Early warning method for reservoir bank slope

Abstract

The invention discloses a reservoir bank slope early warning method, and relates to the field of reservoir bank slope monitoring, and the reservoir bank slope early warning method based on a multivariate regression statistical model and a quality control chart comprises the steps: building the multivariate regression statistical model between slope physical quantity monitoring data and environmental factors, introducing a risk rate index, and carrying out the quantitative evaluation of a slope state, meanwhile, the change trend of the physical quantity of the side slope is visually reflected by using a quality control chart, and grading early warning indexes are set. In order to solve the problem of low reservoir bank slope early warning precision in the prior art, the invention provides the reservoir bank slope early warning method, a multivariate regression statistical model between slope physical quantity monitoring data and environmental factors is established, and a risk rate index is introduced to carry out quantitative evaluation on a slope state, so that the reservoir bank slope early warning precision is improved. Meanwhile, the change trend of the physical quantity of the side slope is visually reflected through the quality control chart, graded early warning indexes are set, and the early warning precision is improved.

Description

Technical Field

[0001] The present application relates to the field of reservoir bank slope monitoring, and in particular to an early warning method for reservoir bank slopes. Background Art

[0002] As the scale of various engineering construction continues to expand, the stability issues faced by slope engineering are becoming increasingly prominent. Reservoir bank slopes not only cause significant economic losses, but may also cause serious safety accidents, endangering the lives and property of the people. Therefore, how to timely and accurately warn of reservoir bank slope risks has become an important issue that needs to be urgently addressed in engineering construction and operation management.

[0003] At present, a lot of research has been carried out on reservoir bank slope warning, mainly focusing on slope stability analysis, monitoring data analysis and warning model construction. Traditional reservoir bank slope warning methods mainly include empirical judgment method, statistical analysis method and numerical simulation method. Among them, the empirical judgment method mainly relies on the experience and judgment of experts and is highly subjective; the statistical analysis method establishes a reservoir bank slope warning index system by statistically analyzing historical monitoring data, but the warning accuracy is greatly affected by data quality and selected indicators; the numerical simulation method analyzes the stress and strain state of the slope by establishing a slope stability calculation model, but there is uncertainty in the parameter selection and boundary condition setting of the model.

[0004] In recent years, with the rapid development of technologies such as the Internet of Things and big data, the collection and transmission of reservoir bank slope monitoring data has become more convenient, providing massive data support for reservoir bank slope early warning. However, how to effectively use these monitoring data to establish a high-precision and reliable early warning model is still an urgent problem to be solved. Most of the existing early warning models use a single physical quantity indicator, such as displacement, cracks, etc., while ignoring the influence of environmental factors, resulting in low early warning accuracy. In addition, the existing early warning methods lack quantitative risk assessment methods, and the credibility of the early warning results needs to be improved. Summary of the invention

[0005] In response to the problem of low accuracy of reservoir bank slope warning in the prior art, the present application provides an early warning method for reservoir bank slopes. A multivariate regression statistical model is established between the slope physical quantity monitoring data and environmental factors, and a risk rate indicator is introduced to quantitatively evaluate the slope status. At the same time, a quality control chart is used to intuitively reflect the changing trend of the slope physical quantity, and a graded early warning indicator is set to improve the early warning accuracy.

[0006] The purpose of this application is achieved through the following technical solutions.

[0007] The present application provides an early warning method for a reservoir bank slope, including: dividing the slope state into three warning levels: safe state, early warning state and emergency state according to the safety level of the slope project, and setting a first-level early warning indicator Pf1 ' and the secondary warning index P f2 '; among them, P f1 ' is the risk rate threshold in the warning state, and P f2 ' is the risk rate threshold in the emergency state; obtain the historical monitoring data of slope physical quantities and the environmental factor data affecting slope stability; among them, the slope physical quantities include slope displacement, seepage pressure, and reservoir water level; the environmental factor data include reservoir water level, rainfall, temperature, and monitoring time; use the historical monitoring data of slope physical quantities and environmental factor data to establish a multiple regression statistical model between slope physical quantities and environmental factors, and obtain the multiple correlation coefficient R and regression standard deviation S of the multiple regression statistical model; among them, R reflects the goodness of fit of the multiple regression statistical model, and the value range is 0 to 1; S reflects the degree of dispersion of the multiple regression statistical model; when R is greater than the threshold, use the multiple regression statistical model to calculate the fitted value of the slope physical quantity, and calculate the root mean square error RMSE between the fitted value and the measured value; when RMSE is less than the threshold, perform the next step, otherwise adjust the parameters of the multiple regression statistical model; according to the obtained fitted value of the slope physical quantity, calculate the center line X CL of the quality control chart, and the center line X CL is the fitted value of the multiple regression statistical model, and calculate the upper and lower control lines X LCL1 , X UCL1 and the secondary upper and lower control lines X LCL2 , X UCL2 respectively; according to the calculated center line X CL and the two-level control lines, generate the quality control chart of slope physical quantities; among them, the horizontal axis of the quality control chart is time, and the vertical axis is the slope physical quantity, and the quality control chart is used to reflect the change trend of the slope physical quantity and the slope safety state; for a newly obtained set of slope physical quantity monitoring data, use the obtained multiple regression statistical model to calculate the fitted value of the corresponding physical quantity, and use the normal distribution function to calculate the risk rate P f ' of the current slope; according to the calculated risk rate P f ', the set risk rate thresholds P f1 ' and P f2 ', and the quality control chart, judge the safety state of the measuring point at the current moment; according to the judgment result, give the corresponding early warning prompt information.

[0008] Further, use the historical monitoring data of slope physical quantities and environmental factor data to establish a multiple regression statistical model between slope physical quantities and environmental factors, including: The expression of the multiple regression statistical model is as follows:

[0009]

[0010] Among them, represents the water pressure component, H represents the upstream water depth, and x, y, z represent the spatial coordinate components in three directions of the monitoring point; k, l, m, n respectively represent the influence coefficients of the water pressure component in the x, y, z directions, reflecting the influence degree of the upstream water depth and the monitoring point position on the water pressure component; represents the rainfall component, P w represents the rainfall amount; represents the temperature component, and t represents the cumulative number of days from the monitoring date to the starting measurement date; represents the aging component, and θ = t / 100 represents the dimensionless time.

[0011] Furthermore, to solve the parameters (A i , B i , C i , D i ) of the multiple regression statistical model, it includes: setting the objective optimization function as follows:

[0012] Q = ∑[X(H, P, T, θ, x, y, z) - f(H, P, T, θ, x, y, z)] 2 , and by taking the partial derivatives of the objective function, the model parameters (A i , B i , C i , D i ) are obtained: Furthermore, using the normal distribution function to calculate the risk rate P f ' of the current slope, it includes: performing standard normalization processing on the reservoir bank slope monitoring sequence {Xi} to form a standard normalization sequence that follows N(0, 1) where, σ = S, S represents the standard deviation of the multiple regression model; ^X_i represents the fitted value of the reservoir bank slope monitoring sequence; according to the target reliability β t set the attribute judgment threshold ±β′, and generate the safety interval (-β′, β′) and the risk interval (-∞, -β′] ∪ [β′, +∞) according to the standard normal distribution curve; where: (-∞, -β′] is the minimum value interval of the risk interval, and [β′, +∞) is the maximum value interval of the risk interval; calculate the real-time risk rate P f ′ of the measured value of the measuring point: where, P rf represents the allowable risk rate of the preset project; is the interval distribution probability of the measured value X i ; Φ is the standard normal distribution function.

[0013] Furthermore, according to the obtained fitted values of the slope physical quantities, calculate the center line XCL of the quality control chart. The center line XCL is the fitted value of the multiple regression statistical model, and calculate the upper and lower control lines X LCL1, X UCL1 and the secondary upper and lower control lines X LCL2 , X UCL2 , including: Among them, represents the simulated fitting value, X i+1 represents the measured value at the measuring point, and S represents the standard deviation of the regression statistical model.

[0014] Furthermore, to judge the safety status of the measuring point at the current moment, including: for the real-time risk rate P f ' of the measuring point i, judge P f ' whether it is less than the first-level warning index P f1 ', if so, judge that the measuring point i is in a safe state; if P f ' is greater than P f1 ', then judge whether P f ' exceeds the first-level upper and lower control lines X LCL1 or X UCL1 , if so, judge that the measuring point is in a warning state; when the measuring point P f ' is greater than P f2 ', or P f ' exceeds the secondary upper and lower control lines X LCL2 or X UCL2 , judge that the measuring point is in an emergency state. Furthermore, P f1 ' is calculated by the following formula: P f1 ' = 1 - Φ(βt1), where Φ is the standard normal distribution function and βt1 is the reliability index corresponding to the first-level warning index.

[0015] Furthermore, the value range of βt1 is 3.7 to 4.2.

[0016] Furthermore, P f2 ' is calculated by the following formula: P f2 ' = 1 - Φ(βt2), where Φ is the standard normal distribution function and βt2 is the reliability index corresponding to the second-level warning index.

[0017] Furthermore, the value range of βt2 is 3.2 to 3.7.

[0018] Compared with the prior art, the advantages of this application are:

[0019] By obtaining the historical monitoring data of the slope physical quantities and the environmental factor data affecting the slope stability, a multiple regression statistical model between the slope physical quantities and the environmental factors is established, comprehensively considering multiple physical quantity indexes such as slope displacement, seepage pressure, reservoir water level, etc. and environmental factors such as rainfall, temperature, time effect, etc., improving the accuracy and applicability of the warning model.

[0020] When establishing a multiple regression statistical model, the water pressure component, rainfall component, temperature component, and aging component are introduced, fully considering various factors affecting the slope stability, and improving the fitting degree and prediction ability of the model. Solving the model parameters through the objective optimization function ensures the robustness and reliability of the model.

[0021] The risk rate of the current slope is calculated using the normal distribution function, and a risk rate index is introduced to quantitatively evaluate the risk of the reservoir bank slope. By setting the target reliability index associated with the engineering safety level, combining the risk rate with the reliability index makes the evaluation result more reasonable and credible, providing a quantitative basis for the early warning of the reservoir bank slope.

[0022] According to the fitting values of the slope physical quantities, the center line and classification control lines of the quality control chart are calculated, generating a quality control chart that intuitively reflects the change trend of the slope physical quantities. Through the quality control chart, the abnormal changes of the slope physical quantities can be monitored in real time, providing an intuitive judgment basis for early warning analysis.

[0023] Based on the normal distribution theory, the real-time risk rate is calculated. At the same time, two-level early warning indicators are set in combination with the center line and classification control lines of the quality control chart, realizing the automatic identification and classification early warning of the slope safety state. Combining the risk rate index with the quality control chart improves the sensitivity and reliability of the early warning.

[0024] For the calculated risk rate and quality control chart, clear slope state discrimination conditions are set. By comparing with the early warning indicators, the safety state of the slope can be automatically judged, and corresponding early warning prompt information can be given, realizing the real-time early warning and risk prompt of the reservoir bank slope. Brief Description of the Drawings

[0025] This application will be further described in the form of exemplary embodiments, which will be described in detail through the drawings. These embodiments are not restrictive. In these embodiments, the same numbers represent the same structures, where:

[0026] Figure 1 is a schematic diagram of the relationship between the reliability β and the risk rate P shown in some embodiments of this application f ;

[0027] Figure 2 is the probability distribution of the measured values of the slope operation shown in some embodiments of this application

[0028] Figure 3 is a schematic diagram showing that the measured value of the measuring point is in a safe state shown in some embodiments of this application;

[0029] Figure 4 is a schematic diagram showing that the measured value of the measuring point is in a risk state shown in some embodiments of this application;

[0030] Figure 5 It is a schematic diagram showing that the measured values of the measurement points are in a critical state according to some embodiments of the present application;

[0031] Figure 6 It is a schematic diagram of the allowable risk rate when the target reliability is 3.7 according to some embodiments of the present application;

[0032] Figure 7 It is a schematic diagram of the allowable risk rate when the target reliability is 2.7 according to some embodiments of the present application;

[0033] Figure 8 It is a schematic diagram for judging the attribute of the measured value according to some embodiments of the present application;

[0034] Figure 9 It is an exemplary flowchart of the method for quantifying the real-time risk rate of slope operation according to some embodiments of the present application;

[0035] Figure 10 It is a typical quality control chart according to some embodiments of the present application;

[0036] Figure 11 It is a schematic diagram of the warning level of the slope project according to some embodiments of the present application;

[0037] Figure 12 It is a schematic diagram of the method for dynamically warning the risk rate of slope operation according to some embodiments of the present application;

[0038] Figure 13 It is a quality control chart of the horizontal displacement measurement points in Area I according to some embodiments of the present application;

[0039] Figure 14 It is a quality control chart of the vertical displacement measurement points in Area I according to some embodiments of the present application;

[0040] Figure 15 It is the horizontal displacement of the real-time risk rate quantification model of slope operation according to some embodiments of the present application;

[0041] Figure 16 It is the vertical displacement of the real-time risk rate quantification model of slope operation according to some embodiments of the present application. Detailed implementation manners

[0042] The methods and systems provided by the embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0043] Considering the clustering results of measurement points, a spatio-temporal regression statistical model is established. On this basis, a quantitative model for the real-time risk rate of slope operation is constructed based on monitoring data, and the risk characteristics of slope operation are measured by the determined probability calculation value. Considering that the traditional method of calculating the risk rate of slope operation based on the theory of probabilistic reliability has been clearly calibrated by specifications, the internal relationship and transformation method between the real-time risk rate and the operation risk rate are explored, so as to provide a quantitative evaluation standard for slope projects with different safety levels according to the requirements of the specifications. For the spatio-temporal regression model of the monitoring effect quantity of the reservoir bank slope, most of the traditional models that construct the functional relationship between the monitoring information and influencing factors based on the measured data of the slope are mainly based on single measurement points, and the connection between multiple measurement points is insufficiently considered, making it difficult to reflect the load response of the slope from an overall perspective. Therefore, the spatial coordinates of the measurement points are taken into account, and considering influencing factors such as environmental load, monitoring time, and measurement point location, a spatio-temporal regression model of the monitoring effect quantity of the reservoir bank slope is constructed to dynamically monitor the change trend of the monitoring effect quantity of the reservoir bank slope under any load at any time.

[0044] According to the causes, the composition of the monitoring effect quantity of the reservoir bank slope can be divided into four parts: water pressure, rainfall, temperature, and aging components. Thus, the spatio-temporal regression statistical model is established as follows: X = X H + X P + X T + X θ + X0 + e, where: X is the estimated value of the monitoring point at a certain moment, X H , X P , X T , X θ represent the water pressure, rainfall, temperature, and aging components of the monitoring effect quantity respectively, X0 is the constant term, and e is the residual term.

[0045] Water pressure component. In the spatio-temporal regression statistical model of the monitoring effect quantity of the reservoir bank slope, the expression of the water pressure component is: X H = f1[f(H), f(x, y, z)] = f1[H, x, y, z], where f(H) is the water pressure component of a single measurement point, usually expressed as a power polynomial: where: ai is the regression coefficient of the water pressure component, and H is the upstream water depth. f(x, y, z) represents the displacement field generated by the slope under a certain water level. Since f(x, y, z) is continuous in the region Ω, it is represented by a multivariate power series: where: x, y, z are the spatial coordinate variables in three directions of the measurement point. The expression of the water pressure component is obtained:

[0046] Rainfall component. In the spatio-temporal regression statistical model of the monitoring effect quantity of the reservoir bank slope, the expression of the rainfall component is: X P= f2[f(P), f(x, y, z)] = f2[P, x, y, z], where f(P) is the rainfall component at a single measurement point, usually expressed by the following formula: In the formula: b w is the regression coefficient of the rainfall component, and P w are the average rainfall amounts on the monitoring day, 2 days before the monitoring day, 3 to 4 days before the monitoring day, 5 to 15 days before the monitoring day, and 16 to 30 days before the monitoring day, respectively. f(x, y, z) represents the displacement field generated by the slope at a certain water level, and the rainfall component expression is obtained:

[0047] Temperature component. In the spatio-temporal regression statistical model of the monitoring effect quantity of the reservoir bank slope, the expression of the temperature component is: X T = f3[f(T), f(x, y, z)] = f3[T, x, y, z], where f(T) is the temperature component at a single measurement point, usually expressed by the sine-cosine composite periodic function of formula (5.3.10): In the formula: c 1i and c 2j are the regression coefficients of the temperature component; t is the cumulative number of days from the monitoring day to the starting measurement day. f(x, y, z) represents the displacement field generated by the slope at a certain water level, and the temperature component expression is obtained:

[0048] Aging component. In the spatio-temporal regression statistical model of the monitoring effect quantity of the reservoir bank slope, the expression of the aging component is: X θ = f4[f(θ), f(x, y, z)] = f4(θ, x, y, z), where f(θ) is the aging component at a single measurement point, usually expressed by the following formula: f(θ) = d1θ + d2lnθ. In the formula: d1 and d2 are the regression coefficients of the aging component, and θ = t / 100. Similarly, we get: Substitute them into the spatio-temporal regression statistical model of the monitoring effect quantity of the reservoir bank slope respectively, and we get:

[0049]

[0050] Use the least squares method to solve the spatio-temporal regression statistical model parameters (A i , B i , C i , D i ). Set the target optimization function as follows: Q = Σ[X(H, P, T, θ, x, y, z) - f(H, P, T, θ, x, y, z)] 2 . To make the sum of squared residuals reach the minimum, it is necessary to satisfy: Obtain the model parameters (A i , B i , C i , Di ) and substituting the model parameters solved by the least squares method into the spatio-temporal distribution model expression of the monitoring effect quantity of the reservoir bank slope, the fitting values of the mathematical model at each measuring point can be calculated and obtained.

[0051] Based on the slope operation risk rate of the probability reliability theory, the calculation of the slope risk rate and reliability accompanies each other. Among them, the probability that the slope completes the predetermined function is commonly represented by the reliability P r denoted, while the probability that the slope cannot complete the predetermined function is denoted by the risk rate P f denoted, and the two are mutually independent incompatible events, and there is P r +P f = 1.

[0052] Assume that a certain function of the slope is related to n basic random variables X1, X2, …, X n , then the slope function Z can be expressed as Z = g(X1, X2, …, X n ). Z > 0 is the reliable state of the slope, Z < 0 is the failure state of the slope, and Z = 0 is the limit state of the slope. Based on the principle of probability conservation, in a general stochastic dynamic system, the random variables can be transformed into partial differential equations for time-evolution analysis through the probability density function of the random response, that is, the slope risk rate can be calculated according to the probability distribution form of the random variables, namely:

[0053] Assume that the slope load is S and the resistance is R, then the slope function Z can be comprehensively expressed by the load S and the resistance R as: Z = R - S. R > S is the reliable state of the slope, R < S is the failure state of the slope, and R = S is the limit state of the slope. At this time, the slope operation risk rate formula is: Considering that S and R are independent random variables, the risk rate formula is rewritten as: In the formula: f R (r) is the probability density function of the resistance R, f S (s) is the probability density function of the load S, f RS (r, s) is the joint probability density function of R and S, F RS (r, s) is the joint cumulative distribution function, and Ω f is the failure domain, representing R ≤ S.

[0054] Based on the probability reliability theory, calculating the slope risk rate mostly adopts probability distribution forms such as normal distribution, uniform distribution, extreme value distribution, and triangular distribution. Among them, the normal distribution is a two-parameter probability distribution form, with characteristics such as symmetry, centrality, and uniform variability, and its probability density function is: If the normal distribution form is used to describe the slope function Z, the corresponding mean is μ z , and the standard deviation is σ z, at this time the slope operation risk rate P f exist Figure 1 is based on the reliability β and risk rate P shown in some embodiments of the present application f The relationship diagram is shaded and the expression is written as: In addition, since the numerical result of the slope operation risk rate calculation is small, it brings certain difficulties to practical application. Reliability index is often used to quantitatively evaluate the operation risk of the structure, and β = μ z / σ z , and obtain the reliability β and risk rate P f The relationship is as follows: Commonly used reliability β and risk rate P f The comparison relationship is shown in Table 1.

[0055] Table 1 Reliability β and risk rate P f Comparison table

[0056] β 1.00 2.00 3.00 3.50 4.00 4.50 <![CDATA[P f > 1.59×10-1 2.27×10-2 1.35×10-3 1.04×10-4 2.32×10-5 3.40×10-6

[0057] Although the risk rate P based on the probability reliability theory f It is consistent with the design principles of slope engineering and gives the reservoir bank slope a probabilistic meaning. However, it is essentially still based on the traditional limit equilibrium method. It measures and evaluates the failure state of the slope on the basis of solving the slope safety factor. It is unable to calculate the slope operation risk rate in real time based on the continuously updated monitoring data. There are certain subjective limitations in practical applications. Therefore, this paper considers an intuitive method to quantify the real-time risk rate of slope operation from the perspective of measured data.

[0058] The real-time risk rate quantitative model of slope operation based on monitoring data has many aspects that affect the risk of slope system failure. The most important and basic characteristics are environmental quantities such as water level, rainfall and temperature upstream of the slope, as well as effect quantities such as deformation and seepage. Their changing laws can reflect the structural state of the slope itself and the characteristics and changing laws of external environmental loads. Therefore, it is of great significance to select the above characteristics as independent variables to establish various slope risk rate analysis models. This section combines the measured data of the slope to construct a real-time risk rate quantitative model of slope operation based on the normal distribution form.

[0059] First, input the bank slope monitoring sequence {X i}, i = 1, 2, ..., n, combined with the measurement point partitioning results based on spatiotemporal data mining in the previous article, for different regional measurement points, the corresponding reservoir water level {H i}、Rainfall {P i}、Temperature{T i}、Time {θ i}, and the measuring point coordinates x, y, z, the spatiotemporal regression model of the reservoir bank slope monitoring effect quantity is established in different regions.

[0060] In the field of hydraulic safety monitoring, before conducting regression analysis on monitoring data, there is a basic assumption that the observation error follows a normal distribution. Therefore, the following combines the probability density function of the normal distribution and uses the measured value X of the measuring point i , the model fitting value and the model standard deviation S to calculate the real-time risk rate P' of the measured value of this measuring point f . Where: X i is the measured value of the measuring point at time i; is the model fitting value corresponding to time i; S is the standard deviation of the spatio-temporal regression statistical model; A is the risk rate correction coefficient; is the interval distribution probability of the measured value X i .

[0061] According to the constructed normal distribution curve with σ = S, as Figure 2 shown by the probability distribution of the measured values of slope operation shown in some embodiments of the present application shown, the distribution probability of the measuring point X i in the interval in the interval in the interval

[0062] Considering that the risk rate P based on the probability reliability theory f has been clearly specified in the current standards, it is possible to study the relationship between the risk rate P f and the real-time risk rate P' f and refer to the existing engineering specifications to quantitatively design the warning index of the real-time operation risk rate of the slope. The following discusses the relationship and transformation between the risk rate P and the real-time risk rate P' f from aspects such as the interval distribution probability f the risk rate conversion coefficient B, and the attribute judgment threshold β'.

[0063] Interval distribution probability First, discuss the relationship between the interval distribution probability and the risk rate P f in combination with Table 2. The interval distribution probability of the measured value of the measuring point is the area of the diagonal region shown in the figure, and the risk rate P f of the slope engineering operation is the area of the red region shown in the figure. The state of the measured value X i is different, corresponding and P f , X i and The relational expressions between them are different. The judgment relational expressions corresponding to the measured value safety state, risk state, and critical state are as follows. For example, Figures 3 to 5 :

[0064] Table 2 Judgment relational expressions for the measured value state of measuring points

[0065]

[0066] Based on the above relationships, the risk rate P formulated based on the probability reliability theory can be realized f and the interval distribution probability of the measured values of the measuring points can be transformed. It can be seen from the above table that when the measured value of the measuring point is in the critical state between safety and risk, there is If the risk rate P at this time f is equal to the allowable risk rate P of the project rf the interval distribution probability and the allowable risk rate P rf can be obtained, and the transformation relationship is: Therefore, the calculation formula for the risk rate correction coefficient A is: And it can be rewritten as: Calculate the real-time risk rate P′ f . When P′ f > 1, it indicates that the measured value of the measuring point is in the risk state; when P′ f < 1, it indicates that the measured value of the measuring point maintains the safety state; when P′ f = 1, it indicates that the measured value of the measuring point is in the critical state between risk and safety.

[0067] The risk rate conversion coefficient B. To achieve the unified measurement of slope projects with different safety levels, the risk rate conversion coefficient B is further introduced to convert and compare the risk rates of projects with different safety levels. In the existing engineering specifications, the target reliability β t and the allowable risk rate P rf provide a clear evaluation standard for quantitatively measuring the ultimate limit state of the continuous bearing capacity of different projects. Referring to the "Unified Standard for Reliability Design of Water Conservancy and Hydropower Engineering Structures (GB50199-2013)", the target reliability β t and the allowable risk rate P rf of water conservancy and hydropower projects with different safety levels are summarized in Table 3. Among them, type I failure is a non-sudden failure with obvious omens before failure, and type II failure is a sudden failure without obvious omen information.

[0068] Table 3 Target reliability β t and allowable risk rate P rf

[0069]

[0070] Taking slope projects with safety levels of first and third grades as examples, for a first-grade slope project, the target reliability β t1 is 3.7, and the allowable risk rate P rf1 is 1.0×10-4. For a third-grade slope project, the target reliability β t3 is 2.7, and the allowable risk rate P rf3 is 3.4×10-3. When the project safety level is higher, the target reliability is higher, and the corresponding allowable risk rate is lower.

[0071] In Figure 6 and Figure 7 for an intuitive comparison, the distribution probabilities of the measured values calculated in the left and right figures are equal, and the areas of the corresponding measured value distribution intervals are equal, which are all represented by the shaded parts in the figures. When the target reliability is larger, the area of the corresponding measured value safety interval is larger, which is all represented by the blue parts in the figures. It can be seen that for different projects, even if the distribution probability of the measured values is certain, the measured project risk characteristics are not the same. Therefore, to uniformly measure the risk characteristics of slope projects with different safety levels, it is necessary to define a risk rate conversion coefficient B to convert and compare the real-time risk rates of projects with different safety levels. Here, taking slope projects with safety levels of first and third grades as examples again, the calculation method of the risk rate conversion coefficient B between the two is as follows:

[0072] When a third-grade slope project needs to be compared with a first-grade slope project, then multiply the original real-time risk rate P′ f by the calculated risk rate conversion coefficient B. Combining Table 4, the risk rate conversion coefficients B between slope projects with different safety levels are sorted out and summarized as shown in Table 4.

[0073] Table 4 Risk rate conversion coefficient B between projects with different safety levels

[0074] Risk rate conversion coefficient B Level 1 Level 2 Level 3 Level 1 1 0.998 660 0.993 220 Level 2 1.001 342 1 0.994 553 Level 3 1.006 826 1.005 477 1

[0075] The attribute judgment threshold β', because the random variable X i follows a non-standard normal distribution form, it is difficult to uniformly measure the risk attributes of the measured values at each measuring point of the slope. Therefore, let σ = S, and perform standard normal transformation on the monitoring sequence {X i} of the reservoir bank slope: Thus, a standard normal sequence {Z i} that follows N(0,1) is formed, and the sequence {Z i} is mapped to the Figure 8 standard normal distribution curve shown. At the same time, the real-time risk rate calculation formula is: In the formula, when Z i = β t , P' f = 1, and the measured value is in the critical state between safety and risk; when Z i > βt t , P' f > 1, which real-time characterizes the occurrence of risk in the measured value. Therefore, the target reliability βt t specified in the code can be referred to set the attribute judgment threshold ±β', and thus generate the safety interval (-β', β') and the risk interval (-∞, -β'] ∪ [β', +∞), where: (-∞, -β'] is the minimum value interval of the risk interval, and [β', +∞) is the maximum value interval of the risk interval.

[0076] As Figure 8 shown, by comparing the sequence value Z i and the attribute judgment threshold ±β', the safety and risk attributes of the measured values at each measuring point can be intuitively judged: when -β' ≤ Z i ≤ β, that is, Z i ∈[-β', β'], the measured value is called a safe measured value; when Z i < -β' or Z i > β', Z i ∈(-∞, -β') ∪ (β', +∞), the measured value is called a risk measured value.

[0077] Furthermore, mapping all the measured values of the measuring points on the standard normal distribution curve and combining with the attribute judgment threshold ±β can not only intuitively judge the safety state of the measured values of the measuring points, but also overall measure the risk characteristics of the monitoring project.

[0078] Now, combined with Figure 9 the method flow chart, the main steps of the method are introduced as follows: spatially and temporally construct a regression statistical model for the monitoring effect quantity of the reservoir bank slope in partitions, and calculate the model fitting value i corresponding to the measured value X of the measuring point. According to the safety level of the slope project, select the corresponding target reliability β t and the allowable risk rate P rf ; use the measured value X i of the measuring point, the model fitting value and the model standard deviation S, and combine with the slope target reliability β t and the allowable risk rate P rf to calculate the real-time risk rate P' f of the measured value of this measuring point; perform standard normalization processing on the input sequence {X i} to generate the standard normal sequence {Z i}; refer to the target reliability β tSet the attribute judgment threshold ±β′, and generate the safety interval and risk interval; compare the sequence value Z i The safety and risk attributes of the measured values ​​at each measuring point can be intuitively judged based on the attribute judgment threshold ±β′. When it is judged as a risk measurement, the real-time risk rate of the risk measurement is output. Mapping the measured values ​​of all measuring points of the same monitoring project on the standard normal distribution curve can not only intuitively judge the safety status of the measured values ​​of the measuring points, but also measure the risk characteristics of the monitoring project as a whole.

[0079] A dynamic early warning method for slope risk rate based on quality control chart is proposed based on the real-time risk rate quantification model of slope operation and the theory of quality control chart. It is suitable for the operation monitoring process of high slope on reservoir bank. The status of measuring points can be judged in time according to the continuously updated measured data, and the actual operation situation of slope can be diagnosed online, providing an effective way for risk early warning of slope engineering.

[0080] Quality control chart theory, quality control chart theory was first applied in factory production. It is a graphical method that uses statistical hypothesis principles to screen out monitoring data in a controlled state, and then discover abnormal conditions in the production process, ultimately achieving the purpose of quality control.

[0081] Typical quality control charts are as follows: Figure 10 As shown. The three straight lines from top to bottom in the figure are the upper control line UCL, the center line CL and the lower control line LCL, and the middle broken line is the measurement line formed by the measured values ​​of the measuring points arranged in chronological order from left to right. In the quality control chart, the measured values ​​will most likely be within the envelope range of the upper and lower control lines, and will be randomly distributed on both sides of the center line. On the contrary, when the measured values ​​exceed the envelope range of the upper and lower control lines, or the measured values ​​show significant regularity on both sides of the center line, it indicates that an abnormal situation has occurred. The reason is that when there are abnormal factors, the original random distribution law of the measured values ​​may be destroyed, the overall distribution center of the measured values ​​may also shift, or the standard deviation of the measured values ​​may change, resulting in the abnormal situation that the measured value distribution shows a significant regularity and the measured value exceeds the control line range.

[0082] The control objectives of traditional quality control charts are often rigid, that is, the positions of the upper and lower control lines and the center line in the control chart are relatively fixed. However, for slope projects in actual operation, there is often no clear rigid requirement for risk rate warning, and the only standard is to meet flexible needs.

[0083] The dynamic early warning method of slope operation risk rate based on quality control chart, taking into account the actual operation of slope engineering, the following article intends to improve the quality control chart theory. First, the warning level of slope engineering operation is reasonably divided, such as Figure 11 As shown in the figure, the warning level of slope engineering can be divided into three levels: safety state, warning state and emergency state.

[0084] Among them: the safe state refers to the stable and safe state of the slope project; the warning state refers to the unstable and critical safety state of the slope project; the emergency state refers to the extremely unstable safety emergency state of the slope. According to different warning states of the slope, corresponding risk rate warning indicators are quantitatively designed. Among them, the risk rate warning indicator for distinguishing the safe state and the warning state of the slope project is P′ f1 , named the first-level warning indicator; the corresponding risk rate warning indicator for distinguishing the warning and emergency states of the slope project is P′ f2 , named the second-level emergency indicator.

[0085] On the basis of dividing the warning levels, how to scientifically and reasonably distinguish the safe state, warning state and emergency state of the slope project becomes the key issue. Here, combined with Table 3, considering that the reservoir bank high slope is a project with a first-level safety level, the limit state of the slope project is further subdivided by using the first type of failure and the second type of failure; at the same time, αβ t is selected as the reliability index for the state assessment of the in-service slope project, and the reduction coefficient α = 0.85 is selected with reference to "Theory and Application of Engineering Structure Reliability". Therefore, when distinguishing the safe state and the warning state of the reservoir bank high slope, the reliability index αβ t1 = 3.7×0.85≈3.0, that is, the allowable risk rate P rf = Φ(3.0) = 1.35×10 -3 , and the corresponding risk rate correction coefficient A is 0.9973; when distinguishing the warning state and the emergency state of the slope, the reliability index αβ t2 = 4.2×0.85≈3.5, that is, the allowable risk rate P rf = φ(3.0) = 1.04×10 -4 , and the corresponding risk rate correction coefficient A is 0.999792. Thus, the two-level real-time risk rate indicators P′ f1 and P′ f2 can be calculated.

[0086] Now, combined with Figure 12 , the principle of the dynamic warning method for the slope operation risk rate is introduced as follows: In the training area, assume that there is a certain measurement point sequence {X i}, i = 1, 2, …, n, where n is the time series length. Based on the historical measured data of the measurement points, a spatio-temporal regression statistical model is constructed to obtain the multiple correlation coefficient R and the standard deviation S of the mathematical model. Among them, the measurement points with a multiple correlation coefficient R>0.8 have a higher fitting accuracy, which helps to improve the accuracy and reliability of the subsequent risk rate calculation results. Therefore, when it is judged that the multiple correlation coefficient R of the measurement point >0.8, the historical measured values and model fitting values of this measurement point are input into Figure 12 (a) The model training area shown.

[0087] Judgment area: As the measured value is updated, input the measured value X of the measuring point i+1 and the corresponding reservoir water level H i+1 , rainfall P i+1 , temperature T i+1 , time θi i+1 and the measuring point coordinates x, y, z, and use the above spatio-temporal regression statistical model to calculate the model prediction value of the measuring point And input the updated measured value and model prediction value of this measuring point Figure 12 into the model judgment area shown in (c).

[0088] Control area: Generate the center line X according to the model fitting value CL , and combine the target reliability index βt corresponding to the two-level warning index t1 and βt t2 , and generate the upper and lower control lines X LCL1 , X LCL2 , X UCL1 and X UCL2 respectively, as follows: At the same time, use the model fitting value to update the measured value X i+1 and the standard deviation S of the regression statistical model, and calculate the real-time risk rate P′ of the measuring point value fi .

[0089] Reference Figure 12 to the model control area shown in (b), use the measured value center line and the two-level upper and lower control lines to form a quality control chart for slope operation risk rate warning, and combine the real-time risk rate P′ of the measuring point fi , and the current safety status, warning status or emergency status of the measuring point can be visually judged. The judgment method is as follows:

[0090] When the real-time risk rate P′ of the measuring point fi does not exceed the first-level warning index P′ f1 , that is, the measured value is within the envelope area of the first-level upper and lower control lines, indicating that the measuring point is in a safe state;

[0091] When the real-time risk rate P′ of the measuring point fi exceeds the first-level warning index P′ f1 , it means that in the quality control chart, when the measured value exceeds the first-level upper and lower control lines X LCL1 or X UCL1 , the measuring point is in a warning state, and the change trend of this measuring point needs to be closely monitored;

[0092] When the real-time risk rate P′ of the measuring point fi exceeds the second-level emergency index P′ f2 , it means that in the quality control chart, when the measured value exceeds the second-level upper and lower control lines X LCL2 or X UCL2When the measuring point is in an emergency state, it is necessary to immediately take engineering measures to prevent and control the risks of this measuring point.

[0093] Engineering example

[0094] For the spatio-temporal regression statistical model of slope displacement, considering the distance of the spatial position of the measuring points and the similarity degree of the attribute characteristics of the measuring points, a spatio-temporal data mining algorithm is used to partition the slope displacement measuring points. Now, based on the measuring point partition results based on the partitioning clustering algorithm in the previous text, the spatial coordinates (x, y, z) of each measuring point, the water depth H, rainfall P, temperature T, and time θ corresponding to the monitoring effect quantity X are used to establish a spatio-temporal regression statistical model based on the actual measured data of the slope in different partitions. The model fitting accuracies of the displacement measuring points in each area are shown in Table 5 - Table 7:

[0095] Table 5 Modeling accuracy of the spatio-temporal regression statistical model of displacement measuring points in Area Ⅰ

[0096]

[0097]

[0098] Table 6 Modeling accuracy of the spatio-temporal regression statistical model of displacement measuring points in Area Ⅱ

[0099]

[0100]

[0101] Table 7 Modeling accuracy of the spatio-temporal regression statistical model of displacement measuring points in Area Ⅲ

[0102]

[0103]

[0104] It can be seen that the multiple correlation coefficients R of the mathematical models of the displacement measuring points in Area Ⅰ are all greater than 0.9262, those in Area Ⅱ are all greater than 0.8762, and those in Area Ⅲ are all greater than 0.9026. Through the model parameters, it can be judged that: the fitting results of the spatio-temporal regression models established based on the monitoring data of the displacement measuring points in the three areas are true and reliable, which helps to improve the effectiveness and credibility of the subsequent risk rate warning indicators established based on the fitting results of the mathematical models.

[0105] For the dynamic warning of the slope risk rate based on the quality control chart, on the basis of constructing the spatio-temporal regression statistical model of the slope, then the dynamic warning method of the slope risk rate based on the quality control chart proposed in this paper is adopted, and the monitoring data of the measuring points in Area Ⅰ with the highest overall credibility of the model are selected for analysis.

[0106] The following analysis is mainly based on the horizontal and vertical displacement monitoring data of the measuring points before December 31, 2018, and the mathematical model fitting values of each measuring point are fitted according to the spatio-temporal regression statistical model. The historical measured values and model fitting values of the measuring points with higher fitting accuracy are selected into the model training area; at the same time, the measured value data updated after January 1, 2019 of the measuring points are selected, and the mathematical model prediction values of each measuring point are predicted based on the spatio-temporal regression statistical model. And the updated measured values and model prediction values of the measuring points are selected into the model judgment area.

[0107] Based on the quality control chart theory, according to the mathematical model fitting values Generate the center line X CL , combined with the target reliability index β corresponding to the two-level early warning index T1 and β T2 , successively generate the upper and lower control lines X of the first level and the second level CL , X LCL1 , X LCL2 , X UCL1 and X UCL2 , arrange the measuring point data in chronological order from left to right, and then the quality control chart for slope operation risk rate warning can be drawn. Among them, the quality control charts of the horizontal and vertical displacement measuring points in Area I are respectively as Figure 13 , 14 shown.

[0108] In the figure: on the left side are the model training areas drawn based on the historical measured values and model fitting values of each measuring point, and on the right side are the model judgment areas drawn based on the updated measured values and model prediction values of each measuring point; the blue dotted line represents the first-level control lines X LCL1 and X UCL1 , the purple dotted line represents the second-level control lines X LCL2 and X UCL2 , the red scatter line is the model fitting value of the measuring point, also known as the center line X CL of the measured value, and the black scatter line is the actual monitoring value of the measuring point.

[0109] At present, the historical measured values, fitting values, updated measured values, and prediction values of the horizontal and vertical displacement measuring points in Area I basically fall within the envelope area of the first-level upper and lower control lines, indicating that the measuring points in Area I are overall in a safe state, and the changes in the measured values of each measuring point are relatively stable. Therefore, the calculation of the real-time risk rate of slope operation is not elaborated in detail. Only taking the measuring point Jiao 16 as an example, the real-time risk rate is quantified by using the measured value X i of the measuring point, the model fitting value and the model standard deviation S.

[0110] Based on the displacement monitoring data of the horizontal and vertical directions of the Jiao 16 measuring point, a quantitative model for the real-time risk rate of slope operation is constructed, as Figure 15and 16 As shown in 16 . Among the monitoring data of the horizontal displacement at the 16 measuring points of the intersection, the measured value on June 2, 2003 had the greatest risk, and the distribution probability of the measured value reached 97.66%. The calculated real-time risk rate P′ f was 0.979 < 1, lower than the first-level warning index; the time when the measured value of the vertical displacement had the greatest risk occurred on March 31, 1998, and the distribution probability of the measured value was 73.41%. The real-time risk rate P′ f was 0.736 < 1, lower than the first-level warning index. It can be seen that the measured values of the horizontal displacement and vertical displacement of the measuring points have always been in a safe state, and the time when the measured value had the greatest risk was relatively far from now. According to the updated measured values after 2019, the real-time risk rate was calculated, and the distribution probability of the measured value remained between 30% and 40%. The calculated real-time risk rate was much lower than the first-level warning index and the second-level emergency index. Thus, the safety attribute of the measured value of the measuring point can be judged.

[0111] Mapping the measured value of a single measuring point on the standard normal distribution curve and combining with the attribute judgment threshold β′ can intuitively judge the safety state of the measured value of the measuring point; mapping the measured values of multiple measuring points of the same monitoring project on the same standard normal distribution curve can also comprehensively measure the risk characteristics of the monitoring project.

[0112] The real-time risk rate quantification model of the slope relies on the monitoring data, can calculate the corresponding risk rate in real time according to the updated measured values of the measuring points, and at the same time combines the traditional method of calculating the operation risk rate of the slope based on the probability reliability theory, and reasonably quantifies the operation risk of the slope with reference to the engineering specifications. It should be noted that although there are currently no obvious abnormal phenomena of accelerating deformation at some measuring points, the displacement measured values have not yet converged, such as measuring points Jiao 9, Jiao 13, and Jiao 16, etc. It is necessary to continue to closely monitor them using the slope real-time risk rate quantification and risk rate dynamic warning methods proposed in this paper to achieve the purpose of online diagnosis of measuring point anomalies and dynamic warning of slope danger situations.

Claims

1. An early warning method for the reservoir bank slope, comprising: According to the safety level of the slope project, the slope state is divided into three warning levels: safe state, early warning state, and emergency state, and the first-level warning index P f1 ′ and the second-level warning index P f2 ′ are set; among them, P f1 ′ is the risk rate threshold for the early warning state, and P f2 ′ is the risk rate threshold for the emergency state; Obtaining historical monitoring data of slope physical quantities and environmental factor data affecting slope stability; wherein, the slope physical quantities include slope displacement, seepage pressure, and reservoir water level; the environmental factor data includes reservoir water level, rainfall, temperature, and monitoring time; Using the historical monitoring data of slope physical quantities and environmental factor data to establish a multiple regression statistical model between slope physical quantities and environmental factors, and obtaining the multiple correlation coefficient R and regression standard deviation S of the multiple regression statistical model; wherein, R reflects the goodness of fit of the multiple regression statistical model, and its value range is 0 to 1; S reflects the dispersion degree of the multiple regression statistical model; When R is greater than the threshold, use the multiple regression statistical model to calculate the fitted value of the slope physical quantity, and calculate the root mean square error RMSE between the fitted value and the measured value; when RMSE is less than the threshold, perform the next step, otherwise adjust the parameters of the multiple regression statistical model; Calculate the center line X of the quality control chart based on the fitted values of the slope physical quantities obtained CL , the center line X CL is the fitted value of the multiple regression statistical model, and calculate the upper and lower control lines X LCL1 , X UCL1 and the upper and lower control lines X of the second level LCL2 , X UCL2 ; According to the calculated center line X CL and two-level control lines, a quality control chart of slope physical quantities is generated; wherein, the horizontal axis of the quality control chart is time, and the vertical axis is the slope physical quantity, and the quality control chart is used to reflect the change trend of the slope physical quantity and the slope safety state; For a newly obtained set of monitoring data of slope physical quantities, the fitted values of the corresponding physical quantities are calculated using the obtained multiple regression statistical model, and the risk rate P of the current slope is calculated using the normal distribution function f ′; According to the calculated risk rate P f ′, the set risk rate threshold P f1 ′ and P f2 ′, as well as the quality control chart, judge the safety status of the measurement point at the current moment; According to the judgment result, give the corresponding early warning prompt information.

2. The early warning method for the reservoir bank slope according to claim 1, characterized in that: Using the historical monitoring data of slope physical quantities and environmental factor data to establish a multiple regression statistical model between slope physical quantities and environmental factors, including: The expression of the multiple regression statistical model is as follows: Wherein, Denote the water pressure component, H denote the upstream water depth, and x, y, z denote the spatial coordinate components in three directions of the monitoring point position; k, l, m, n respectively denote the influence coefficients of the water pressure component in the x, y, z directions, reflecting the influence degrees of the upstream water depth and the monitoring point position on the water pressure component; represents the rainfall component, P w represents the rainfall amount; represents the temperature component, and t represents the cumulative number of days from the monitoring date to the starting measurement date; Indicates the aging component, and θ = t / 100 represents the dimensionless time.

3. The early warning method for the reservoir bank slope according to claim 2, characterized in that: Solving for the parameters (A i , B i , C i , D i ) of a multiple regression statistical model, including: Set the target optimization function as follows: Q = ∑[X(H, P, T, θ, x, y, z) - f(H, P, T, θ, x, y, z)] 2 By taking the partial derivative of the objective function, the model parameters (A i , B i , C i , D i ) are obtained:

4. The early warning method for the reservoir bank slope according to claim 2, characterized in that: Calculating the risk rate P of the current slope using the normal distribution function f ', including: Perform standard normal transformation on the monitoring sequence {Xi} of the reservoir bank slope to form a standard normal sequence {Z i} that follows N(0,1): Among them, σ = S, S represents the standard deviation of the multiple regression model; ^X_i represents the fitted value of the monitoring sequence of the reservoir bank slope; According to the target reliability β t Set the property judgment threshold ±β′, and generate the safety interval (-β′, β′) and the risk interval (-∞, -β′] ∪ [β′, +∞) according to the standard normal distribution curve; where: (-∞, -β′] is the minimum value interval of the risk interval, and [β′, +∞) is the maximum value interval of the risk interval; Calculate the real-time risk rate P of the measured value at the measurement point f ': Among them, P rf represents the allowable risk rate of a preset project; is the interval distribution probability of the measured value X i ; Φ is the standard normal distribution function.

5. The early warning method for the reservoir bank slope according to claim 4, characterized in that: According to the fitted values of the slope physical quantities obtained, calculate the center line XCL of the quality control chart. The center line XCL is the fitted value of the multiple regression statistical model. Calculate the upper and lower control lines X LCL1 and X UCL1 for the first level, and the upper and lower control lines X LCL2 and X UCL2 for the second level. Including: Among them, represents the simulated fitting value, and X i+1 represents the measured value at the measuring point, and S represents the standard deviation of the regression statistical model.

6. The early warning method for the reservoir bank slope according to claim 5, characterized in that: Judging the safety state of the measuring point at the current moment, including: For the real-time risk rate P of measurement point i f ′, judge P f ' whether it is less than the first-level warning index P f1 ′. If so, it is judged that the measurement point i is in a safe state; If P f ' is greater than P f1 ′, then it is judged whether P f ' exceeds the upper and lower control lines of the first level X LCL1 or X UCL1 . If so, it is judged that the measuring point is in the warning state; When the measurement point P f ' is greater than P f2 ′, or P f ' exceeds the upper and lower control lines of the second level X LCL2 or X UCL2 When this occurs, it is determined that the measurement point is in an emergency state.

7. The early warning method for the reservoir bank slope according to claim 6, characterized in that: P f1 is calculated by the following formula: P f1 P' = 1 - Φ(βt1) Wherein, Φ is the standard normal distribution function, and βt1 is the reliability index corresponding to the first-level early warning index.

8. The early warning method for the reservoir bank slope according to claim 7, characterized in that: The value range of βt1 is 3.7 to 4.

2.

9. The early warning method for the reservoir bank slope according to claim 7, characterized in that: P f2 is calculated by the following formula: P f2 P' = 1 - Φ(βt2) Wherein, Φ is the standard normal distribution function, and βt2 is the reliability index corresponding to the second-level early warning index.

10. The early warning method for the reservoir bank slope according to claim 9, characterized in that: The value range of βt2 is 3.2 to 3.7.

Images