Dynamic prediction method for residual life of side slope of water conservancy project

Abstract

The invention relates to the technical field of hydraulic engineering safety monitoring, in particular to a hydraulic engineering slope residual life dynamic prediction method, which comprises the steps of influence factor selection; slope displacement is predicted; determining a slope failure threshold value; establishing a binary Wiener process model; the slope residual life dynamic prediction method has the beneficial effects that displacement prediction is carried out on the slope with the creep trend, then the predicted value is substituted into the binary Wiener process residual life prediction model, slope residual life predicted values corresponding to different periods in the future can be obtained, and meanwhile, a sample library is continuously updated by utilizing a new monitoring result; a dynamic residual life prediction value is obtained, and the future state of the slope can be mastered more accurately in real time.

Description

Technical Field

[0001] The invention relates to the technical field of water conservancy project safety monitoring, and in particular to a method for dynamically predicting the remaining life of a water conservancy project slope. Background Art

[0002] In order to prevent landslide disasters on reservoir slopes, channel slopes, etc. and reduce the loss of life and property caused by landslides, it is particularly important to conduct long-term monitoring of the slopes and understand the changes in the safety status of the slopes.

[0003] In the prediction of the remaining life of slopes, a remaining life value is often predicted based on the monitoring data of slope displacement, deformation, etc. through regression analysis, grey theory, time series analysis, artificial neural network method, etc.

[0004] In fact, the state change of the slope is a long-term nonlinear process, and the remaining life prediction value will also show a trend of continuous change. In order to obtain the future state of the slope more accurately, it is necessary to carry out dynamic prediction of the remaining life of the slope. Summary of the invention

[0005] The purpose of the present invention is to provide a method for dynamically predicting the remaining life of a water conservancy project slope, so as to solve the problems raised in the above-mentioned background technology.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A method for dynamically predicting the remaining life of a water conservancy project slope comprises the following steps:

[0008] S1: Selection of impact factors;

[0009] S2: slope displacement prediction;

[0010] S3: Determination of slope failure threshold;

[0011] S4: Building a binary Wiener process model;

[0012] S5: Dynamic prediction of slope remaining life.

[0013] Preferably, step S1 specifically includes: selecting average rainfall data of a recent time period as a rainfall factor, taking the daily average water level as a water level factor, taking the time function as a time factor, and taking the displacements in two directions of the slope as two performance indicators of the slope.

[0014] Preferably, step S2 specifically includes the following steps:

[0015] S21: Data sample preprocessing: normalize the impact factor data;

[0016] S22: Initialization of particle swarm algorithm parameters: setting the initial values ​​of penalty coefficient, kernel function parameters, maximum number of particle swarm populations, maximum number of evolutions, learning factor, and inertia weight parameters;

[0017] S23: Model training: Search the support vector regression machine parameter penalty factor and kernel function parameters through the particle swarm algorithm to obtain the optimal parameters.

[0018] Preferably, step S3 specifically includes: referring to the requirements of the specification, adopting the slope displacement monitoring and early warning criterion based on the safety factor as the slope failure threshold.

[0019] Preferably, step S4 specifically includes the following steps:

[0020] S41: Establishing the life distribution function of each performance index;

[0021] S42: Establish the slope joint life distribution function;

[0022] S43: Establish the life probability density function of the slope binary Wiener process;

[0023] S44: Use the maximum likelihood method to estimate unknown parameters;

[0024] S45: Prediction of remaining life of slope.

[0025] Preferably, step S5 specifically includes: incorporating the latest acquired slope monitoring data into the sample, repeating S1-S4, and obtaining the latest predicted remaining life of the slope.

[0026] Preferably, step S45 specifically includes: taking different moments in a future period of time as the current moment to make predictions, predicting the remaining life results corresponding to the different moments in the future, and taking the date closest to the current moment as the currently predicted remaining life of the slope.

[0027] Compared with the prior art, the present invention has the following beneficial effects:

[0028] The present invention predicts the displacement of a slope with an existing creep tendency and then substitutes the predicted value into a binary Wiener process remaining life prediction model to obtain the corresponding remaining life prediction values ​​of the slope at different future periods. At the same time, the sample library is continuously updated with new monitoring results to obtain a dynamic remaining life prediction value, which can more accurately grasp the future state of the slope in real time. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 is a flow chart of the present invention;

[0030] Figure 2 It is the x-direction displacement fitting diagram of the present invention;

[0031] Figure 3It is the y-direction displacement fitting diagram of the present invention. DETAILED DESCRIPTION

[0032] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0033] For example, see Figure 1 The present invention provides a technical solution: a method for dynamically predicting the remaining life of a water conservancy project slope, the method comprising the following steps:

[0034] 1. Selection of impact factors

[0035] The average rainfall data of the most recent period is selected as the rainfall factor, the daily average water level is taken as the water level factor, the time function is taken as the timeliness factor, and the displacement in two directions of the slope are taken as the two performance indicators of the slope.

[0036] 2. Slope displacement prediction

[0037] (1) Data sample preprocessing: normalize the impact factor data;

[0038] (2) Initialization of particle swarm algorithm parameters: setting the initial values ​​of parameters such as penalty coefficient, kernel function parameters, maximum number of particle swarm populations, maximum number of evolutions, learning factor, inertia weight, etc.

[0039] (3) Model training: Use particle swarm algorithm to search for support vector regression machine parameter penalty factor and kernel function parameters to obtain the optimal parameters. Use support vector machine model to establish channel slope displacement and influencing factors

[0040] The nonlinear response model between them is then used to learn the sample data using the obtained optimal parameters.

[0041] 3. Determination of slope failure threshold

[0042] Referring to relevant specifications, the slope displacement monitoring and early warning criterion based on safety factor is used as the slope failure threshold S cr :

[0043]

[0044] Where: F cr is the slope safety factor; S 0 is the initial elastic deformation of the slope; S cr is the slope failure threshold.

[0045] 4. Establish a binary Wiener process model

[0046] (1) Establish the life distribution function of each performance index;

[0047] According to the properties of the Wiener process, the slope life T k If it obeys the inverse Gaussian distribution, then the life span of the kth performance index of the slope is T k The distribution function and probability density function are

[0048]

[0049] In the formula: In the formula, Φ(·) represents the standard normal distribution function; ω k is the slope failure threshold corresponding to the kth performance index; μ k , σ k are the drift coefficient and diffusion coefficient respectively.

[0050] (2) Establish the joint life distribution function of slopes;

[0051] Select the displacements in two mutually perpendicular directions at a certain measuring point on the slope as the performance degradation index, then the performance degradation trajectory of the slope at time t is X(t) = (X 1 (t),X 2 (t)), if the slope is still in service at time t, the slope joint life distribution can be expressed as:

[0052] F(t)=(T <t)= 1 (t)+ 2 (t)-(F 1 (t), 2 (t);)

[0053] Where: θ is the parameter of the Copula function; F 1 (t), F 2 (t) represents the lifetime distribution function in the x and y directions, respectively. In this paper, it corresponds to the lifetime distribution function of displacement in the two directions.

[0054] Taking the Clayton Copula function as an example, there is a correlation between the displacements in the two directions of the slope, and the distribution function is

[0055]

[0056] Then the life distribution function expression of the slope based on the binary Wiener process is:

[0057]

[0058] (3) Establish the life probability density function of the slope binary Wiener process;

[0059] The probability density function of the slope life based on the binary Wiener process is:

[0060]

[0061] The life probability density function indicates the possibility of the life at a certain value point, so the horizontal axis corresponding to the maximum value of the slope life probability density is taken as the life prediction value.

[0062] (4) Use the maximum likelihood method to estimate unknown parameters

[0063] Note ΔX ki is the slope at time t i-1 With t i The degradation increment between i =t i -t i-1 is time t i-1 With t i The time interval between them. The maximum likelihood method is used to estimate the unknown parameters:

[0064]

[0065] Where n is the sample size of the monitoring data corresponding to the degradation amount of the performance indicator.

[0066] According to the obtained parameter estimate μ k , σ k , the marginal distribution function F 1 (t), F 2 (t) are used as the input values ​​of the Copula function. The maximum likelihood method is also used to estimate the parameter θ of the Copula function. That is, the following formula L(θ) is differentiated with respect to the parameter θ, and the derivative is set to 0 to obtain the parameter estimate.

[0067]

[0068] (5) Prediction of remaining life of slope.

[0069] The prediction is made by taking different moments in the future as the current moment, predicting the remaining life results corresponding to different moments in the future, and taking the date closest to the current time as the currently predicted remaining life of the slope.

[0070] 5. Dynamic prediction of slope remaining life

[0071] The latest acquired slope monitoring data is included in the sample, and S1 to S4 are repeated to obtain the latest predicted remaining life of the slope.

[0072] Embodiment 2: Refer to the attached Figure 2-Figure 3As shown, based on the first embodiment, a slope monitoring data is selected to perform dynamic prediction of remaining life, and the specific steps are as follows:

[0073] 1. Selection of impact factors

[0074] The average rainfall data of the previous 30 days is taken as the rainfall factor, the daily average channel water level is taken as the water level factor, the time function t is taken as the time factor, and the displacement in two directions of the slope is taken as the two performance indicators of the slope.

[0075] 2. Slope displacement prediction

[0076] (1) Data sample preprocessing: normalize the impact factor data;

[0077] (2) Initialization of PSO algorithm parameters: Set the penalty coefficient C = [0.1, 100], kernel function parameter γ = [0.1, 500], the maximum number of particle swarm populations to 20, the maximum number of evolutions to 100, and the learning factor c 1 =1.2, c 2 =1.9, inertia weight w=1.

[0078] (3) Model training: The particle swarm algorithm is used to search for the support vector regression machine parameter penalty factor and kernel function parameters. The optimal parameters finally obtained are shown in the following table.

[0079] Table 1 Optimal parameters of support vector machine

[0080] direction c γ X direction 17.2088 3.9064 Y direction 15.7674 3.7827

[0081] The support vector machine model is used to establish a nonlinear response model between the channel slope displacement and the influencing factors, and then the sample data is learned using the obtained optimal parameters to output the fitting results, such as Figure 2 and Figure 3 .

[0082] Using the trained displacement prediction model, the displacement changes of this slope in the x and y directions in July 2019 are predicted: x = 78.49 mm; y = 40.19 mm.

[0083] 3. Determination of slope failure threshold

[0084] The measured displacement data from May 1978 to May 1982 were used to model the Wiener process degradation, and the time interval was 1 day. The landslide was divided into four deformation stages: the initial deformation stage (before November 1979), the constant deformation stage (November 1979 to August 1982), the accelerated deformation stage (August 1982 to May 1985), and the landslide entered the pre-slip stage in May 1985. The time when the initial deformation stage ended was determined to be t 0=518d, the corresponding initial elastic deformations are:

[0085] S 0x =404.50mm, S 0y =152.93mm

[0086] Refer to the "Slope Design Code for Water Conservancy and Hydropower Projects" (SL386-2007) to comprehensively determine the safety factor F cr =1.25, the displacement warning criteria (i.e., thresholds) of the slope are calculated as follows:

[0087] S crx =2022.50mm, S cry =764.65mm

[0088] 4. Establish a binary Wiener process model

[0089] Substituting the above displacement prediction values ​​into the constructed binary Wiener process remaining life prediction model, the corresponding model parameters for July 2019 are as follows:

[0090] Table 2 Model parameter estimates

[0091] parameter <![CDATA[μ 1 ]]> <![CDATA[μ 2 ]]> <![CDATA[σ 1 ]]> <![CDATA[σ 2 ]]> θ Estimated value 0.0526 0.0257 0.2865 0.5192 13.0131

[0092] Taking July 2019 as the current time, the remaining life prediction value of the channel slope is:

[0093] T RUL =1569d

[0094] Using the displacement prediction model established in this paper, we continue to predict the displacement change values ​​in August and September 2019:

[0095] Predicted displacement values ​​in the x and y directions at future times

[0096]

[0097]

[0098] Substitute the above displacement prediction values ​​into the binary Wiener process remaining life prediction model constructed based on the slope, and use the future moment as the current moment to predict the remaining life of the slope. The results are summarized in the following table:

[0099] Table 3 Remaining life prediction results

[0100] Current moment <![CDATA[T RUL ]]> Corresponding date June 2019 3197 December 29, 2023 July 2019 1569 October 18, 2023 August 2019 1503 September 19, 2023 September 2019 1433 August 21, 2023

[0101] The closest date to the present, August 21, 2023, is taken as the currently predicted remaining life of the slope.

[0102] 5. Dynamic prediction of slope remaining life

[0103] According to the above results combined with the engineering survey data of the slope, it can be seen that the slope is still in a creep state and has a further sliding trend, that is, the life prediction value will gradually decrease over time. Therefore, it is necessary to further repeat the above steps in combination with the updated monitoring values, re-predict the remaining life of the slope at regular intervals, obtain the dynamic prediction results of the remaining life of the slope, and grasp the future state of the slope in real time.

[0104] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for dynamically predicting the remaining life of a water conservancy project slope, characterized by: The method comprises the following steps: S1: Selection of impact factors; S2: slope displacement prediction; S3: Determination of slope failure threshold; S4: Building a binary Wiener process model; S5: Dynamic prediction of slope remaining life.

2. A method for dynamically predicting the remaining life of a water conservancy project slope according to claim 1, characterized in that: Step S1 specifically includes: selecting the average rainfall data of a recent time period as the rainfall factor, taking the daily average water level as the water level factor, taking the time function as the timeliness factor, and taking the displacements in two directions of the slope as two performance indicators of the slope.

3. A method for dynamically predicting the remaining life of a water conservancy project slope according to claim 2, characterized in that: Step S2 specifically includes the following steps: S21: Data sample preprocessing: normalize the impact factor data; S22: Initialization of particle swarm algorithm parameters: setting the initial values ​​of penalty coefficient, kernel function parameters, maximum number of particle swarm populations, maximum number of evolutions, learning factor, and inertia weight parameters; S23: Model training: Search the support vector regression machine parameter penalty factor and kernel function parameters through the particle swarm algorithm to obtain the optimal parameters.

4. A method for dynamically predicting the remaining life of a water conservancy project slope according to claim 3, characterized in that: Step S3 specifically includes: referring to the requirements of the specification, adopting the slope displacement monitoring and early warning criterion based on the safety factor as the slope failure threshold.

5. A method for dynamically predicting the remaining life of a water conservancy project slope according to claim 4, characterized in that: Step S4 specifically includes the following steps: S41: Establishing the life distribution function of each performance index; S42: Establish the slope joint life distribution function; S43: Establish the life probability density function of the slope binary Wiener process; S44: Use the maximum likelihood method to estimate unknown parameters; S45: Prediction of remaining life of slope.

6. A method for dynamically predicting the remaining life of a water conservancy project slope according to claim 5, characterized in that: Step S5 specifically includes: incorporating the latest acquired slope monitoring data into the sample, repeating S1-S4, and obtaining the latest predicted remaining life of the slope.

7. A method for dynamically predicting the remaining life of a water conservancy project slope according to claim 5, characterized in that: Step S45 specifically includes: taking different moments in a future period of time as the current moment to make predictions, predicting the remaining life results corresponding to the different moments in the future, and taking the date closest to the current moment as the currently predicted remaining life of the slope.

Images