series extended version of the Bayesian network. Unlike the static Bayesian network, the dynamic Bayesian network can capture the dynamic changes of the system over time and represent the causal relationship between variables that change over time. It is suitable for systems involving uncertainty, noise and time dependence. In the present invention, it is used to update structural parameters to infer the evolution of the structural health state.

[0100] The dynamic Bayesian network model realizes the dynamic update of material parameters by maximizing the posterior probability of parameters. Due to the differences in sensor types and computer vision target domains, the acquired response data is heterogeneous in dimension. Therefore, the first response data and the second response data are fused through data fusion, and the multi-source response data are fused and converted into homogeneous response data (i.e., bridge observation data). The bridge observation data is used as the input of the dynamic Bayesian network model. According to the joint likelihood function and prior distribution of the bridge observation data, the posterior distribution of the material parameters is calculated by Bayesian theorem, and a dynamic Bayesian network model is constructed to realize the dynamic real-time update of bridge material parameters. The dynamic Bayesian network model-driven parameter update is an existing technology. For details, please refer to the literature: "Finite element model update method of aqueduct structure based on Bayesian theory" proposed by Wu Xijie et al., Journal of Hydraulic Engineering, 2024(6): 1-9; and the literature: "Joint Robin sparse Bayesian learning structural damage identification model based on mixed Gaussian distribution" proposed by Li Rongpeng et al., China Journal of Highway and Transport, May 2024.

[0101] Step 7: Input the structural material parameter vector obtained in step 6 into the target proxy model in step 5 to obtain the predicted response data (ie, the updated response data).

[0102] Step 8: Determine whether the optimal fit is achieved based on the predicted response data obtained in step 7 and the structural observed response data obtained in step 3. If so, input the structural material parameter vector into the bridge finite element model to obtain the bridge digital twin model, such as Figure 8 If not, repeat steps 6 to 8 until the best fit is achieved.

[0103] Figures 9 to 11 The modal parameters of the field test model and the first three modal parameters of the main beam of the digital twin model after the material parameters are updated are compared. Figures 9 to 11 It can be seen that the vibration modes of the two are very similar; at the same time, combined with the specific values ​​of the first three eigenfrequencies of the two (as shown in Table 1), it can be calculated that the maximum error is only 4.10%, which shows that the digital twin model can effectively reflect the dynamic characteristics of the bridge structure.

[0104] Table 1 Comparison of characteristic frequencies of the main beam of the field test model and the digital twin model of the bridge structure

[0105] Degree Test model characteristic frequency Digital Twin Model Eigenfrequencies error First level 10.485 10.715 2.15％ Second order 21.786 20.892 4.10％ Third level 46.251 44.649 3.46％

[0106] The present invention does not use an objective function or a purely data-driven method to establish a digital twin model. Instead, it establishes a probabilistic model based on physical guidance and multi-source response data to achieve material parameter updates, reduce costs, improve the accuracy of digital twins, and achieve accurate assessment of the health status of bridge structures.

[0107] Example 2

[0108] An embodiment of the present invention also provides an electronic device, which includes: a memory, a processor, and a computer program / instructions stored in the memory, and the processor executes the computer program / instructions to implement the structural digital twin method in the embodiment of the present application.

[0109] Although not shown, the electronic device includes a processor, which can perform various appropriate operations and processes according to the program and / or data stored in the read-only memory (ROM) or the program and / or data loaded from the storage portion into the random access memory (RAM). The processor can be a multi-core processor, or it can include multiple processors. In some embodiments, the processor can include a general main processor and one or more special coprocessors, such as a central processing unit, a graphics processing unit (GPU), a neural network processor (NPU), a digital signal processor (DSP), etc. In RAM, various programs and data required for device operation are also stored. The processor, ROM and RAM are connected to each other via a bus. The input / output (I / O) interface is also connected to the bus.

[0110] The processor and memory are used together to execute the program / instructions stored in the memory. When the program / instructions are executed by the computer, the methods, steps or functions described in the above embodiments can be implemented.

[0111] Although not shown, an embodiment of the present invention also provides a computer-readable storage medium having a computer program / instruction stored thereon, which, when executed by a processor, implements the structural digital twin method in the embodiment of the present application.

[0112] Readable storage media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer-readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer-readable media does not include temporary computer-readable media (transitory media), such as modulated data signals and carrier waves.

[0113] Although not shown, an embodiment of the present invention also provides a computer program product, including: a computer program / instructions, which, when executed by a processor, implement the structural digital twin method in the embodiment of the present application.

[0114] What is disclosed above is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of changes or modifications within the technical scope disclosed in the present invention, which should be covered within the protection scope of the present invention.

Claims

1. A structural digital twin method, characterized in that: The twinning method comprises: Acquire a dynamic response signal of the actual structure, and process the dynamic response signal to obtain first response data; wherein the dynamic response signal is acquired by a sensor arranged on the actual structure; Acquire a vibration response video stream of the actual structure, and process the vibration response video stream to obtain second response data; Aligning and fusing the first response data with the second response data to obtain structural observation data; Constructing a sample data set; wherein each sample in the sample data set includes structural material parameters and corresponding response data; Loading the proxy model, and using the sample data set to train and verify the proxy model to obtain a target proxy model; Loading a dynamic Bayesian network model, inputting the structural observation data into the dynamic Bayesian network model, and obtaining a structural material parameter vector; Inputting the structural material parameter vector into the target proxy model to obtain predicted response data; Whether the optimal fit goodness of fit is achieved is determined based on the predicted response data and the structural observed response data. If so, the structural digital twin is realized based on the structural material parameter vector; if not, the steps of obtaining the structural material parameter vector, obtaining the predicted response data and determining whether the optimal fit accuracy is achieved are repeated.

2. The structural digital twin method according to claim 1, characterized in that: Processing the vibration response video stream includes: A motion amplification algorithm is used to amplify the vibration response video stream; The amplified vibration response video stream is framed to obtain a vibration response image; Dividing the target area on the vibration response image to obtain a region of interest; Performing displacement calculation on the region of interest to obtain structural displacement; The structural displacement is subjected to Fourier transformation to obtain the structural vibration frequency.

3. The structural digital twin method according to claim 1, characterized in that: The first response data and the second response data both include structural displacement and structural vibration frequency.

4. The structural digital twin method according to claim 1, characterized in that: Aligning and fusing the first response data with the second response data to obtain structural observation data specifically includes: Performing temporal and spatial alignment processing on the first response data and the second response data of the same part; The first response data and the second response data after alignment are fused to obtain structural observation data.

5. The structural digital twin method according to claim 1, characterized in that: The process of constructing the sample data set includes: Construct structural finite element models; Determine the value range of structural material parameters; Sampling the value range of the structural material parameters to obtain different material parameter combinations; Utilizing the structural finite element model, calculating responses under different material parameter combinations, and obtaining response data under different structural material parameters; The sample data set is constructed according to response data under different structural material parameters.

6. The structural digital twin method according to claim 1, characterized in that: The structural material parameters include elastic modulus, density and Poisson's ratio.

7. The structural digital twin method according to claim 1, characterized in that: The proxy model uses the Kriging model.

8. An electronic device comprising a memory, a processor, and a computer program / instruction stored in the memory, characterized in that: The processor executes the computer program / instructions to implement the target state detection method according to any one of claims 1 to 7.

9. A computer-readable storage medium having a computer program / instruction stored thereon, characterized in that: When the computer program / instruction is executed by a processor, the target state detection method according to any one of claims 1 to 7 is implemented.

10. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instruction is executed by a processor, the target state detection method according to any one of claims 1 to 7 is implemented.

Images