Seamed arch dam displacement interval prediction method based on separation modeling and fractal theory

Through separation modeling and fractal theory, combined with deep learning model and Bootstrap method, the displacement sequence is decomposed and multi-fractal analysis is solved, and the problem of difficulty in evaluating change trends and interval predictions in the existing technology is solved, achieving more accurate and reliable displacement prediction.

CN119004946APending Publication Date: 2024-11-22YANGZHOU UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410957602.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-17
Publication Date
2024-11-22

AI Technical Summary

Technical Problem

Existing displacement prediction methods are difficult to effectively evaluate the changing trend of dam displacement, and there are challenges in optimizing from single point value prediction to interval prediction.

Method used

Using the method based on separation modeling and fractal theory, the displacement sequence is STL decomposed and decomposed into trend components, seasonal components and residual terms. Through multi-fractal analysis, combined with crack opening degree as an influencing factor, a deep learning model CNN-LSTM is established. Make predictions and finally use the Bootstrap method to make interval predictions.

Benefits of technology

It improves the accuracy and reliability of displacement prediction, can better handle the nonlinear characteristics and uncertainties of displacement, and provides richer information and high confidence prediction intervals.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119004946A_ABST
    Figure CN119004946A_ABST
Patent Text Reader

Abstract

The invention discloses a slit arch dam displacement interval prediction method based on separation modeling and a fractal theory, and relates to the technical field of concrete dam safety monitoring, and the method comprises the steps: carrying out the STL decomposition of a displacement sequence, decomposing the displacement into a trend component, a season component and a remainder, and carrying out the multi-fractal analysis of the displacement and the change trend of the components; an initial modeling factor set is obtained based on an HST model, the fracture opening degree serves as an influence factor of displacement, a displacement monitoring model considering fracture influence is established, a primary selection factor set is obtained, a variable selection method mRMR is adopted for optimizing the influence factor of displacement, the remaining modeling factors are combined into a screened factor set, and normalization processing is carried out; a deep learning model CNN-LSTM is established to predict a trend component and a seasonal component of displacement and then superpose the trend component and the seasonal component to obtain superposed displacement, interval prediction is performed based on a Bootstrap method, evaluation indexes are calculated, and performance of each model prediction interval is compared.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the technical field of concrete dam safety monitoring, and in particular to a displacement interval prediction method for a seamed arch dam based on separation modeling and fractal theory. Background Art

[0002] During the operation cycle, the arch dam will be subjected to the long-term effects of various environmental loads such as hydrostatic pressure and temperature. In order to timely grasp the operating status of the arch dam, a large number of instruments are buried inside it to monitor variables such as displacement, cracks, stress, strain and seepage. Among them, displacement is the most intuitive indicator to measure the overall operating status of the arch dam. In long-term operating concrete arch dams, the emergence of large-scale cracks is not accidental. These deep cracks will greatly reduce the overall bearing capacity of the dam body, and are more likely to deform when stressed, which will lead to a significant increase in the displacement of the dam body. Therefore, it is necessary to establish a reasonable and high-precision displacement model of the arch dam with seams.

[0003] Displacement monitoring data usually exhibits non-stationarity and is often accompanied by noise interference. Therefore, it is difficult to accurately extract valuable information from noisy data by relying on a single algorithm for analysis. In previous displacement prediction studies, static fitting of the displacement of the measuring point is usually performed, which makes it difficult to effectively evaluate the changing trend of the dam displacement. Therefore, it is necessary to combine denoising technology and fractal theory to deeply explore the overall pattern of arch dam displacement changes. In the existing arch dam displacement monitoring model research, scholars generally use mathematical models to obtain point value estimates of the displacement of the measuring point, and pay less attention to the uncertainty and random factors in the displacement change. Compared with a single point value prediction, interval prediction can provide richer information and show its inherent uncertainty through the prediction interval. Summary of the invention

[0004] In view of the above-mentioned problems, the present invention is proposed.

[0005] Therefore, the technical problem solved by the present invention is that it is difficult to effectively evaluate the changing trend of dam displacement in the existing displacement prediction method and how to optimize the single point value prediction to interval prediction.

[0006] To solve the above technical problems, the present invention provides the following technical solutions: a method for predicting the displacement interval of a seamed arch dam based on separation modeling and fractal theory, comprising performing STL decomposition on a displacement sequence, decomposing the displacement into a trend component, a seasonal component and a remainder, and performing multi-fractal analysis on the displacement and component change trends; obtaining an initial set of modeling factors based on an HST model, taking the crack opening and closing degree as an influencing factor of the displacement, establishing a displacement monitoring model considering the influence of cracks, obtaining a preliminary factor set, adopting a variable selection method mRMR to optimize the influencing factors of the displacement, and forming the remaining modeling factors into a screened factor set for normalization; establishing a deep learning model CNN-LSTM to predict the trend component and the seasonal component of the displacement and then superimpose them to obtain the superimposed displacement, performing interval prediction based on the Bootstrap method, calculating evaluation indicators, and comparing the performance of the prediction intervals of each model.

[0007] As a preferred solution of the displacement interval prediction method of the seamed arch dam based on separation modeling and fractal theory described in the present invention, wherein: the STL decomposition of the displacement sequence to decompose the displacement into trend components, seasonal components and residuals includes STL decomposing the displacement time series Yt into trend components, seasonal components and residuals through inner and outer loops, assuming T t k is the seasonal component and trend component of the kth iteration, then the k+1th iteration process includes calculating the detrended sequence Y t -T t k , perform Loess smoothing on the subsequence of the detrended sequence to obtain a temporary seasonal sequence Pair Sequence Applying a low-pass filter and the Loess process yields The seasonal component of the k+1th iteration is The original series removes the seasonal component The Loess smoothing is used to process the sequence of the deseasonalized component, and the trend component of the k+1 iteration is T t k+1 .

[0008] In the outer loop, the seasonal component and trend component obtained after the inner loop are used to calculate the remainder It is expressed as:

[0009]

[0010] As a preferred solution of the displacement interval prediction method of the seamed arch dam based on separation modeling and fractal theory described in the present invention, wherein: the multifractal analysis of the displacement and component change trends includes calculating the displacement trend component, seasonal component and the Hurst index, Renyi index and multifractal spectrum of the displacement sequence itself using the MF-DFA method based on fractal theory, performing multifractal analysis on the displacement and its component change trends, and given a displacement time series x(t), the sequence mean is μx, and the cumulative deviation sequence y(t) of x(t) and the mean is expressed as:

[0011]

[0012] By using a given time scale s, y(t) is divided equally into subintervals y v (t), and perform trend fitting, and subtract the fitting part from the original basis Get the residual sequence z v (t), expressed as:

[0013]

[0014] Calculate the residual sequence z v (t) and the mean square error F2(s,v) and the q-order fluctuation function F q (s), expressed as:

[0015]

[0016] If there is long-range correlation in the displacement time series, then F q The relationship between (s) and the time scale s is expressed as:

[0017] IgF q (s) = h (q) lgs + lgb

[0018] Among them, F q (s) is the q-order wave function of the displacement sequence, h(q) is the Hurst exponent, and b is a constant coefficient;

[0019] The fractal strength and fractal singularity of the displacement sequence are characterized by the multifractal spectrum f(α), which is expressed as:

[0020] τ(q)=qh(q)-1

[0021] α=dτ(q) / dq

[0022] f(α)=qα-τ(q)

[0023] Among them, τ(q) is the Renyi index, α is the singular intensity, and f(α) is the multifractal spectrum.

[0024] As a preferred solution of the displacement interval prediction method of the seamed arch dam based on separation modeling and fractal theory described in the present invention, wherein: the variable selection method mRMR is used to optimize the influencing factors of the displacement, the remaining modeling factors are formed into a screened factor set, and the normalization processing includes using the variable selection method mRMR to optimize the influencing factors of the displacement, deleting the modeling factors that do not meet the threshold requirements after calculation by the mRMR method, and forming the remaining modeling factors into a screened factor set. The factor elimination of the variable selection method mRMR includes mRMR using the mutual information method to measure the dependency between two variables, which is expressed as:

[0025]

[0026] Among them, x i is the influencing factor of displacement, y is the measured displacement, p(x i ) and p(y) represent x i and the marginal probability density of y, p(x i ,y) represents their joint probability density.

[0027] The maximum correlation D is expressed as the average value of mutual information, and the minimum redundancy R is expressed as x i and x j The minimum correlation between them is obtained, and the operator φ is defined to jointly operate D and R, which can be expressed as:

[0028]

[0029] Where F is x i is a feature set of F, where |F| is equal to the number of features in F.

[0030] The value of φ is calculated for each factor in turn, and the influencing factors corresponding to the value less than 1 are eliminated to form a candidate feature set, which is used as the input of the subsequent prediction model.

[0031] As a preferred solution of the displacement interval prediction method of the seamed arch dam based on separation modeling and fractal theory described in the present invention, wherein: the deep learning model CNN-LSTM is established to predict the trend component and the seasonal component of the displacement and then superimpose them to obtain the superimposed displacement, including taking the daily value corresponding to the displacement influencing factor as the input of the CNN-LSTM model, and the daily measured displacement as the output of the CNN-LSTM model. At the same time node, the measured data of each influencing factor and the displacement are divided according to the ratio of 8:2, respectively as a training set and a test set, the training set is used for training the corresponding CNN-LSTM model, and the test set is used to evaluate the prediction accuracy of the corresponding CNN-LSTM model;

[0032] The accuracy of the CNN-LSTM model is verified by calculating the root mean square error RMSE, mean absolute error MAE and determination coefficient R2 of the measured and predicted values ​​in the test set. The calculation of the three indicators is expressed as:

[0033]

[0034] Among them, n represents the length of the time series, y i and are monitoring data and predicted values, respectively. and are the mean of the observed data series and the mean of the predicted data series, respectively.

[0035] As a preferred solution of the displacement interval prediction method of the seamed arch dam based on separation modeling and fractal theory described in the present invention, the interval prediction based on the Bootstrap method includes a given displacement and an influencing factor, which is expressed as:

[0036] (x s ,y s )(s=1,2,…,t)

[0037] Among them, t is the time length of the input sequence, x s is the set of factors affecting the dam displacement, y s is the displacement value;

[0038] By resampling the displacement data training set B times using the Bootstrap method, we get B sets of training sets. It is expressed as:

[0039]

[0040] in, is the predicted value output by the lth model,

[0041] Variance of model uncertainty It is expressed as:

[0042]

[0043] Keep input x s and replace the displacement sequence with Get the transformed data set D ξ , expressed as:

[0044]

[0045] Right D ξ Resample to obtain the estimate of noise variance and the variance of uncertainty in the noise regression model The variance of the data noise is obtained by superposition, which is expressed as:

[0046]

[0047] Assume the estimated error and data noise ξ(x s ) are statistically independent, and the variance of the total prediction error It is expressed as:

[0048]

[0049] in, and are the variance of model uncertainty and data noise, respectively.

[0050] As a preferred solution of the displacement interval prediction method of the seamed arch dam based on separation modeling and fractal theory described in the present invention, the interval prediction based on the Bootstrap method also includes when the significance level is set to α, (1-α)×100% is the nominal confidence of the prediction interval, then the displacement y s Prediction interval at time step s It is expressed as:

[0051]

[0052] in, and are the lower and upper bounds of the constructed prediction interval, is the output of the prediction model, is the variance of the total forecast error, z 1-α / 2 is the 1-α / 2 percentile of the standard normal distribution. In order to evaluate the overall quality of the constructed interval, the prediction interval coverage probability PICP, the average prediction interval width MPIW and the coverage and width criterion CWC are introduced to measure, which is expressed as:

[0053]

[0054]

[0055] CWC=MPIW[1+γ(PICP)e -η(PICP-PINC) ]

[0056]

[0057] Among them, N t is the number of samples. When the displacement value is within the interval, c s The value of is 1, otherwise it is 0, and η is the penalty parameter, which is set to 50.

[0058] Another object of the present invention is to provide a displacement interval prediction system for a seamed arch dam based on separation modeling and fractal theory, which can fully consider the influence of cracks on the displacement of the arch dam, add the crack opening and closing degree to the influencing factors of the displacement, construct a displacement monitoring model HSCT considering the influence of cracks, and use mRMR to perform feature selection on the displacement influencing factors of the seamed arch dam. After eliminating irrelevant variables, the optimal set of displacement influencing factors is obtained, which solves the problem that the current displacement prediction method contains the difficulty in predicting the displacement trend of the dam and the large amount of data processing.

[0059] As a preferred solution of the displacement interval prediction system of the arch dam with seams based on separation modeling and fractal theory described in the present invention, it includes: a sequence decomposition module, a factor screening module, and a model search and training module; the path planning module is used to perform STL decomposition on the displacement sequence, decompose the displacement into trend components, seasonal components and remainders, and perform multi-fractal analysis on the displacement and component change trends; the factor screening module is used to obtain an initial modeling factor set based on the HST model, take the crack opening and closing degree as the influencing factor of the displacement, establish a displacement monitoring model considering the influence of cracks, obtain a preliminary factor set, use the variable selection method mRMR to optimize the influencing factors of the displacement, and form the remaining modeling factors into a screened factor set for normalization; the model search and training module is used to establish a deep learning model CNN-LSTM to predict the trend component and seasonal component of the displacement and then superimpose them to obtain the superimposed displacement, perform interval prediction based on the Bootstrap method, calculate evaluation indicators, and compare the performance of the prediction interval of each model.

[0060] A computer device includes a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement a method for predicting displacement intervals of a seamed arch dam based on separation modeling and fractal theory.

[0061] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of a displacement interval prediction method for a seamed arch dam based on separation modeling and fractal theory.

[0062] Beneficial effects of the present invention: The present invention provides a displacement interval prediction method for a seamed arch dam based on separation modeling and fractal theory. The present invention provides a displacement interval prediction method for a seamed arch dam based on separation modeling and fractal theory. In order to eliminate the influence of noise, STL is used to decompose the displacement to obtain trend components, seasonal components and remainders. Through separation modeling, the nonlinear characteristics of the displacement can be better handled. In order to identify the nonlinear evolution law of the displacement, MF-DFA is used to analyze the multiple fractal characteristics of the displacement. Secondly, the influence of cracks on the displacement of the arch dam is fully considered, the crack opening and closing degree is added to the influencing factors of the displacement, and a displacement monitoring model HSCT considering the influence of cracks is constructed. The mRMR is used to select the features of the displacement influencing factors of the seamed arch dam. After eliminating irrelevant variables, the optimal set of displacement influencing factors is obtained. Finally, in order to consider the uncertainty of the displacement, interval prediction is performed based on the Bootstrap method and the CNN-LSTM model, which improves the reliability of the displacement prediction. The present invention achieves better results in terms of stability and accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0064] Figure 1 An overall flow chart of a method for predicting displacement intervals of a seamed arch dam based on separation modeling and fractal theory is provided for the first embodiment of the present invention.

[0065] Figure 2 Another overall flow chart of a method for predicting displacement intervals of a seamed arch dam based on separation modeling and fractal theory provided for the first embodiment of the present invention.

[0066] Figure 3 A diagram of the arrangement of vertical monitoring points for an arch dam according to a displacement interval prediction method for a seamed arch dam based on separation modeling and fractal theory is provided in accordance with a second embodiment of the present invention.

[0067] Figure 4 A schematic diagram of components after STL decomposition of two selected measuring points of a method for predicting displacement intervals of a jointed arch dam based on separation modeling and fractal theory provided in the second embodiment of the present invention.

[0068] Figure 5 A schematic diagram of the Hurst index, Renyi index and fractal spectrum of two selected measuring points after MF-DFA calculation for a method for predicting displacement intervals of a jointed arch dam based on separation modeling and fractal theory provided in the second embodiment of the present invention.

[0069] Figure 6 A multi-model prediction performance index diagram after STL decomposition of the measuring point PL8-U of a method for predicting displacement intervals of a jointed arch dam based on separation modeling and fractal theory provided in the second embodiment of the present invention.

[0070] Figure 7 A multi-model prediction performance index diagram after STL decomposition of the measuring point PL18-U of a method for predicting displacement intervals of a jointed arch dam based on separation modeling and fractal theory provided in the second embodiment of the present invention.

[0071] Figure 8 A multi-model prediction interval and quantitative evaluation diagram of the measuring point PL8-U of a displacement interval prediction method for a jointed arch dam based on separation modeling and fractal theory are provided in the second embodiment of the present invention.

[0072] Fig. 9 A multi-model prediction interval and quantitative evaluation diagram of the measuring point PL18-U of a displacement interval prediction method for a jointed arch dam based on separation modeling and fractal theory are provided in the second embodiment of the present invention. DETAILED DESCRIPTION

[0073] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0074] Example 1, reference Figure 1-Figure 2 , which is an embodiment of the present invention, provides a displacement interval prediction method for a seamed arch dam based on separation modeling and fractal theory, comprising:

[0075] S1: Perform STL decomposition on the displacement sequence, decompose the displacement into trend component, seasonal component and remainder, and perform multifractal analysis on the changing trend of displacement and components.

[0076] Furthermore, the displacement series is decomposed into trend component, seasonal component and residual term by STL through inner loop and outer loop, and the displacement time series Yt is decomposed into trend component, seasonal component and residual term. T t k is the seasonal component and trend component of the kth iteration, then the k+1th iteration process includes calculating the detrended sequence Y t -T t k, perform Loess smoothing on the subsequence of the detrended sequence to obtain a temporary seasonal sequence Pair Sequence Applying a low-pass filter and the Loess process yields The seasonal component of the k+1th iteration is The original series removes the seasonal component The Loess smoothing is used to process the sequence of the deseasonalized component, and the trend component of the k+1 iteration is T t k+1 .

[0077] In the outer loop, the seasonal component and trend component obtained after the inner loop are used to calculate the remainder It is expressed as:

[0078]

[0079] It should be noted that the multifractal analysis of the displacement and component trends includes calculating the displacement trend component, seasonal component and the Hurst index, Renyi index and multifractal spectrum of the displacement sequence itself using the MF-DFA method based on fractal theory. The multifractal analysis of the displacement and component trends is performed. Given a displacement time series x(t), the sequence mean is μx, and the cumulative deviation sequence y(t) from the mean is expressed as:

[0080]

[0081] By using a given time scale s, y(t) is divided equally into subintervals y v (t), and perform trend fitting, and subtract the fitting part from the original basis Get the residual sequence z v (t), expressed as:

[0082]

[0083] Calculate the residual sequence z v (t) and the mean square error F2(s,v) and the q-order fluctuation function F q (s), expressed as:

[0084]

[0085] If there is long-range correlation in the displacement time series, then F q The relationship between (s) and the time scale s is expressed as:

[0086] IgF q (s) = h (q) lgs + lgb

[0087] Among them, F q(s) is the q-order wave function of the displacement sequence, h(q) is the Hurst exponent, and b is a constant coefficient. The fractal intensity and fractal singularity of the displacement sequence are characterized by the multifractal spectrum f(α), which is expressed as:

[0088] τ(q)=qh(q)-1

[0089] α=dτ(q) / dq

[0090] f(α)=qα-τ(q)

[0091] Among them, τ(q) is the Renyi index, α is the singular intensity, and f(α) is the multifractal spectrum.

[0092] S2: Based on the HST model, the initial set of modeling factors is obtained. The crack opening and closing degree is taken as the influencing factor of displacement. A displacement monitoring model considering the influence of cracks is established to obtain the preliminary factor set. The variable selection method mRMR is used to optimize the influencing factors of displacement. The remaining modeling factors are combined into a screened factor set for normalization.

[0093] Furthermore, the initial modeling factor set includes water pressure component, temperature component and time-effect component. The crack opening and closing degree is taken as the influencing factor of displacement, and the displacement monitoring model (HSCT) considering the influence of cracks is established, and the preliminary factor set is obtained. Among them, the preliminary factor set includes water pressure component, temperature component, time-effect component and crack component.

[0094] It should be noted that the variable selection method mRMR is used to optimize the influencing factors of displacement, and the remaining modeling factors are combined into a screened factor set for normalization. The variable selection method mRMR is used to optimize the influencing factors of displacement, and the modeling factors that do not meet the threshold requirements after calculation by the mRMR method are deleted, and the remaining modeling factors are combined into a screened factor set. The factor elimination of the variable selection method mRMR includes mRMR using the mutual information method to measure the dependency between two variables, which is expressed as:

[0095]

[0096] Among them, x i is the influencing factor of displacement, y is the measured displacement, p(x i ) and p(y) represent x i and the marginal probability density of y, p(x i ,y) represents their joint probability density.

[0097] The maximum correlation D is expressed as the average value of mutual information, and the minimum redundancy R is expressed as x i and x j The minimum correlation between them is obtained, and the operator φ is defined to jointly operate D and R, which can be expressed as:

[0098]

[0099] Where F is x i is a feature set of F, where |F| is equal to the number of features in F.

[0100] The value of φ is calculated for each factor in turn, and the influencing factors corresponding to the value less than 1 are eliminated to form a candidate feature set, which is used as the input of the subsequent prediction model.

[0101] It should be noted that a deep learning model CNN-LSTM is established to predict the trend component and seasonal component of the displacement and then superimpose them. The superimposed displacement includes taking the daily value corresponding to the displacement influencing factor as the input of the CNN-LSTM model, and the actual displacement measured on the day as the output of the CNN-LSTM model. At the same time node, the measured data of each influencing factor and displacement are divided into a ratio of 8:2 as a training set and a test set respectively. The training set is used for training the corresponding CNN-LSTM model, and the test set is used to evaluate the prediction accuracy of the corresponding CNN-LSTM model.

[0102] The accuracy of the CNN-LSTM model is verified by calculating the root mean square error RMSE, mean absolute error MAE and determination coefficient R2 of the measured and predicted values ​​in the test set. The calculation of the three indicators is expressed as:

[0103]

[0104]

[0105] Among them, n represents the length of the time series, y i and are monitoring data and predicted values, respectively. and are the mean of the observed data series and the mean of the predicted data series, respectively.

[0106] S3: A deep learning model CNN-LSTM is established to predict the trend component and seasonal component of the displacement and then superimpose them to obtain the superimposed displacement. The interval prediction based on the Bootstrap method is used to calculate the evaluation index and compare the performance of the prediction interval of each model.

[0107] Furthermore, the interval prediction based on the Bootstrap method includes the given displacement and the influencing factor, which is expressed as:

[0108] (x s ,y s )(s=1,2,…,t)

[0109] Among them, t is the time length of the input sequence, xs is the set of factors affecting the dam displacement, y s is the displacement value.

[0110] By resampling the displacement data training set B times using the Bootstrap method, we get B sets of training sets. It is expressed as:

[0111]

[0112] in, is the predicted value output by the lth model,

[0113] Variance of model uncertainty It is expressed as:

[0114]

[0115] Keep input x s and replace the displacement sequence with Get the transformed data set D ξ , expressed as:

[0116]

[0117] Right D ξ Resample to obtain the estimate of noise variance and the variance of uncertainty in the noise regression model The variance of the data noise is obtained by superposition, which is expressed as:

[0118]

[0119] Assume the estimated error and data noise ξ(x s ) are statistically independent, and the variance of the total prediction error It is expressed as:

[0120]

[0121] in, and are the variance of model uncertainty and data noise, respectively.

[0122] It should be noted that the interval prediction based on the Bootstrap method also includes when the significance level is set to α, (1-α)×100% is the nominal confidence level of the prediction interval, then the displacement y s Prediction interval at time step s It is expressed as:

[0123]

[0124] in, and are the lower and upper bounds of the constructed prediction interval, is the output of the prediction model, is the variance of the total forecast error, z 1-α / 2 is the 1-α / 2 percentile of the standard normal distribution. In order to evaluate the overall quality of the constructed interval, the prediction interval coverage probability PICP, the average prediction interval width MPIW and the coverage and width criterion CWC are introduced to measure, which is expressed as:

[0125]

[0126] CWC=MPIW[1+γ(PICP)e -η(PICP-PINC) ]

[0127]

[0128] Among them, N t is the number of samples. When the displacement value is within the interval, c s The value of is 1, otherwise it is 0, and η is the penalty parameter, which is set to 50.

[0129] Example 2, reference Figure 3-Figure 9 , which is an embodiment of the present invention, provides a displacement interval prediction method for a seamed arch dam based on separation modeling and fractal theory. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments.

[0130] First, collect data of a concrete gravity arch dam, the maximum dam height is 76.30m, the dam top elevation is 126.30m, the dam top width is 8.00m, and the maximum dam bottom width is 53.50m. The arch dam vertical monitoring point layout is shown in the figure below. Figure 3 As shown in the figure. Due to construction conditions and sudden drop in temperature, cracks appeared at the 105m elevation of the downstream face of the arch dam during construction, and the crack opening displacement continued to increase after reinforcement. In order to consider the impact of cracks on the displacement of nearby measuring points, this study selected the horizontal displacements of measuring points PL8-U and PL18-U near the 105m elevation from January 1, 2006 to December 31, 2015 for analysis.

[0131] The STL algorithm is used to decompose the displacement of the measuring points, and the displacement series is decomposed into trend component, seasonal component and remainder. Figure 4 The STL decomposition results of the measurement points PL8-U and PL18-U are shown. Figure 4It can be seen that there is an increasing trend in the trend components of the two measuring points. There is a strong periodicity in the seasonal components of the two measuring points. In addition, the residuals of the two measuring points have no obvious regularity and are highly irregular. Therefore, they are removed in the subsequent prediction. To further analyze the changing trend of displacement, MF-DFA is used to analyze the displacement and its trend component and seasonal component. The fluctuation order q ranges from [-10,10]. The analysis results are as follows: Figure 5 shown.

[0132] Figure 5 (a) is the change of Hurst index of displacement and its trend component and seasonal component, Figure 5 (a) It can be seen that when q changes between [-10,10], the Hurst exponent of the measuring point displacement and its components shows a nonlinear decreasing trend, indicating that the measuring point displacement and its components have obvious multi-fractal characteristics. Under different fluctuation orders q, the h(q) value is significantly greater than 0.5, indicating that the displacement and its components have good memory and long-range correlation. Based on h(q), the Renyi index, i.e. the scaling function τ(q), is calculated, as follows: Figure 5 (b) The scaling functions of the measuring point displacement and its components have good consistency, the middle part is convex, and satisfies τ(0) = -1. The overall relationship is nonlinear, which further verifies that the measuring point displacement has multiple fractal characteristics. Figure 5 (c) is a multifractal spectrum. Each multifractal spectrum image is a single-peak convex distribution, which is similar to a quadratic function curve, depicting the diversity of local changes at different times. The singular intensity is mainly concentrated on both sides of the image, reflecting the multifractal characteristics of the displacement sequence of the measuring point. The above analysis shows that the changes in the displacement of the measuring point and its components are consistent and gradually rising, and have self-similarity and regularity. In addition, compared with the direct multifractal analysis of the displacement, the multifractal analysis of the trend component and seasonal component of the displacement after STL decomposition can also draw the same conclusion.

[0133] The measuring points of the arch dam near the elevation of 105 m will be significantly affected by the cracks. Therefore, the crack opening displacement is introduced and a displacement monitoring model (HSCT) considering the influence of cracks is established. Its mathematical expression is:

[0134]

[0135] Where H is the upstream water depth, a0 is a constant term, and a i , b 1j , b 2j , c1, c2, d k is the fitting coefficient, t is the cumulative number of days from the selected displacement observation date to the initial displacement monitoring date, θ = t / 100, p is the number of joint meters, J k It is the crack opening degree of each crack gauge.

[0136] Based on the HSCT regression model, the monitoring data of 19 joint meters were selected, that is, p is 19, and the factor set affecting the displacement can be expressed as Where J = {J1, J2, ..., J 19 In order to obtain the optimal factor set, the mRMR method was used to eliminate the factors corresponding to φ less than 1. Table 1 shows the results of mRMR calculation of measuring points PL8-U and PL18-U under the HSCT model.

[0137] Table 1 mRMR calculation results of factors at two measurement points under HSCT model

[0138] Impact Factor PL8-U PL18-U Impact Factor PL8-U PL18-U x1 3.947 1.266 J6 1.527 1.390 x2 1.108 2.885 J7 0.864 0.531 x3 1.528 0.838 J8 1.527 1.390 x4 0.875 0.662 J9 1.584 1.442 x5 19.641 17.617 J10 3.042 5.740 x6 1.701 1.002 J11 0.583 0.772 x7 4.540 10.262 J12 5.069 1.389 x8 34.841 1.329 J13 0.625 0.569 x9 1.528 1.389 J14 5.456 10.563 x10 0.735 0.669 J15 1.527 1.388 J1 7.400 6.787 J16 18.267 4.252 J2 5.074 1.390 J17 0.639 0.581 J3 1.528 1.389 J18 0.862 1.402 J4 3.044 2.764 J19 5.078 1.390 J5 1.005 12.943 - - -

[0139] For measuring point PL8-U, after mRMR calculation, factors x4, x10, J7, J11, J13, J17, and J18 were eliminated. For measuring point PL18-U, after mRMR calculation, factors x3, x4, x10, J7, J11, J13, and J17 were eliminated. Table 2 shows the final factor set after variable selection for different measuring points. In the subsequent point value prediction and interval prediction, the factor set in Table 2 is used as the influencing factor set of the displacement of the arch dam with joints.

[0140] Table 2 Factor set obtained after screening by mRMR method

[0141] Measuring point The selected factor set PL8-U {x1-x3,x5-x9,J1-J6,J8-J10,J12,J14-J16,J19} PL18-U {x1,x2,x5-x9,J1-J6,J8-J10,J12,J14-J16,J18,J19}

[0142] In order to verify that STL decomposition can effectively eliminate the influence of noise and is conducive to separation modeling, SSA-ELM, LSTM and CNN-LSTM models are established to predict the displacement trend component and seasonal component after STL decomposition and superimpose them to obtain the superimposed displacement. For the purpose of comparison, the above three models are also used to predict the displacement without STL decomposition. The first 80% of the data set is used as the training set of the model, and the last 20% is used as the test set to evaluate the accuracy of the model. Figure 6 and Figure 7 Multi-model statistical metrics for the two-point test set are shown.

[0143] according to Figure 6From the statistical indicators, we can see that compared with SSA-ELM, the RMSE of the CNN-LSTM model is reduced by 18.0% on average, and the MAE is reduced by 20.1% on average. Compared with LSTM, the RMSE of the CNN-LSTM model is reduced by 6.9% on average, and the MAE is reduced by 7.7% on average. The optimal model is CNN-LSTM, with a goodness of fit of up to 0.961, which reflects the advantage of CNN-LSTM in processing high-dimensional nonlinear relationships. From the above analysis, it can be seen that compared with directly predicting and analyzing the displacement, after decomposing the displacement by STL, the trend component and seasonal component of the displacement are separately modeled and superimposed, which can effectively improve the prediction accuracy.

[0144] The traditional point value prediction method only outputs a deterministic value as the final prediction result, but the interval prediction can construct a high-confidence interval to represent the uncertainty of the displacement. In order to study the performance of different prediction algorithms in quantifying the impact of uncertainty, after STL eliminates the influence of noise, the interval prediction of three models, SSA-ELM, LSTM and CNN-LSTM, is implemented based on the Bootstrap method at 95% confidence. Figure 8 and Fig. 9 The prediction intervals of the two measuring points and their quantitative evaluation are given. Table 3 shows the quantitative evaluation indicators of the interval prediction of the displacement of the two measuring points under multiple models.

[0145] Table 3 Performance indicators of interval prediction between two measuring points

[0146]

[0147]

[0148] Table 3 further shows that the constructed prediction intervals can cover most of the samples. At 95% confidence, the PICP values ​​of the three models are all greater than the specified PINC. Since PICP is greater than PINC, CWC will not be affected by the penalty term and its value is equal to MPIW, which shows that the obtained prediction intervals are valid. For the measurement point PL8-U, the coverage of the three models is as high as 98%. The MPIW of SSA-ELM is greater than that of the other two models, and the MPIW of CNN-LSTM is the smallest among the three models. For the measurement point PL18-U, the MPIW of SSA-ELM is still greater than that of the other two models, and the coverage rate just exceeds PICP. Although LSTM has a greater coverage rate than SSA-ELM, its MPIW inevitably increases. The coverage rate of CNN-LSTM is the highest and has the smallest MPIW. The prediction results of CNN-LSTM are relatively consistent with the true value, so the model can obtain a narrower prediction interval. The above results show that different prediction algorithms have differences in uncertainty quantification, but the proposed interval prediction based on the Bootstrap method and CNN-LSTM model has a satisfactory performance, that is, the uncertainty of displacement change is accurately quantified by the prediction interval.

[0149] Embodiment 3, an embodiment of the present invention, provides a displacement interval prediction system for a seamed arch dam based on separation modeling and fractal theory, comprising a sequence decomposition module, a factor screening module, and a model search and training module.

[0150] Among them, the path planning module is used to perform STL decomposition on the displacement sequence, decompose the displacement into trend components, seasonal components and remainders, and perform multi-fractal analysis on the displacement and component change trends. The factor screening module is used to obtain the initial modeling factor set based on the HST model, take the crack opening and closing degree as the influencing factor of the displacement, establish a displacement monitoring model considering the influence of cracks, obtain the preliminary factor set, use the variable selection method mRMR to optimize the influencing factors of the displacement, and form the remaining modeling factors into a screened factor set for normalization. The model search training module is used to establish a deep learning model CNN-LSTM to predict the trend component and seasonal component of the displacement and then superimpose them to obtain the superimposed displacement, calculate the evaluation index based on the interval prediction of the Bootstrap method, and compare the performance of the prediction interval of each model.

[0151] If the function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods of each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0152] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0153] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering, or processing in another suitable manner as necessary, and then stored in a computer memory.

[0154] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc. It should be noted that the above embodiments are only used to illustrate the technical solution of the present invention and are not limited. Although the present invention is described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical solution of the present invention can be modified or replaced by equivalents without departing from the spirit and scope of the technical solution of the present invention, which should be included in the scope of the claims of the present invention.

[0155] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A displacement interval prediction method for a seamed arch dam based on separation modeling and fractal theory, characterized in that: include: Perform STL decomposition on the displacement sequence, decompose the displacement into trend component, seasonal component and remainder, and perform multi-fractal analysis on the changing trend of displacement and components; Based on the HST model, the initial modeling factor set is obtained. The crack opening and closing degree is taken as the influencing factor of displacement. The displacement monitoring model considering the influence of cracks is established to obtain the preliminary factor set. The variable selection method mRMR is used to optimize the influencing factors of displacement. The remaining modeling factors are combined into a screened factor set and normalized. A deep learning model CNN-LSTM is established to predict the trend component and seasonal component of the displacement and then superimpose them to obtain the superimposed displacement. The interval prediction based on the Bootstrap method is used to calculate the evaluation index and compare the performance of the prediction interval of each model.

2. The displacement interval prediction method for a seamed arch dam based on separation modeling and fractal theory according to claim 1 is characterized in that: The method of performing STL decomposition on the displacement sequence to decompose the displacement into trend component, seasonal component and residual term includes decomposing the displacement time series Yt into trend component, seasonal component and residual term through inner loop and outer loop of STL. is the seasonal component and trend component of the kth iteration, then the k+1th iteration process includes calculating the detrended sequence Perform Loess smoothing on the subsequence of the detrended sequence to obtain a temporary seasonal series Pair Sequence Applying a low-pass filter and the Loess process yields The seasonal component of the k+1th iteration is The original series removes the seasonal component The Loess smoothing is used to process the sequence of the deseasonalized component, and the trend component of the k+1 iteration is T t k+1 ; In the outer loop, the seasonal component and trend component obtained after the inner loop are used to calculate the remainder It is expressed as:

3. The displacement interval prediction method for a seamed arch dam based on separation modeling and fractal theory as claimed in claim 2 is characterized by: The multifractal analysis of the displacement and component change trends includes calculating the displacement trend component, the seasonal component and the Hurst index, the Renyi index and the multifractal spectrum of the displacement sequence itself using the MF-DFA method based on the fractal theory, and performing multifractal analysis on the displacement and component change trends. Given a displacement time series x(t), the sequence mean is μx, and the cumulative deviation sequence y(t) between x(t) and the mean is expressed as: By using a given time scale s, y(t) is divided equally into subintervals y v (t), and perform trend fitting, and subtract the fitting part from the original basis Get the residual sequence z v (t), expressed as: Calculate the residual sequence z v (t) and the mean square error F2(s,v) and the q-order fluctuation function F q (s), expressed as: If there is long-range correlation in the displacement time series, then F q The relationship between (s) and the time scale s is expressed as: lgF q (s)=h(q)lgs+lgb Among them, F q (s) is the q-order wave function of the displacement sequence, h(q) is the Hurst exponent, and b is a constant coefficient; The fractal strength and fractal singularity of the displacement sequence are characterized by the multifractal spectrum f(α), which is expressed as: τ(q)=qh(q)-1 α=dτ(q) / dq f(α)=qα-τ(q) Among them, τ(q) is the Renyi index, α is the singular intensity, and f(α) is the multifractal spectrum.

4. The displacement interval prediction method for a seamed arch dam based on separation modeling and fractal theory as claimed in claim 3 is characterized by: The variable selection method mRMR is used to optimize the influencing factors of the displacement, and the remaining modeling factors are combined into a screened factor set, and the normalization processing includes using the variable selection method mRMR to optimize the influencing factors of the displacement, deleting the modeling factors that do not meet the threshold requirements after calculation by the mRMR method, and combining the remaining modeling factors into a screened factor set. The factor elimination of the variable selection method mRMR includes mRMR using the mutual information method to measure the dependency between two variables, which is expressed as: Among them, x i is the influencing factor of displacement, y is the measured displacement, p(x i ) and p(y) represent x i and the marginal probability density of y, p(x i ,y) represents their joint probability density. The maximum correlation D is expressed as the average value of mutual information, and the minimum redundancy R is expressed as x i and x j The minimum correlation between them is obtained, and the operator φ is defined to jointly operate D and R, which can be expressed as: Where F is x i is a feature set of , where |F| is equal to the number of features in F. The value of φ is calculated for each factor in turn, and the influencing factors corresponding to the value less than 1 are eliminated to form a candidate feature set, which is used as the input of the subsequent prediction model.

5. The displacement interval prediction method for a seamed arch dam based on separation modeling and fractal theory as claimed in claim 4 is characterized by: The deep learning model CNN-LSTM is established to predict the trend component and the seasonal component of the displacement and then superimpose them to obtain the superimposed displacement, including taking the daily value corresponding to the displacement influencing factor as the input of the CNN-LSTM model, and the daily measured displacement as the output of the CNN-LSTM model. At the same time node, the measured data of each influencing factor and the displacement are divided into a training set and a test set in a ratio of 8:2, respectively. The training set is used for training the corresponding CNN-LSTM model, and the test set is used to evaluate the prediction accuracy of the corresponding CNN-LSTM model; The accuracy of the CNN-LSTM model is verified by calculating the root mean square error RMSE, mean absolute error MAE and determination coefficient R2 of the measured and predicted values ​​in the test set. The calculation of the three indicators is expressed as: Among them, n represents the length of the time series, y i and are monitoring data and predicted values, respectively. and are the mean of the observed data series and the mean of the predicted data series, respectively.

6. The displacement interval prediction method for a seamed arch dam based on separation modeling and fractal theory as claimed in claim 5 is characterized by: The interval prediction based on the Bootstrap method includes a given displacement and an influencing factor, which is expressed as: (x s ,y s )(s=1,2,…,t) Among them, t is the time length of the input sequence, x s is the set of factors affecting the dam displacement, y s is the displacement value; By resampling the displacement data training set B times using the Bootstrap method, we get B sets of training sets. It is expressed as: in, is the predicted value output by the lth model, Variance of model uncertainty It is expressed as: Keep input x s and replace the displacement sequence with Get the transformed data set D ξ , expressed as: Right D ξ Resample to obtain the estimate of noise variance and the variance of uncertainty in the noise regression model The variance of the data noise is obtained by superposition, which is expressed as: Assume the estimated error and data noise ξ(x s ) are statistically independent, and the variance of the total prediction error It is expressed as: in, and are the variance of model uncertainty and data noise, respectively.

7. The displacement interval prediction method for a seamed arch dam based on separation modeling and fractal theory as claimed in claim 6 is characterized by: The interval prediction based on the Bootstrap method also includes that when the significance level is set to α, (1-α)×100% is the nominal confidence level of the prediction interval, then the displacement y s Prediction interval at time step s It is expressed as: in, and are the lower and upper bounds of the constructed prediction interval, is the output of the prediction model, is the variance of the total forecast error, z 1-α / 2 is the 1-α / 2 percentile of the standard normal distribution. In order to evaluate the overall quality of the constructed interval, the prediction interval coverage probability PICP, the average prediction interval width MPIW and the coverage and width criterion CWC are introduced to measure, which is expressed as: CWC=MPIW[1+γ(PICP)e -η(PICP-PINC) ] Among them, N t is the number of samples. When the displacement value is within the interval, c s The value of is 1, otherwise it is 0, and η is the penalty parameter, which is set to 50.

8. A system using the method for predicting displacement intervals of a seamed arch dam based on separation modeling and fractal theory as claimed in any one of claims 1 to 7, characterized in that: Including sequence decomposition module, factor screening module, model search and training module; The path planning module is used to perform STL decomposition on the displacement sequence, decompose the displacement into trend components, seasonal components and remainders, and perform multi-fractal analysis on the change trends of the displacement and components; The factor screening module is used to obtain an initial set of modeling factors based on the HST model, take the crack opening degree as the influencing factor of displacement, establish a displacement monitoring model considering the influence of cracks, obtain a preliminary factor set, use the variable selection method mRMR to optimize the influencing factors of displacement, and form the remaining modeling factors into a screened factor set for normalization. The model training module is used to establish a deep learning model CNN-LSTM to predict the trend component and seasonal component of the displacement and then superimpose them to obtain the superimposed displacement, perform interval prediction based on the Bootstrap method, calculate evaluation indicators, and compare the performance of each model prediction interval.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method for predicting displacement intervals of a jointed arch dam based on separation modeling and fractal theory as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the displacement interval prediction method of a jointed arch dam based on separation modeling and fractal theory as described in any one of claims 1 to 7 are implemented.