Underwater bridge pier erosion state discrimination method based on flow field characteristics

Abstract

The invention provides an underwater bridge pier erosion state discrimination method based on flow field characteristics. The method comprises the following steps: arranging data measurement points of a flow field time history around a bridge pier; through numerical simulation or on-site actual measurement, a streaming time history sample of the complete bridge pier and the bridge pier with the erosion defect is obtained; constructing a time convolutional neural network model, taking the streaming time history sample as input, training the time convolutional neural network model, and outputting a pier erosion state discrimination result through feature extraction and classification; collecting time history data of a flow field around a to-be-evaluated bridge pier; and inputting the collected time history data of the flow field around the to-be-evaluated bridge pier into the trained deep learning model, and outputting an erosion state discrimination result. Intelligent and high-precision discrimination of the underwater erosion state is realized by analyzing the characteristics (pressure, flow velocity and the like) of the flow field around the bridge pier, and the problems of poor accessibility, low precision and insufficient environmental adaptability of a traditional detection means at a hidden part are solved.

Description

Technical Field

[0001] This invention relates to the field of on-site detection technology for the erosion state of underwater bridge piers, and more particularly to a method for determining the erosion state of underwater bridge piers based on flow field characteristics. Background Technology

[0002] As a critical supporting component of bridge structures, underwater bridge piers are submerged in complex aquatic environments for extended periods, continuously enduring the combined effects of water erosion, sediment abrasion, and chemical corrosion. Pier erosion refers to the gradual loss of pier material under hydrodynamic forces, leading to a reduction in effective cross-sectional area. Its extent directly impacts the pier's load-bearing capacity, structural stability, and overall bridge safety. Erosion not only degrades material properties but also alters the flow field characteristics around the pier, causing localized flow turbulence, exacerbating structural vibration risks, and potentially triggering bridge collapse in severe cases. Therefore, real-time and accurate monitoring of underwater bridge pier erosion is a core technological requirement for ensuring the safe operation of bridges throughout their entire lifecycle. The detection of underwater bridge pier erosion currently relies primarily on three traditional methods: first, contact measurement, such as underwater exploration by divers or robots equipped with sensors directly contacting the pier surface to obtain erosion depth and morphology data; second, optical detection, such as underwater photography and laser scanning, which record surface defects through optical imaging; and third, indirect assessment methods, which estimate erosion depth based on empirical formulas or simplified models. In recent years, some studies have attempted to introduce flow field analysis techniques, such as acoustic Doppler current meters (ADV) or particle image velocimetry (PIV), to collect flow velocity field data and combine this with computational fluid dynamics (CFD) simulations to analyze the impact of erosion on the flow field, providing auxiliary data for condition assessment. These methods have improved detection efficiency to some extent, but a systematic intelligent discrimination system has not yet been formed. Existing technologies suffer from the following core shortcomings: First, contact-based detection is limited by the high-risk nature of the underwater environment (turbulent currents, low visibility) and poor accessibility of concealed areas, making it difficult to obtain complete erosion information, and also incurring high detection costs and long cycles. Second, optical detection is affected by water turbidity and light attenuation, resulting in insufficient accuracy in identifying minute erosion defects, and it cannot penetrate sediment cover layers. Finally, existing flow field analysis methods are limited to qualitative correlation, lacking quantitative modeling of erosion state and flow field characteristic parameters (such as pressure and vorticity), and have not established standardized intelligent discrimination models, making it difficult to achieve efficient and accurate automated monitoring in practical engineering. These shortcomings severely restrict the real-time assessment and early warning capabilities of underwater bridge pier erosion status, necessitating a breakthrough in traditional technological bottlenecks to develop an intelligent discrimination method for underwater bridge pier erosion status based on flow field characteristics and deep learning. Summary of the Invention To address the aforementioned technical problems, this invention provides a method for determining the erosion state of underwater bridge piers based on flow field characteristics. This invention utilizes flow field time history data and a deep learning model (TCNN) to achieve intelligent and high-precision determination of the erosion state of underwater bridge piers through non-contact flow field feature analysis.

[0003] The technical means employed in this invention are as follows: A method for determining the erosion state of underwater bridge piers based on flow field characteristics includes: S1. Arrange flow field time history data measurement points around the bridge piers; S2. Obtain flow time history samples of complete bridge piers and bridge piers with erosion defects through numerical simulation or field measurement; S3. Construct a temporal convolutional neural network model, using the flow time history sample as input, train the temporal convolutional neural network model, and output the pier erosion status discrimination result through feature extraction and classification; S4. Collect time history data of the flow field around the bridge pier to be evaluated; S5. Input the time history data of the flow field around the bridge pier to be evaluated into the trained deep learning model and output the erosion state discrimination result.

[0004] Further, step S1 specifically includes: S11. Determine the flow range for arranging measuring points. The measuring points should cover the area downstream of the bridge pier with a transverse flow length ≥ 4D and a longitudinal flow length ≥ 4D, where D is the characteristic dimension of the bridge pier. S12. Within the defined flow range of the measurement points, generate the coordinates of the measurement points by random or regular point placement; S13. Remove invalid measuring point coordinates inside the bridge piers to obtain the measuring point set.

[0005] Further, step S2 specifically includes: S21. Determine the shape and size of the bridge pier to be identified; S22. Determine the parameters and medium properties for the numerical simulation, including Reynolds number, simulated medium, density, viscosity and flow rate of the simulated medium; S23. Determine the computational domain of the numerical simulation to fully capture the changes in the flow field around the bridge pier; S24. Based on the geometric shape and flow field variation characteristics of the piers, mesh division is performed on both intact piers and intact piers with erosion defects. S25. Set the boundary conditions for the numerical simulation according to the actual situation, including: setting the inlet boundary as a velocity inlet with the direction of flow; setting the outlet boundary as a pressure outlet; and setting the pier surface as a no-slip boundary condition, that is, the velocity of the fluid on the pier surface is zero. S26. Start numerical simulation calculation, set a reasonable time step and number of iterations according to the actual situation, and let the software solve the fluid control equations iteratively according to the time step. Extract the pressure time history data of specific measuring points around the complete and erosion-defect complete bridge piers from the results to obtain time history samples for subsequent analysis. The sample set of time history signals is N.

[0006] Further, step S3 specifically includes: S31. Construct a Time-CNN (TCNN) model for deep learning of flow field data using one-dimensional convolution; S32. Input the flow field pressure time history signal set N as input sample into the time convolutional neural network model; S33. Define the loss function L of the model as follows: L=∑|OG| Where O is the target shape label calculated and predicted by the model, and G is the actual shape label; S34. Conduct iterative training. During this process, the loss function value defined in step S33 continues to decrease. When the loss function value decreases to a certain extent, the network model parameters suitable for shape recognition within the application scope are obtained.

[0007] Further, step S4 specifically includes: S41. Obtain the underwater erosion status of bridge piers that require identification. Obtain the underwater erosion status identification requirements of bridge inspection companies. Combine the service life of the bridge piers, the intensity of water flow scouring and surface erosion characteristics to screen target bridge piers that have been subjected to long-term mainstream scouring and show signs of erosion. Establish an information archive that includes location, structure and surrounding hydrological parameters. S42. Install sensors at key locations around the bridge piers, ensure accurate underwater positioning and waterproofing, debug the equipment, check sensor sensitivity indicators, and establish a real-time monitoring system. S43. Collect flow field time history data in real time according to the set sampling frequency (100 times per second), and record key parameters, including the acquisition time, simultaneously.

[0008] Further, step S5 specifically includes: S51. Input the flow field time history data collected in step S43 into the time convolutional neural network model trained in step S3, and use the network model parameters obtained in step S34 to perform calculations. S52. Based on the calculated result vector, determine its similarity with each label in the pier to be identified, find the label m that is most similar to it, and output the sample shape of the corresponding label m in the pier to be identified as the identification result, thereby completing the identification of the target shape and outputting the erosion state discrimination result.

[0009] Compared with the prior art, the present invention has the following advantages: 1. The present invention provides a method for determining the erosion state of underwater bridge piers based on flow field characteristics. It uses flow field characteristic analysis to replace traditional contact detection, solving the problem of underwater low visibility, high flow velocity and hidden parts that cannot be directly detected, and significantly reducing detection risks and costs.

[0010] 2. The present invention provides a method for determining the erosion state of underwater bridge piers based on flow field characteristics, which fills the gap in existing methods that cannot effectively identify erosion defects of complete bridge piers through flow field data. It uses flow field velocity time history signals as the data basis to study the erosion state of bridge piers. This data acquisition and analysis method can accurately capture the flow field change characteristics caused by erosion and ensure the correlation with the actual erosion state of the bridge pier.

[0011] 3. The present invention provides a method for judging the erosion state of underwater bridge piers based on flow field characteristics, which fills the gap in existing research that cannot compare the impact of different deep learning architectures on the accuracy of bridge pier erosion recognition. It can accurately determine the performance difference between fully connected neural networks (MLP) and temporal convolutional neural networks (TCNN) when processing flow field data, and clarify the advantages and limitations of different models in feature extraction and classification.

[0012] 4. The present invention provides an underwater bridge pier erosion state discrimination method based on flow field characteristics, which fills the gap in existing research that makes it difficult to quantify and analyze the adaptability of different models to complex water flow environments. It can obtain the stability variation law of the TCNN model for bridge pier erosion identification under different water flow velocities and flow directions, and provide data support for the actual engineering application of the model.

[0013] Based on the above reasons, this invention can be widely applied in fields such as intelligent monitoring of underwater bridge pier erosion and bridge structural health assessment. Attached Figure Description

[0014] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0015] Figure 1 This is a flowchart of the method of the present invention.

[0016] Figure 2 This is a complete diagram of the bridge pier measuring point layout provided for an embodiment of the present invention.

[0017] Figure 3 This is a diagram showing the layout of measuring points for erosion-defect bridge piers, provided in an embodiment of the present invention.

[0018] Figure 4 The target shape provided for embodiments of the present invention.

[0019] Figure 5 Example diagram of complete pier pressure time history curve provided for embodiments of the present invention.

[0020] Figure 6 Example diagram of pressure time history curve of erosion defect bridge pier provided in the embodiment of the present invention.

[0021] Figure 7 The Temporal Convolutional Neural Network (TCNN) model structure diagram provided for embodiments of the present invention.

[0022] Figure 8 This is a diagram showing the accuracy of the TCNN model in recognizing complete bridge piers, as provided in an embodiment of the present invention.

[0023] Figure 9 The image shows the accuracy of the TCNN model for identifying erosion defects on bridge piers, as provided in this embodiment of the invention. Detailed Implementation

[0024] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0025] It should be noted that the terms "comprising" and "having" and any variations thereof in the specification, claims and accompanying drawings of this invention are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are explicitly listed, but may include other steps or units that are not explicitly listed or that are inherent to such processes, methods, products or devices.

[0026] like Figure 1 As shown, this invention provides a method for determining the erosion state of underwater bridge piers based on flow field characteristics, including: S1. Arrange flow field time history data measurement points around the bridge piers; S2. Obtain flow time history samples of complete bridge piers and bridge piers with erosion defects through numerical simulation or field measurement; S3. Construct a temporal convolutional neural network model, using the flow time history sample as input, train the temporal convolutional neural network model, and output the pier erosion status discrimination result through feature extraction and classification; S4. Collect time history data of the flow field around the bridge pier to be evaluated; S5. Input the time history data of the flow field around the bridge pier to be evaluated into the trained deep learning model and output the erosion state discrimination result.

[0027] In a specific implementation, as a preferred embodiment of the present invention, step S1 specifically includes: S11. Determine the flow range for arranging measuring points. The measuring points should cover the area downstream of the bridge pier with a transverse flow length ≥ 4D and a longitudinal flow length ≥ 4D, where D is the characteristic dimension of the bridge pier. In this embodiment, the bridge pier is taken as the research object. D The diameter of the bridge pier is given. In the downstream region of the bridge pier, the transverse flow length is set to 6. D The length along the flow direction is 6 D A rectangular area was designated as the specific range for the measurement points. This area was chosen based on research into the flow field characteristics around the underwater bridge piers, which can effectively capture information on flow field changes caused by pier erosion.

[0028] S12. Within the defined flow range of the measured points, the coordinates of the measured points are generated by random or regular point placement. In this embodiment, the measured point placement calculation is performed for two examples. 5040 measured point coordinates are randomly generated within the range of 2 to 8 on the horizontal axis and -3 to 3 on the vertical axis. Among them, 5040 measured points are placed on the intact bridge piers and 5040 measured points are placed on the bridge piers with erosion defects. Finally, a total of 10800 measured points are obtained. Random point placement can enable the measured points to cover a wider area in the study area, which helps to capture possible local special changes in the flow field and avoid information omissions caused by regular point placement. S13. Remove invalid measuring point coordinates inside the pier to obtain the measuring point set. In this embodiment, for the 10800 measuring point coordinates generated in step S12, based on the position and size of the complete pier, determine whether each measuring point is inside the pier. If the measuring point is inside the pier, it is deleted from the measuring point set; if it is outside the pier, it is retained. Through filtering, a valid measuring point set containing only the outside of the pier is obtained, wherein the measuring point arrangement of the complete pier is as follows. Figure 2 As shown, the arrangement of measuring points on the erosion-damaged bridge pier is as follows: Figure 3 As shown; In a specific implementation, as a preferred embodiment of the present invention, step S2 specifically includes: S21. Determine the shape and size of the pier to be identified; in this embodiment, the pier to be identified covers two types of target shapes: intact piers with complete structures and piers with erosion defects. Specific shapes can be found in the attached document. Figure 4 As shown. It should be noted that the technical solution of this patent has wide applicability. In different application scenarios, the shape of the target to be identified will be flexibly adjusted according to the actual working conditions and detection requirements, and is not limited to the shape listed in this embodiment; S22. Determine the parameters and medium properties for the numerical simulation, including the Reynolds number, the simulated medium, and the density, viscosity, and flow rate of the simulated medium. In this embodiment, the Reynolds number is set to Re=3900, and water is selected as the simulated medium with a density of 1000 kg / m³. 3 The viscosity is 1.005 × 10⁻⁶. -5 Pa·s, flow velocity of 1m / s, these parameters can be determined to simulate the stress and flow field of bridge piers under actual working conditions more accurately; S23. Determine the computational domain range of the numerical simulation to fully capture the changes in the flow field around the bridge pier. In this embodiment, the flow length of the computational domain is 10D and the lateral length is 6D, where D is the diameter of the complete bridge pier. The size of the computational domain is chosen to ensure that the changes in the flow field around the bridge pier can be fully captured, avoiding the influence of the flow field boundary effect due to the computational domain being too small, and also avoiding the increase in computational cost due to the computational domain being too large. S24. Based on the geometry and flow field characteristics of the piers, mesh generation is performed on both intact piers and intact piers with erosion defects. In this embodiment, the mesh of the intact pier contains 625,332 nodes and 85,160 elements, while the mesh of the intact pier with erosion defects contains 604,627 nodes and 44,320 elements. During mesh generation, the mesh density and distribution need to be adjusted appropriately according to the geometry and flow field characteristics of the piers. For the pier surface and areas with drastic flow field changes, a denser mesh is used to improve the simulation accuracy; for areas with relatively gentle flow field changes, a sparser mesh is used. S25. Set the boundary conditions for numerical simulation according to the actual situation, including: setting the inlet boundary as a velocity inlet with a velocity of 1 m / s and a flow direction; setting the outlet boundary as a pressure outlet with a pressure value of 0; setting the pier surface as a no-slip boundary condition, that is, the velocity of the fluid on the pier surface is zero; in this embodiment, the setting of these boundary conditions can accurately simulate the interaction between the fluid and the pier, ensuring the authenticity of the simulation results. S26. Start the numerical simulation calculation. Set a reasonable time step and number of iterations based on the actual situation. Let the software iteratively solve the fluid control equations according to the time steps. Extract pressure time history data from specific measuring points around the intact and erosion-defect-incomplete bridge piers from the results to obtain time history samples for subsequent analysis. The sample set of time history signals is N. In this embodiment, there are 2 samples of the bridge piers to be identified, each containing 5400 flow field signals. Therefore, the total signal set N in this embodiment contains 10800 sample time histories. Figure 5 As shown, six time-history curves under different pressures are listed for the complete bridge pier, such as... Figure 6 As shown, six time-history curves under different pressures are listed for bridge piers with erosion defects; In a specific implementation, as a preferred embodiment of the present invention, step S3 specifically includes: S31. Construct a Time-CNN (TCNN) model for deep learning of flow field data using one-dimensional convolution; in this embodiment, the model structure is as follows: Figure 7 As shown; the model input layer is a one-dimensional flow field pressure time history signal; convolution operation is performed on the time history signal of the input layer to obtain convolutional layer 1 of the model; convolution operation is performed again on the output data of convolutional layer 1 to obtain convolutional layer 2 of the model; convolution operation is performed again on the output data of convolutional layer 2 to obtain convolutional layer 3 of the model; convolution operation is performed again on the output data of convolutional layer 3 to obtain convolutional layer 3 of the model; global pooling operation is performed on the output data of convolutional layer 4 to obtain pooling layer 1 of the model; fully connected layer operation is performed on the output data of pooling layer to obtain the output layer of the model; S32. The flow field pressure time history signal set N is used as the input sample and input into the time convolutional neural network model. In this embodiment, during the training process, a certain proportion of the samples are randomly selected from the entire sample as the training set, and the remaining samples are used as the validation set. This ensures that the training set and the test set contain the sample features of the flow field at each location relatively evenly. The specific calculation example is set as follows: the proportion of training samples is 10%, and the number of training samples is 358.

[0029] S33. Define the loss function L of the model as follows: L=∑|OG| Where O is the target shape label predicted by the model calculation, and G is the actual shape label; in this embodiment, the initial label of the model is set to label 0, and the actual labels are 0 and 1, which correspond to the structurally complete bridge piers and the bridge piers with erosion defects in the bridge piers to be identified, respectively. S34. Iterative training is conducted. During this process, the loss function value defined in step S33 continuously decreases. When the loss function value decreases to a certain extent, the network model parameters suitable for shape recognition within the application scope are obtained. In this embodiment, reverse iteration is used to eliminate errors in the model output results, prompting the model to gradually converge. In this embodiment, a total of 200 iterations were performed. At this point, the loss function in step S23 has decreased to a sufficiently small value, fully meeting the preset accuracy requirements. Thus, the model training is successfully completed.

[0030] In a specific implementation, as a preferred embodiment of the present invention, step S4 specifically includes: S41. Select the test site and the bridge piers to be evaluated for underwater erosion according to the requirements; obtain the bridge piers with underwater erosion status that need to be identified; obtain the underwater erosion status identification requirements of the bridge inspection company; combine the service life of the bridge piers, the intensity of water flow scouring and surface erosion characteristics; screen the target bridge piers that have been subjected to long-term mainstream scouring and have signs of erosion; and establish an information archive that includes location, structure and surrounding hydrological parameters. S42. Equipment installation and commissioning: Prepare sensors, data acquisition instruments and other equipment; install sensors in key locations around the bridge piers; ensure accurate underwater positioning and waterproofing; commission the equipment; check sensor sensitivity indicators; and establish a real-time monitoring system. S43. Collect flow field time history data in real time according to the set sampling frequency (100 times per second), and record key parameters, including the collection time, simultaneously. In this embodiment, the data is simultaneously saved to local storage devices and cloud servers to achieve dual backup. The acquisition device is inspected at fixed intervals to ensure the integrity and accuracy of the data collection. In a specific implementation, as a preferred embodiment of the present invention, step S5 specifically includes: S51. Input the flow field time history data collected in step S43 into the time convolutional neural network model trained in step S3, and use the network model parameters obtained in step S34 to perform calculations. In this embodiment, a portion of the data collected in S43 can be used as sample data to be verified. Since its data characteristics are not exactly the same as the training samples, it can represent new samples that the model has not learned and be used to test the model accuracy. S52. Based on the calculated result vector, determine its similarity with each label in the pier to be identified, find the label m that is most similar to it, and output the sample shape of the corresponding label m in the pier to be identified as the identification result, thereby completing the identification of the target shape and outputting the erosion state discrimination result.

[0031] In summary, the TCNN model shows that the confidence level of most measurement points is concentrated around 90%, with a small number of misclassified measurement points located near the downstream area of ​​the bridge pier. The results further demonstrate the feasibility of the underwater bridge pier erosion state identification method proposed in this invention. Furthermore, the TCNN model exhibits higher accuracy; the TCNN model's accuracy in identifying intact bridge piers is shown in the figure. Figure 8 The TCNN model's accuracy in identifying erosion-damaged bridge piers is shown in [reference needed]. Figure 9 .

[0032] Finally, 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 foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for determining the erosion state of underwater bridge piers based on flow field characteristics, characterized in that, include: S1. Arrange flow field time history data measurement points around the bridge piers; S2. Obtain flow time history samples of complete bridge piers and bridge piers with erosion defects through numerical simulation or field measurement; S3. Construct a temporal convolutional neural network model, using the flow time history sample as input, train the temporal convolutional neural network model, and output the pier erosion status discrimination result through feature extraction and classification; S4. Collect time history data of the flow field around the bridge pier to be evaluated; S5. Input the time history data of the flow field around the bridge pier to be evaluated into the trained deep learning model and output the erosion state discrimination result.

2. The method for determining the erosion state of underwater bridge piers based on flow field characteristics according to claim 1, characterized in that, Step S1 specifically includes: S11. Determine the flow range for arranging measuring points. The measuring points should cover the area downstream of the bridge pier with a transverse flow length ≥ 4D and a longitudinal flow length ≥ 4D, where D is the characteristic dimension of the bridge pier. S12. Within the defined flow range of the measurement points, generate the coordinates of the measurement points by random or regular point placement; S13. Remove invalid measuring point coordinates inside the bridge piers to obtain the measuring point set.

3. The method for determining the erosion state of underwater bridge piers based on flow field characteristics according to claim 1, characterized in that, Step S2 specifically includes: S21. Determine the shape and size of the bridge pier to be identified; S22. Determine the parameters and medium properties for the numerical simulation, including Reynolds number, simulated medium, density, viscosity and flow rate of the simulated medium; S23. Determine the computational domain of the numerical simulation to fully capture the changes in the flow field around the bridge pier; S24. Based on the geometric shape and flow field variation characteristics of the piers, mesh division is performed on both intact piers and intact piers with erosion defects. S25. Set the boundary conditions for the numerical simulation according to the actual situation, including: setting the inlet boundary as a velocity inlet with the direction of flow; setting the outlet boundary as a pressure outlet; and setting the pier surface as a no-slip boundary condition, that is, the velocity of the fluid on the pier surface is zero. S26. Start numerical simulation calculation, set a reasonable time step and number of iterations according to the actual situation, and let the software solve the fluid control equations iteratively according to the time step. Extract the pressure time history data of specific measuring points around the complete and erosion-defect complete bridge piers from the results to obtain time history samples for subsequent analysis. The sample set of time history signals is N.

4. The method for determining the erosion state of underwater bridge piers based on flow field characteristics according to claim 1, characterized in that, Step S3 specifically includes: S31. Construct a temporal convolutional neural network model for deep learning of flow field data using one-dimensional convolution; S32. Input the flow field pressure time history signal set N as input sample into the time convolutional neural network model; S33. Define the loss function L of the model as follows: L=∑|OG| Where O is the target shape label calculated and predicted by the model, and G is the actual shape label; S34. Conduct iterative training. During this process, the loss function value defined in step S33 continues to decrease. When the loss function value decreases to a certain extent, the network model parameters suitable for shape recognition within the application scope are obtained.

5. The method for determining the erosion state of underwater bridge piers based on flow field characteristics according to claim 1, characterized in that, Step S4 specifically includes: S41. Obtain the underwater erosion status of bridge piers that require identification. Obtain the underwater erosion status identification requirements of bridge inspection companies. Combine the service life of the bridge piers, the intensity of water flow scouring and surface erosion characteristics to screen target bridge piers that have been subjected to long-term mainstream scouring and show signs of erosion. Establish an information archive that includes location, structure and surrounding hydrological parameters. S42. Install sensors at key locations around the bridge piers, ensure accurate underwater positioning and waterproofing, debug the equipment, check sensor sensitivity indicators, and establish a real-time monitoring system. S43. Collect flow field time history data in real time according to the set sampling frequency, and record key parameters, including the acquisition time, simultaneously.

6. The method for determining the erosion state of underwater bridge piers based on flow field characteristics according to claim 1, characterized in that, Step S5 specifically includes: S51. Input the flow field time history data collected in step S43 into the time convolutional neural network model trained in step S3, and use the network model parameters obtained in step S34 to perform calculations. S52. Based on the calculated result vector, determine its similarity with each label in the pier to be identified, find the label m that is most similar to it, and output the sample shape of the corresponding label m in the pier to be identified as the identification result, thereby completing the identification of the target shape and outputting the erosion state discrimination result.

Images