Pile foundation scouring depth prediction model construction method based on multi-modal parameters

Abstract

The embodiment of the invention discloses a method for constructing a pile foundation scouring depth prediction model based on multi-modal parameters. The method comprises the steps that acceleration and scouring depth time history data of a pile foundation in the scouring process are obtained; for each scouring depth, the following operations are executed: according to the acceleration time history data, the frequencies of the pile foundation before and after scouring and the vibration modes of a plurality of measuring points on the pile foundation are identified; calculating a frequency change rate according to the frequencies before and after washing; calculating modal curvature and diagonal flexibility of each measuring point according to the vibration mode of each measuring point; performing principal component analysis on the modal curvature and the diagonal flexibility of each measuring point, retaining the principal components of the first plurality of levels as curvature features, and forming a feature vector with the frequency change rate; and training an ANFIS model by using each scouring depth and the corresponding feature vector, wherein the trained model is used for automatically predicting the corresponding scouring depth according to one feature vector. According to the embodiment of the invention, through multi-modal parameters, the uncertainty influence is reduced, and the prediction reliability and accuracy are improved.

Description

Technical Field

[0001] This invention relates to the field of scour depth prediction technology, and in particular to a method for constructing a pile foundation scour depth prediction model based on multimodal parameters. Background Technology

[0002] Foundation scour is the leading cause of decreased bridge safety and durability. A significant number of bridge structural failures are related to foundation scour. Therefore, accurately assessing scour depth has become a crucial aspect of bridge health monitoring and safe operation, and the accuracy of its predictions directly impacts the reliability of bridge life assessments.

[0003] Currently, methods for detecting bridge scour depth mainly include traditional methods and non-destructive testing methods. Among them, traditional methods rely on manual underwater operations, and the accuracy of the detection is highly dependent on human experience. They cannot be implemented during flood season or in harsh environments, and have poor applicability.

[0004] Non-destructive testing methods mainly include gravity-based scour depth monitoring, acoustic monitoring, and time-domain reflectometry. Gravity-based methods assess scour depth by monitoring the settling distance of heavy objects; their principle is simple and they allow for continuous monitoring, and they have been applied in some projects. However, during the silt backfilling stage, heavy objects are often buried, making it difficult to reflect the backfilling process. Ultrasonic and sonar detection are currently common underwater monitoring methods, allowing for multiple measurements. However, these methods are sensitive to hydrological conditions and weather, which can easily lead to inaccurate predictions.

[0005] The aforementioned methods suffer from drawbacks in practical applications, including cumbersome underwater installation, poor durability, significant susceptibility to environmental factors, and high costs, making large-scale application in engineering projects difficult. To meet engineering needs, several novel scour monitoring methods exist in practical engineering. For example, patent CN107179172A discloses a bridge pier scour monitoring system and method based on an impact hammer, which uses vibration data to determine whether the pier frequency has changed, thereby accurately and sensitively monitoring the scour condition. Patent CN118607340B discloses a bridge single-phase flow calculation method based on a Physical Information Neural Network (PINN), which can quickly and accurately predict the scour depth and flow field distribution of underwater bridge foundations using information such as flow field and riverbed. However, most existing prediction methods rely on single-modal parameters, making them susceptible to interference from uncertainties such as temperature, leading to decreased prediction accuracy. How to introduce multi-modal parameters and reduce the impact of uncertainties to improve prediction reliability and accuracy is an urgent problem to be solved. Summary of the Invention

[0006] This invention provides a method for constructing a pile foundation scour depth prediction model based on multimodal parameters to solve the above-mentioned technical problems.

[0007] In a first aspect, embodiments of the present invention provide a method for constructing a pile foundation scour depth prediction model based on multimodal parameters, including:

[0008] Obtain time history data of acceleration and scour depth of pile foundation during the scour process;

[0009] For each scour depth, the following operations are performed: Based on the acceleration time history data, the frequencies before and after scour of the pile foundation, as well as the mode shapes of multiple measuring points on the pile foundation, are identified; based on the frequencies before and after scour, the frequency change rate is calculated; based on the mode shapes of each measuring point, the modal curvature and diagonal compliance of each measuring point are calculated; principal component analysis is performed on the modal curvature and diagonal compliance of each measuring point, and the first few levels of principal components are retained as curvature features, which, together with the frequency change rate, form a feature vector;

[0010] An ANFIS model is trained using each scour depth and its corresponding feature vector. The trained model is then used to automatically predict the corresponding scour depth based on a feature vector.

[0011] In a second aspect, embodiments of the present invention provide an electronic device, the electronic device comprising:

[0012] One or more processors;

[0013] Memory, used to store one or more programs;

[0014] When the one or more programs are executed by the one or more processors, the one or more processors implement the pile foundation scour depth prediction model construction method based on multimodal parameters as described in any embodiment.

[0015] Thirdly, embodiments of the present invention also provide a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the pile foundation scour depth prediction model construction method based on multimodal parameters as described in any embodiment.

[0016] In summary, this invention provides a method for constructing a pile foundation scour depth prediction model based on multimodal parameters. By introducing structural multimodal parameters, a high-precision regression model is constructed, thereby significantly improving the accuracy of bridge scour depth prediction in complex environments and providing a reliable basis for bridge safety and durability assessment. Specifically, the method of this embodiment has the following advantages:

[0017] 1) This embodiment uses multimodal features such as frequency change rate, modal curvature, and diagonal compliance as input parameters, breaking through the limitations of traditional methods that rely on the frequency or displacement characteristics of different components. This allows for a more comprehensive characterization of the dynamic properties of the pile foundation scour process. Through the complementarity of multimodal parameters, it not only captures the dynamic characteristics changes during the scour process more comprehensively and accurately, achieving high-precision prediction of scour depth, but also effectively suppresses the influence of external interferences such as temperature changes and environmental noise on the prediction results, thereby significantly improving the stability and applicability of the method.

[0018] 2) The ANFIS model used in this embodiment integrates the self-learning and nonlinear mapping capabilities of artificial neural networks, as well as the advantages of fuzzy inference in handling uncertainty. It can effectively cope with the high nonlinearity and uncertainty problems in the pile foundation scour process in marine and river environments, thereby improving the applicability and robustness of the model. Attached Figure Description

[0019] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific 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 from these drawings without creative effort.

[0020] Figure 1 This is a flowchart of a method for constructing a pile foundation scour depth prediction model based on multimodal parameters, provided by an embodiment of the present invention.

[0021] Figure 2 This is a schematic diagram of the scour depth and measuring points of a pile foundation provided in an embodiment of the present invention;

[0022] Figure 3 This is a schematic diagram of the training process of an ANFIS (Adaptive Neuro-Fuzzy Inference System) model provided in an embodiment of the present invention;

[0023] Figure 4 This is a flowchart of an ANFIS model for predicting scour depth provided in an embodiment of the present invention;

[0024] Figure 5 This is a performance diagram of an ANFIS model provided in an embodiment of the present invention;

[0025] Figure 6 This is a comparison chart of different model indicators provided in an embodiment of the present invention;

[0026] Figure 7This is a comparison chart provided by an embodiment of the present invention for identifying scour depth based on curvature and curvature difference;

[0027] Figure 8 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0029] In the description of this invention, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing the invention and for simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the invention. Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0030] In the description of this invention, it should also be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0031] Figure 1 This is a flowchart illustrating a method for constructing a pile foundation scour depth prediction model based on multimodal parameters, provided by an embodiment of the present invention. This method is executed by electronic equipment. Figure 1 As shown, the method specifically includes:

[0032] S110. Obtain the acceleration and scour depth time history data of the pile foundation during the scour process.

[0033] The scour depth of a pile foundation refers to the depth of erosion caused by high flow rates and velocities in the riverbed surrounding the pile foundation. Figure 2As indicated by the red double-headed arrows in the diagram. Optionally, the original acceleration time history signal of the pile foundation and the corresponding scour depth data (i.e., the pile foundation acceleration and scour depth at each moment) can be collected through indoor water tank scour tests and field measurements, serving as the data source for the entire method.

[0034] S120. Based on the collected data, construct the pile foundation feature vector corresponding to each scour depth.

[0035] This feature vector includes multimodal features of the pile foundation, which are used as the basis for predicting scour depth.

[0036] Specifically, for the scour depth collected at each moment, the following steps are performed:

[0037] Step 1: Based on the acceleration time history data, identify the frequencies of the pile foundation before and after scour, as well as the mode shapes at multiple measuring points on the pile foundation. Optionally, the original acceleration signal is first preprocessed to fill in missing values ​​and remove outliers; then, ERA (Eigensystem Realization Algorithm) is applied to identify the natural frequencies (or simply frequencies) and mode shapes of the structure (i.e., the pile foundation). Specifically, the pile foundation structure includes multiple frequencies and multiple mode shapes. In this embodiment, unless otherwise specified, relatively accurate first-order frequencies and first-order mode shapes are used for calculation.

[0038] Step 2: Calculate the rate of change of frequency based on the frequencies before and after scouring; calculate the modal curvature and diagonal compliance of each measuring point based on the mode shape of each measuring point.

[0039] Specifically, localized scour weakens foundation constraints, leading to a decrease in structural stiffness and significant changes in structural dynamic parameters. This stiffness change directly causes a change in the structure's natural frequency. This embodiment uses the rate of change of frequency before and after scour to characterize the stiffness change of the pile foundation, serving as a feature for predicting scour depth. Optionally, the rate of change of frequency... The calculation formula is as follows:

[0040] (1)

[0041] In the formula, f s f is the frequency of the pile foundation after scouring; u The frequency at which the pile foundation is scoured.

[0042] At the same time, when the scour depth changes, the measuring points on the pile foundation (such as...) Figure 2The lateral displacement (shown by the red dot in the diagram) will also change, causing changes in the morphology of the pile foundation. In particular, the pile foundation will exhibit a significant curvature change at the interface between the water body and the riverbed. Therefore, this embodiment characterizes the morphology of the pile foundation by measuring the curvature at each measuring point, using curvature as a feature for predicting scour depth. Optionally, the modal curvature of measuring point i... :

[0043] (2)

[0044] In the formula, , , These are the vibration modes after scouring at measuring points i-1, i, and i+1, respectively. i-1 and l i Let the measuring points i-1 and i be located along the pile foundation. This represents the distance between adjacent points along the pile foundation. The modal curvature of each measuring point at each scour depth constitutes a curvature vector.

[0045] Furthermore, scour depth can also be considered as a form of damage to the stiffness of the pile foundation. Modal compliance, in a physical sense, corresponds to the reciprocal of the pile foundation stiffness. Therefore, in this embodiment, the compliance vector formed by the modal compliance values ​​at each measuring point is selected as a feature for predicting scour depth. Optionally, the modal compliance matrix F of the pile foundation is calculated according to the following formula:

[0046] (3)

[0047] In the formula, Ф n This represents the nth mode shape of the pile foundation, where N indicates that there are a total of N mode shapes. , This represents the nth mode shape at measurement point i. , Indicates the total number of measuring points; f ns This represents the nth-order frequency after scouring. The diagonal elements of F are taken as the diagonal compliance at each measurement point:

[0048] (4)

[0049] In the formula, f ii For measuring points diagonal softness, This indicates that the diagonal elements are taken, and the diagonal compliance of each measurement point constitutes a one-dimensional feature vector.

[0050] Step 3: Perform principal component analysis on the modal curvature and diagonal compliance at each measuring point, retaining the first few principal components as curvature features, which, together with the frequency change rate (FCR), form a feature vector. Specifically, due to the large number of measuring points, this embodiment performs feature dimensionality reduction on high-dimensional curvature and diagonal compliance to reduce the computational load in subsequent model construction and prediction. Optionally, the curvature feature vector and diagonal compliance feature vector at the same scour depth can be concatenated into a single vector, and principal component analysis is performed on this vector. The first three principal components (those contributing approximately 70% of the cumulative data) are retained as curvature features, which, together with the frequency change rate (FCR), constitute the final 4-dimensional feature vector X=[x1, x2, x3, x4]. T .

[0051] S130. Train the ANFIS model using each scour depth and its corresponding feature vector. The trained model is used to automatically predict the corresponding scour depth based on a feature vector.

[0052] This step trains the ANFIS model using the feature vectors and scour depths at the same time as samples. The trained model takes the feature vector of the pile foundation at a certain time as input and the scour depth of the pile foundation at that time as output.

[0053] In one specific implementation, the sample data is first normalized to eliminate dimensional differences between different features. Since the input feature scales differ significantly, the Z-score normalization method can be used to eliminate these dimensional differences. In other words, the input feature set data is converted to a standard normal distribution with zero mean and unit variance to accelerate the convergence process of gradient descent. The formula for Z-score normalization is shown below:

[0054] (5)

[0055] In the formula, Z represents the normalized input feature, X represents the original 4-dimensional input feature value, and μ and σ represent the mean and standard deviation of the input feature, respectively.

[0056] Next, the dataset is partitioned and the model is trained. The normalized input feature set Z=[z1, z2, z3, z4] T The dataset is divided into training, validation, and test sets according to a certain ratio. In this embodiment, the training, validation, and test sets account for 85%, 5%, and 10% of the original dataset, respectively. The training and validation sets are used for model training, and the test set is used for model performance evaluation.

[0057] Optionally, the model training process is as follows: Figure 2 As shown, the specific steps include the following:

[0058] Step 1: Define the fuzzy evaluation set of the ANFIS model, which includes multiple scour depth levels. Optionally, each feature in the feature vector (4 features in total) contains m fuzzy evaluation subsets, where m=3, corresponding to the low, medium, and high scour depth levels, respectively.

[0059] Step 2: Based on each scour depth level and the features in the feature vector, define multiple rules for the ANFIS model. As mentioned above, the total number of rules M in the ANFIS model is m. 4 One, rule The antecedent is ( , , , The consequent is a first-order Sugeno:

[0060] (6)

[0061] In the formula, p l , q l , r l , s l , q l All are consequent parameters. For rules The calculated local scour depth.

[0062] Step 3: Calculate the strength of each rule triggered by a given feature vector. Optionally, first, for each input parameter z... j With each fuzzy set Calculate membership degrees (low, medium, high) as follows:

[0063] (7)

[0064] In the formula, It is a fuzzy set The membership degree determines whether the input parameter Z can satisfy The degree; This represents the membership function.

[0065] Then, calculate the rule trigger strength for each node (i.e., each rule). The node parameters are all fixed and do not need to be adjusted through network learning. The calculation formula is as follows:

[0066] (8)

[0067] Calculate each The node is used to obtain the first node. The normalized rule trigger strength of each node is obtained by taking the ratio of the trigger strength of this rule to the sum of the trigger strengths of all rules. :

[0068] (9)

[0069] Step 4: Based on the consequent parameters in the ANFIS model, calculate the prediction depth of the given feature vector under each rule, and then perform a weighted average of the prediction depths under each rule according to the trigger strength of each rule to obtain the final prediction depth of the given feature. Optionally, firstly, the actual output value of each rule is represented by the product of the local prediction depth and the trigger strength of the normalization rule. Based on the output results of the previous step, the actual output value of each rule can be obtained. :

[0070] (10)

[0071] Then, the actual output values ​​of the above rules are summed to obtain the final prediction result. :

[0072] (11)

[0073] Step 5: Update the ANFIS model parameters based on the difference between the final predicted depth and the scour depth corresponding to a certain feature. The model parameters are those in formula (10). , , , and The entire model training process described above is as follows: Figure 3 As shown. After the model is trained, a new feature vector X is input into the model, and the model can automatically predict the scour depth corresponding to that feature vector. The whole process is as follows Figure 4 As shown.

[0074] Furthermore, early stopping can be employed during model training. Early stopping determines whether to terminate training prematurely by monitoring the loss on the validation set. If the loss on the validation set does not decrease within a consecutive set of iterations, training is stopped, effectively preventing overfitting and improving the model's generalization ability. Therefore, processing the data and training the model according to the above steps, while simultaneously applying early stopping, can significantly improve the model's generalization ability.

[0075] At the same time, evaluation indicators are used to assess the optimal model, including but not limited to the coefficient of determination (R²). 2 ), root mean square error (RMSE), and mean absolute error (MAE). 2Reflecting the ANFIS model's fit to the overall trend, RMSE emphasizes the penalty for errors, while MAE measures the magnitude of the average error. In this embodiment, the overall prediction results from three-fold cross-validation are used to calculate the above indicators to obtain the comprehensive performance of the ANFIS model under different parameter combinations. The optimal ANFIS model's prediction results for each indicator are as follows: R 2 The accuracy is 0.93, RMSE is 752.65 mm, and MAE is 492.44 mm, indicating that the ANFIS model (where scour depth and frequency have a non-linear relationship) has high accuracy. The test set predicted values ​​and actual values ​​are compared as follows: Figure 5 As can be observed in the figure, the predicted values ​​and the actual values ​​match well, with most data points falling within the 95% prediction range.

[0076] To further verify the superior performance of the ANFIS model, it was compared with three other regression models: Artificial Neural Network (ANN), Gradient Boosting Tree (GXBoost), and Random Forest (RF). The results are shown in Table 1 and... Figure 6 As shown in the figure, the results indicate that the ANFIS model outperforms other models in all three evaluation metrics, with a test set correlation coefficient of 0.95, and RMSE and MAE of 752.65 mm and 492.44 mm, respectively. The RF and GXBoost models follow, while the ANN model performs the worst.

[0077] Table 1. Performance Comparison of Different Models

[0078]

[0079] In summary, this embodiment provides a method for constructing a pile foundation scour depth prediction model based on multimodal parameters. By introducing structural multimodal parameters, a high-precision regression model is constructed, thereby significantly improving the accuracy of bridge scour depth prediction in complex environments and providing a reliable basis for bridge safety and durability assessment. Specifically, the method of this embodiment has the following advantages:

[0080] 1) This embodiment uses multimodal features such as frequency change rate, modal curvature, and diagonal compliance as input parameters, breaking through the limitations of traditional methods that rely on the frequency or displacement characteristics of different components. This allows for a more comprehensive characterization of the dynamic properties of the pile foundation scour process. Through the complementarity of multimodal parameters, it not only captures the dynamic characteristics changes during the scour process more comprehensively and accurately, achieving high-precision prediction of scour depth, but also effectively suppresses the influence of external interferences such as temperature changes and environmental noise on the prediction results, thereby significantly improving the stability and applicability of the method.

[0081] 2) The ANFIS model used in this embodiment integrates the self-learning and nonlinear mapping capabilities of artificial neural networks, as well as the advantages of fuzzy inference in handling uncertainty. It can effectively cope with the high nonlinearity and uncertainty problems in the pile foundation scour process in marine and river environments, thereby improving the applicability and robustness of the model.

[0082] 3) The ANFIS model is particularly suitable for scour depth prediction in this embodiment. First, there is a complex nonlinear relationship between the rate of change of frequency, modal curvature, curvature difference, and scour depth, and ANFIS excels at capturing this complex relationship. Through fuzzy inference rules and the adaptive capabilities of neural networks, ANFIS can predict scour depth based on changes in these input features.

[0083] While Artificial Neural Networks (ANNs) and XGBoost are powerful at fitting complex nonlinear relationships, their "black box" nature makes it difficult to achieve transparency in how input features affect model predictions. In contrast, ANFIS combines the learning power of neural networks with the interpretability of fuzzy inference systems. By employing IF-THEN fuzzy rules and membership functions, it establishes a mapping between input and output, providing not only powerful nonlinear fitting capabilities but also the ability to express the relationship between modal parameters and scour depth in rule form, offering a degree of engineering interpretability.

[0084] Furthermore, the ANFIS model achieves high prediction accuracy while maintaining engineering interpretability. This embodiment systematically compares the prediction results of ANFIS, ANN, XGBoost, and Random Forest (RF) models using the same dataset. The results show that ANFIS achieves the best performance (R² = 0.95), demonstrating that this model can effectively predict scour depth when multimodal parameters are involved.

[0085] Furthermore, the aforementioned feature vector uses three pile foundation characteristics—rate of change of frequency, modal curvature, and diagonal compliance—as the basis for predicting the scour depth of the pile foundation. These three pile foundation characteristics are not arbitrarily determined, but rather pre-selected from numerous characteristics of the pile foundation through a specific method. In a specific implementation, this specific method includes the following steps:

[0086] Step 1: Determine the multiple pile foundation attributes affected by scour depth, and then determine the pile foundation characteristics under each attribute. Specifically, this embodiment first determines the pile foundation attributes affected by scour depth, including pile foundation stiffness and pile foundation morphology, the specific influence principles of which are explained in S120. Then, the pile foundation characteristics under the pile foundation stiffness attribute are determined, including the frequency change rate characteristic; the pile foundation characteristics under the pile foundation morphology attribute are also determined, including modal curvature characteristics, modal curvature change rate characteristics, and modal execution criterion characteristics. The calculation formula for the modal execution criterion (MAC) is as follows:

[0087] (12)

[0088] In the formula, This indicates the vibration mode of the pile foundation after scouring. It is the vibration mode of the pile foundation before scour. The closer the MAC is to 1, the less likely there is scour. Generally, if it is less than 0.9 or 0.8, it is considered that there is a probability that scour has occurred.

[0089] Step Two: Expand the set of pile foundation features under each pile foundation attribute to ensure that each pile foundation attribute includes at least one pile foundation feature unaffected by environmental interference. Specifically, among the pile foundation features identified in Step One, the rate of change of frequency is significantly affected by temperature, while the other features mainly originate from lateral displacement and are minimally affected by temperature. Therefore, this step expands the set of pile foundation features under the pile foundation stiffness attribute by adding the diagonal compliance feature, which is unaffected by temperature. This feature can be calculated from the displacement of the measuring point and also corresponds to the reciprocal of the pile foundation stiffness, thus simultaneously satisfying the requirement of reflecting the pile foundation stiffness attribute without being affected by temperature.

[0090] Step 3: Verify the sensitivity of each feature in the expanded pile foundation feature set to scour depth, and remove insensitive features. Specifically, after the expansion in Step 2, a complete comparison surface and a stable feature set have been obtained. This step, based on the dataset collected in S110, verifies the sensitivity of each individual feature in the feature set to scour depth, and removes those features that are insensitive to scour depth, including the modal execution criterion features.

[0091] Step 4: Extract multiple feature combinations with replacement from the remaining feature set, and train the ANFIS model based on the feature vectors formed by each feature combination. Select the optimal feature combination that yields the best model performance. In this step, multiple features (2, 3, or 4) are randomly extracted from the remaining feature set to construct the input feature vector of the ANFIS model. Low-dimensional features are directly used as elements of the feature vector, while high-dimensional features are reduced in dimensionality according to the method described in the previous embodiment and then used as elements of the feature vector. Using the feature vectors constructed from each feature combination and the scour depth data, train the ANFIS model separately, and select the combination with the best performance as the optimal combination.

[0092] Step 5: If multiple features in the preferred feature combination have similar relationships with scour depth, retain only one of the multiple features. This step continues to analyze the relationship between each individual feature in the preferred combination and scour depth, finding that the influence of scour depth on modal curvature features is very similar to the influence of scour depth on modal curvature change rate (i.e., curvature difference) features, such as... Figure 7 As shown, curvature exhibits a unidirectional abrupt change at the scour depth, while the curvature difference shows a bidirectional abrupt change at the scour depth, but the abrupt change on the right side primarily reflects the scour depth. Therefore, only one of these two features is retained: the modal curvature feature is preserved, while the modal curvature variation feature is removed.

[0093] Step Six: Incorporate the multiple features retained from the preferred feature combination into the model feature vector used to predict scour depth. After removing the modal curvature variation feature in Step Five, the frequency change rate, modal curvature, and diagonal compliance are ultimately retained as the basis for constructing the scour depth prediction model feature vector in this embodiment.

[0094] This embodiment provides a method for selecting features for predicting scour depth. This method combines theoretical analysis with data analysis, enabling the selection of effective features that match the data characteristics from numerous features of pile foundations for subsequent prediction. This method is applicable to all original feature sets and sample datasets. In practical applications, when the available set of foundation pile features changes, this method can still be used to re-select several features that can effectively predict scour depth from the new feature set, constructing a new input vector and a new ANFIS model, thereby achieving accurate and stable prediction of pile foundation scour depth.

[0095] It should be noted that all data involved in this application are information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with the relevant laws, regulations and standards of the relevant countries and regions, and corresponding operation portals are provided for users to choose to authorize or refuse.

[0096] Figure 8This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention, such as... Figure 8 As shown, the device includes a processor 60, a memory 61, an input device 62, and an output device 63; the number of processors 60 in the device can be one or more. Figure 8 Taking a processor 60 as an example; the processor 60, memory 61, input device 62, and output device 63 in the device can be connected via a bus or other means. Figure 8 Taking the example of a connection between China and Israel via a bus.

[0097] The memory 61, as a computer-readable storage medium, can be used to store software programs, computer-executable programs, and modules, such as the program instructions / modules corresponding to the pile foundation scour depth prediction model construction method based on multimodal parameters in this embodiment of the invention. The processor 60 executes various functional applications and data processing of the device by running the software programs, instructions, and modules stored in the memory 61, thereby realizing the aforementioned pile foundation scour depth prediction model construction method based on multimodal parameters.

[0098] The memory 61 may primarily include a program storage area and a data storage area. The program storage area may store the operating system and at least one application program required for a given function; the data storage area may store data created based on terminal usage. Furthermore, the memory 61 may include high-speed random access memory and non-volatile memory, such as at least one disk storage device, flash memory, or other non-volatile solid-state storage device. In some instances, the memory 61 may further include memory remotely located relative to the processor 60, which can be connected to the device via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0099] Input device 62 can be used to receive input digital or character information, and to generate key signal inputs related to user settings and function control of the device. Output device 63 may include display devices such as a display screen.

[0100] This invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the pile foundation scour depth prediction model construction method based on multimodal parameters of any embodiment.

[0101] The computer storage medium of this invention can be any combination of one or more computer-readable media. A computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. A computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of computer-readable storage media (a non-exhaustive list) include: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this document, a computer-readable storage medium can be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0102] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, capable of sending, propagating, or transmitting programs for use by or in connection with an instruction execution system, apparatus, or device.

[0103] Program code contained on a computer-readable medium may be transmitted using any suitable medium, including but not limited to wireless, wire, optical fiber, RF, etc., or any suitable combination thereof.

[0104] Computer program code for performing the operations of this invention can be written in one or more programming languages ​​or a combination thereof. Programming languages ​​include object-oriented programming languages—such as Java, Smalltalk, and C++—as well as conventional procedural programming languages—such as C or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0105] 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 technical solutions of the embodiments of the present invention.

Claims

1. A method for constructing a pile foundation scour depth prediction model based on multimodal parameters, characterized in that, include: Obtain time history data of acceleration and scour depth of pile foundation during the scour process; For each scour depth, the following operations are performed: Based on the acceleration time history data, the frequencies before and after scour of the pile foundation, as well as the mode shapes of multiple measuring points on the pile foundation, are identified; based on the frequencies before and after scour, the frequency change rate is calculated; based on the mode shapes of each measuring point, the modal curvature and diagonal compliance of each measuring point are calculated; principal component analysis is performed on the modal curvature and diagonal compliance of each measuring point, and the first few levels of principal components are retained as curvature features, which, together with the frequency change rate, form a feature vector; An ANFIS model is trained using each scour depth and its corresponding feature vector. The trained model is then used to automatically predict the corresponding scour depth based on a feature vector.

2. The method according to claim 1, characterized in that, The calculation of the frequency change rate based on the frequencies before and after the scouring includes: Calculate the rate of change of frequency using the following formula. : ， In the formula, f s f is the frequency after scouring of the pile foundation; u The frequency before scouring of the pile foundation.

3. The method according to claim 1, characterized in that, The calculation of modal curvature at each measuring point based on the mode shape includes: Calculate the modal curvature of measuring point i using the following formula. : ， In the formula, , , These are the vibration modes after scouring at measuring points i-1, i, and i+1, respectively. i-1 and l i The positions of measuring points i-1 and i along the pile foundation are given.

4. The method according to claim 1, characterized in that, The calculation of modal curvature and diagonal compliance at each measuring point based on the mode shape includes: Calculate the modal compliance matrix F of the pile foundation using the following formula: ， In the formula, Ф n This represents the nth mode shape of the pile foundation, where N indicates that there are a total of N mode shapes. , This represents the nth mode shape at measurement point i. , Indicates the total number of measuring points; f ns This represents the nth order frequency after scouring of the pile foundation; Take each diagonal element in F as the diagonal compliance of each measurement point.

5. The method according to claim 1, characterized in that, The method of training the ANFIS model using each scour depth and its corresponding feature vector includes: Define a fuzzy evaluation set for the ANFIS model, wherein the fuzzy evaluation set includes multiple scour depth levels; Based on each scour depth level and the features in the feature vector, multiple rules of the ANFIS model are defined. Calculate the strength of each rule triggered by a given feature vector; Based on the training parameters of the ANFIS model, the prediction depth of a certain feature vector under each rule is calculated, and the prediction depth under each rule is weighted and averaged according to the triggering intensity of each rule to obtain the final prediction depth of the certain feature. The training parameters are updated based on the difference between the final predicted depth and the actual scour depth corresponding to a certain feature vector.

6. The method according to claim 1, characterized in that, Before performing the following operations for each scouring depth, the procedure further includes: Determine the multiple pile foundation properties affected by scour depth, and determine the pile foundation characteristics under each property separately; The characteristics of pile foundations under each pile foundation attribute are expanded to ensure that each pile foundation attribute includes at least one pile foundation characteristic that is not affected by the environment. The sensitivity of each expanded pile foundation feature to scour depth was verified separately, and non-sensitive pile foundation features were removed. Multiple feature combinations are extracted with replacement from the remaining set of pile foundation features, and the ANFIS model is trained based on the feature vectors formed by each feature combination. The optimal feature combination with the best model performance is then selected. If multiple pile foundation features in the preferred combination of features have similar relationships with scour depth, only one of the multiple pile foundation features shall be retained. The multiple pile foundation features that remain in the final combination of the preferred features are incorporated into the model feature vector used to predict the scour depth.

7. The method according to claim 6, characterized in that, The determination of multiple pile foundation attributes affected by scour depth, and the determination of pile foundation characteristics under each attribute, includes: Determine the impact of scour depth on the stiffness and morphological properties of the pile foundation; Determine the frequency variation rate characteristics of pile foundations under the stiffness attribute, and the modal curvature characteristics, modal curvature variation rate characteristics, and modal execution criterion characteristics of pile foundations under the morphological attribute.

8. The method according to claim 7, characterized in that, The expansion of pile foundation characteristics under various pile foundation attributes ensures that each pile foundation attribute includes at least one pile foundation characteristic that is not affected by the environment, including: based on the characteristic that the frequency change rate is affected by temperature, expanding the pile foundation stiffness attribute to include a diagonal compliance characteristic that is not affected by temperature. The removal of non-sensitive pile foundation features includes: removing modal execution criterion features that are insensitive to scour depth; If multiple pile foundation features in the preferred feature combination have similar influence patterns on scour depth, only one of the multiple pile foundation features shall be retained, including: in the preferred feature combination, removing the modal curvature variation feature from the modal curvature features and modal curvature change rate features that have similar influence patterns on scour depth, and retaining the modal curvature feature.

9. An electronic device, characterized in that, include: One or more processors; Memory, used to store one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the method for constructing a pile foundation scour depth prediction model based on multimodal parameters as described in any one of claims 1-8.

10. A computer-readable storage medium, characterized in that, It stores a computer program that, when executed by a processor, implements the method for constructing a pile foundation scour depth prediction model based on multimodal parameters as described in any one of claims 1-8.

Images