CN103605903B - Landslide time medium-short term prediction method - Google Patents

Abstract

The present invention relates to a kind of landslide time medium-short term prediction method, be utilize computer program that landslide is carried out medium-short term prediction, estimate the landslide play sliding time. With displacement dynamically for parameter, establishing landslide time medium-short term prediction multinomial model, this model structure is simple, workable, simulates effective, it is possible to achieve real-time tracing prediction of landslides. This medium-short term prediction is a kind of dynamic tracking forecast, in forecasting process, often obtains new observation data, just carries out the process of landslide parameter, substitutes into landslide medium-short term prediction system and forecasts, each monitoring point all can obtain determining solution. Namely forecast with the up-to-date variation characteristic that comes down, accomplish that real-time tracking forecasts, accurately and reliably, medium-short term prediction is had very strong adaptability, the efficiency solving slide prediction is low, reliability is relatively low and calculates the problems such as complicated, can better for the service of preventing and reducing natural disasters.

Description

Landslide time medium-short term prediction method

Technical field:

The present invention relates to a kind of Landslide Hazards forecasting procedure, an especially sliding-type landslide time medium-short term prediction.

Background technology:

Time Prediction of Landslide is to occur the acute sliding time for target, but also not all landslide can occur acute sliding. Saying of outline, landslide motion mode has three kinds: crawling type, interval crawling type and a sliding-type. The above two are typically not occurs play sliding, or a sliding distance is less, although the mankind and economic activity thereof also can be produced the harm that degree is different by them, but typically not causes bigger disaster. The present invention be directed to the medium-short term prediction on a sliding-type landslide.

Time Prediction of Landslide, if from 1969 Japanese scholars vegetarian rattan (M.Saito) publish thesis the 6th world, Montreal soil mechanics and foundation engineering meeting and to count, the history of existing 30 years so far. In recent ten years, since particularly carrying out environmental conservation and natural disaster reduction activity in 10 years, domestic and international many scholars have put into this research field, alleviate with expectation and prevent and treat landslide disaster. Although at present Chinese scholars has been presented for the theoretical model of dozens of slide prediction, but it is up to the present the real dependence successful example of quantitative forecast but without landslide. So-called " successful predicting " at present, mostly simply according to facing the micro-judgment done by sliding phenomenon, adopt the forecast that quantitative forecast model is done to landslide, be all almost some post-hoc tests bar none. From forecasting procedure, mainly there is following aspect:

Phenomenon and experimental forecast, mainly based on phenomenon, experimental forecast, the play sliding time on the macroscopic premonitory phenomenon speculative arbitration landslide before namely sliding according to some the deformation failure signs showed in the evolution process of landslide, play.Such as vegetarian rattan enlightening filial piety proposes creep rupture three-stage theory, and establish tertiary creep forecasting model (hillside plot firm two, cross just bright, island is clear to be controlled. landslide and slope failure and preventing and treating thereof. and " landslide and slope failure and preventing and treating thereof " translation group, translate. Beijing: Science Press, 1980,3:20-254 pages); HockE. according to the displacement monitoring-time graph on landslide, Chile Chuquicamata ore deposit, propose the form utilizing slope body deformability curve and trend carries out extension and inquires into epitaxy (5 rights of acute sliding time, Wang Niansheng. a kind of landslide displacement dynamic forecasting is inquired into. Chinese Geological Disasters and preventing and treating journal .1996,7:269-270 page). These methods are the qualitative methods that thought is taken in qualitative forecast and experimental forecast as the leading factor, it is the forecast directly perceived of landslide omen reflection experience accumulation, is the empirical equation set up under certain artificial condition, is theoretically unsound, the gliding mass creep play the tried to achieve sliding time belongs to budgetary estimate, and reliability is relatively low.

Displacement versus time statistical analysis forecasts, the modern mathematical theory such as join probability opinion, gray system theory, mathematical statistics opinion, fuzzy mathematics opinion establish multiple slide prediction model, for instance the comprehensive forecasting method that Wang Sijing proposes according to total deformation before the body unstability of slope and rate of displacement; Wang Lansheng, Chen Mingdong are according to the GM (1 in gray system theory, 1) model, by the Lycoperdon polymorphum Vitt GM (1 proposed that the matching of gliding mass displacement-time curve is extrapolated, 1) model (Chen Mingdong, Wang Lansheng. the gray prediction method of slope deformation destruction. see: Third National engineering geology conference collection of thesis); Yan Tongzhen is according to gliding mass evolution process and the biological Verhulst biological growth model (Tao Ganqiang breeding, develop, have similar features proposition, Ren Fengyu, Wang Xiaocun. gray theory Verhulst model is for the research of Landslide Prediction, mining industry research and development, 2002,22 (4): 11-13); The Verhulst inverse function model (Wang Nianqin that Li Tianbin proposes according to displacement quantization curve and the Verhulst inverse function curve more similarity of slope body deformability, Wang Yongfeng, Luo Donghai, Yao Yong. Chinese Landslide prediction research is summarized. geology comment .2008,54 (3): 355-360 pages); Open the Fibonacci method mourning unit's proposition (to open and mourn unit, Huang Runqiu. " Fibonacci number " method of system linearity and nonlinear state and time to rupture forecast before Instability of Rock Body. see: whole nation third time engineering geology conference collection of thesis: publishing house of Chengdu Univ. of Science & Technology, 1988.1233��1240); Sun Jingheng propose Pearl growth model (Sun Jingheng, Li Zhenming, Soviet Union ten thousand benefit the application [J] in slope instability Time Forecast of the .Pearl model. Chinese Geological Disasters with preventing and treating journal, 1993.4 (2): 36-41). The mathematical theory of those methods mainly advanced person and the application of method, have ignored some and the closely-related basic problem that comes down, selection such as the analyzing and processing of: monitoring materials, monitoring sequential, and forecast parameter and slope body deformability destroy, the contacting of evolution mechanism, therefore forecast reliability receives large effect.

Nonlinear prediction, based on the fresh approach such as nonlinear science, control theory, such forecasting model is without the mathematic(al) representation determined, or without explicit math equation, model dimension and model parameter are dynamically changes. Such as Qin Siqing is based on nonlinear dynamics theory, it is proposed that the non-linear dynamic model of slide prediction. This model, based on nonlinear dynamics theory, embodies slope body nonlinear dynamic characteristic, extracts the differentiation information of Landslide System, embody slope body deformability failure mechanism, its explicit physical meaning from monitoring time series.But what existence function was big embody form the unknown, state variable and initial condition be more difficult the deficiency such as determines; The calculating of overall approximation this model of model is more complicated, especially when Embedded dimensions is significantly high orCalculate more complicated time very complicated. (Yin Weiguo. the research of the evaluation of rubble rock mass strength and slide prediction technology. North China University of Tech's Master's thesis .2008,5:38-43 page)

Physical parameter measurement is forecast, such as sound reflecting parameter, owing to being still within conceptual phase, therefore also not commonly used.

Macroscopic view judges forecast, is important method of preventing and reducing natural disasters, but scientific knowledge limitation.

Summary of the invention:

The purpose of the present invention is just for above-mentioned the deficiencies in the prior art, it is provided that a kind of landslide time medium-short term prediction method.

It is an object of the invention to be achieved through the following technical solutions:

Landslide time medium-short term prediction method, comprises the following steps:

A, landslide parameter processing, by Slip moinitoring data inputting to data.txt file, first row: natural law; Secondary series: x displacement (horizontal displacement, cm); 3rd row: y displacement (vertical displacement, cm);

B, data fitting: by data.txt file typing landslide time medium-short term prediction multinomial model;

C, data select, including X displacement, Y displacement or resultant displacement;

Data item in d, data file uses, including before using total data item or only using _ _ individual data item;

E, preservation result of calculation;

F, Model Selection: ��=P₀+ P₁T+P₂t²

P in formula₀For initial displacement, P₁For initial velocity, 2P₂For acceleration, �� is displacement, and t monitors the time.

G, whether present acceleration mode, be;

H, multinomial model: $\mathrm{\Δ}=\frac{P_4+P_{5}t+P_6{t}^2}{P_3-t}-\frac{P_4}{P_3}$

In formula: P3, p4, p5, p6, be undetermined coefficient, change along with the change of data.txt file;

I, start calculate;

J, obtain forecast result.

Beneficial effect: the present invention proposes a kind of landslide time medium-short term prediction multinomial model, and this model structure is simple, workable, simulates effective, it is possible to achieve real-time tracing prediction of landslides. This medium-short term prediction is a kind of dynamic tracking forecast, in forecasting process, often obtains new observation data, just carries out the process of landslide parameter, substitutes into landslide medium-short term prediction system and forecasts, each monitoring point all can obtain determining solution. Namely forecast with the up-to-date variation characteristic that comes down, accomplish that real-time tracking forecasts, accurately and reliably, medium-short term prediction is had very strong adaptability, the efficiency solving slide prediction is low, reliability is relatively low and calculates the problems such as complicated, can better for the service of preventing and reducing natural disasters.

Accompanying drawing illustrates:

Accompanying drawing 1 landslide time medium-short term prediction method software flow pattern

The main interface of accompanying drawing 2 software

Accompanying drawing 3 sampled data

The file of " output+ current time " by name is generated under accompanying drawing 4data.txt catalogue

Accompanying drawing 5 comes down Model for Movement Law horizontal displacement fitting result (output.txt and outputdata.txt)

Accompanying drawing 6 comes down Model for Movement Law horizontal displacement-time curve

Accompanying drawing 7 comes down Model for Movement Law vertical displacement fitting result (output.txt and outputdata.txt)

Accompanying drawing 8 comes down Model for Movement Law vertical displacement-time curve

Accompanying drawing 9 comes down Model for Movement Law resultant displacement fitting result (output.txt and outputdata.txt)

Accompanying drawing 10 comes down Model for Movement Law resultant displacement-time curve

Accompanying drawing 11 forecasting model horizontal displacement fitting result (output.txt and outputdata.txt)

Accompanying drawing 12 forecasting model level versus time relation curve

A forecasting model horizontal displacement-time curve, b forecasting model horizontal velocity-time curve, c forecasting model horizontal acceleration-time curve

Accompanying drawing 13 forecasting model vertical displacement fitting result (output.txt and outputdata.txt)

Accompanying drawing 14 forecasting model is vertical-time curve

A forecasting model vertical displacement-time curve, b forecasting model vertical speed-time curve, c forecasting model normal acceleration-time curve

Accompanying drawing 15 forecasting model resultant displacement fitting result (output.txt and outputdata.txt)

Accompanying drawing 16 forecasting model conjunction-time curve

A forecasting model resultant displacement-time curve, b forecasting model sum velocity-time curve, c forecasting model resultant acceleration-time curve

Figure 17 forecasting model fitting result table

Detailed description of the invention:

It is described in further detail below in conjunction with drawings and Examples:

The present invention relates to and a kind of utilize computer program that landslide is carried out medium-short term prediction, estimate the landslide play sliding time.Model of the present invention adopts displacement dynamic parameter, because reflection landslide kinestate is apparent that displacement dynamic parameter most, and displacement dynamic parameter is prone to obtain, and also relatively directly perceived, it reflects the comprehensive function result of various factors. Due to scatterplot certain curve extremely similar of most of landslide displacements-time dynamic monitoring data row, the present invention change curve according to displacement with monitoring time, it is proposed to object of which movement equationSimilar equation $\mathrm{\Δ}=\frac{P_4+P_{5}t+P_6{t}^2}{P_3-t}-\frac{P_4}{P_3}$ As multinomial forecasting model.

In formula, �� is displacement, and t monitors time, P in formula₀For initial displacement, P₁For initial velocity, 2P₂For acceleration; P3, p4, p5, p6, it is undetermined coefficient, changes along with the change of data.txt file.

" motor process " that namely come down that the displacement on landslide and the relation curve between the time characterize, this medium-short term prediction multinomial model is just based on known " motor process " and goes infer following " motor process " and seek the acute sliding time of coming down. This forecasting model function has convergence at y direction, there is incremental time �� t �� 0, i.e. t �� P₀The point of displacement increment �� �� �� ��, namely this function have vertical asymptote (definition according to vertical asymptote, when on curve 1 M along curve infinitely away from initial point time, if M is substantially equal to zero to the distance of straight line, then this straight line is called this asymptote of curve). In displacement-time curve; if deformation trends towards infinity at short notice; extensive slip can be there is in that before illustrating to put between landslide is at this moment; then the time corresponding to asymptote is the landslide play sliding time; namely 90 �� of (the i.e. angle �� of the tangent line of certain point and time abscissa, i.e. ��=tan on displacement-time curve are reached with the displacement-time curve angle of contingence^-1(dx/dt) when tending to 90 ��) as the criterion of Time Prediction of Landslide. The present invention adopts C++ and C# program language (to be called for short LM algorithm by Levenberg-Marquardt algorithm, nonlinear iteration optimization) work out medium-short term prediction program software, landslide displacement dynamic monitoring data is carried out nonlinear regression, determine the undetermined coefficient of model, and the development along with landslide, computer processes in real time, finally determines the landslide play sliding time.

Forecast is relevant with the density gathering data for period. When adopting mid-range forecast, the deformation values of input should in the moon; When adopting short-period forecast, input deformation values is in day. If Slope occurs abnormal, monitoring periods should be encrypted, be monthly encrypted as every 10 days once as being had by monitoring periods; Slope enters sharply after deformation stage, is encrypted as once a day, is beneficial to improve the accuracy of forecast.

Forecasting model displacement versus time matched curve ensures that initial point value is zero, and namely initial time displacement is zero, is consistent with practical situation. After landslide enters acceleration deformation stage, along with being continuously increased of Landslide Deformation speed, its acceleration be on the occasion of, and in the trend being gradually increased, landslide displacement-time curve, Velocity-time relation curve and the acceleration-time curve obtained by forecasting model of the present invention meets this characteristics of motion of landslide, all in non-linear relation, it is consistent with practical situation.

The work process of the present invention:

1, early stage landslide parameter processing: create a deformation measurement data file, set up plain text format .txt document, date corresponding for Slip moinitoring data is converted into natural law, the natural law of monitoring from date is decided to be 1, thereafter monitor the date natural law (namely the difference of monitoring date and from date adds 1 every time) from from date every time, be input in plain text format .txt document first row;Monitoring corresponding x displacement (horizontal displacement) every time and be input in secondary series, the value that from date is corresponding is 0, is followed by accumulative displacement, and unit is cm; Monitoring corresponding y displacement (vertical displacement) every time and be input in secondary series, the value that from date is corresponding is 0, is followed by accumulative displacement, and unit is cm. Time between the column and the column with comma interval (note: input comma, ", keyboard shift is to input under English state, otherwise cannot calculate).

Document format data:

Time, x displacement, y displacement

Time, x displacement, y displacement

Time, x displacement, y displacement

Time, x displacement, y displacement

����

����

2, data fitting: after data file is ready to complete, opens invention software, enters fit procedure interface:

1) click " browsing " and select data file, select .txt document after treatment.

2) data select (namely select x displacement or y displacement or resultant displacement, can only single choice)

3) data item in data file selects, and selection is n item data before matching total data or matching

4) result of calculation is preserved: when selecting "Yes", calculation result data file is saved under the catalogue at deformation measurement data file place, and a newly-built sub-folder named with " output+ current time ", deposit fitting result wherein, including three the large scale png pictures (displacement versus time relation curve, Velocity-time relation curve and acceleration-time curve) being decorated with curvilinear figure, fitting data file outputdata.txt, matching journal file output.txt. Select "No" not preserve result of calculation, namely do not generate new file.

5) Model Selection: select ��=P₀+ P₁T+P₂t²Model (physical equations of motion, come down Model for Movement Law), clicks " starting to calculate ", then eject four windows, respectively fitting result window, displacement versus time relation curve window, Velocity-time relation curve window and acceleration-time curve window. Judge whether landslide enters acceleration deformation stage (Landslide Deformation sustainable growth according to displacement versus time relation curve, displacement-time curve slope dramatically increases, �� > 45 ��, certain acceleration occurs), if the displacement-time curve of Monitoring Data does not present acceleration mode, then this monitoring point is not within the scope of medium-short term prediction, exits software; If the displacement-time curve of Monitoring Data presents acceleration mode, it was shown that monitoring point enters medium-short term prediction scope, then select $\mathrm{\Δ}=\frac{P_4+P_{5}t+P_6{t}^2}{P_3-t}-\frac{P_4}{P_3}$ Model carries out medium-short term prediction. Select $\mathrm{\Δ}=\frac{P_4+P_{5}t+P_6{t}^2}{P_3-t}-\frac{P_4}{P_3}$ After model, click " starting to calculate ", eject four windows, respectively fitting result window, displacement versus time relation curve window, Velocity-time relation curve window and acceleration-time curve window, wherein fitting result window shows: sampled data number, fitting parameter p [3], p [4], p [5], p [6] value, remaining poor (quadratic sum of the difference of measured value and predictive value), average remaining poor (difference of measured value and predictive value square average), correlation coefficient, asymptote (namely the time on landslide occurs in prediction), velocity expression (first derivative of displacement-time curve), acceleration expression formula (second dervative of displacement-time curve).

In a, displacement versus time relation curve window "+" type curve is observational deformation curve; Smoothed curve is matching deformation curve, and real segment is matched curve before the up-to-date monitoring date, and phantom line segments is the prediction matched curve after the up-to-date monitoring date.

B, coefficient R expression formula be:

$R=\frac{\underset{i}{\mathrm{\Σ}}(x_i-\stackrel{\‾}{x})(y_i-\stackrel{\‾}{y})}{\sqrt{\underset{i}{\mathrm{\Σ}}{(x_i-\stackrel{\‾}{x})}^2}\sqrt{\underset{i}{\mathrm{\Σ}}{(y_i-\stackrel{\‾}{y})}^2}}$

Wherein x_iFor digital simulation value,For digital simulation meansigma methods, y_iFor measured value,For actual measurement meansigma methods, remaining difference is the quadratic sum of measured value and the difference of predictive value, and average remaining difference is the average of remaining difference.

The asymptote value of gained is natural law, is converted into date form according to from date, and namely the date of slip occurs in prediction.

Come down for somewhere, Sichuan: as it is shown on figure 3, start monitoring from June 1st, 2010, carry out 66 deformation monitorings altogether. The data that this medium-short term prediction adopts are (t_i,��_i), i=1,2...66. Multinomial is substituted into after Monitoring Data being processedForecasting model carries out medium-short term prediction.

1, landslide parameter processing: by Slip moinitoring data inputting to data.txt file. First row: natural law; Secondary series: x displacement (horizontal displacement, cm); 3rd row: y displacement (vertical displacement, cm). Sampled data number: 66. The individual meaning of this sampled data number 66 Monitoring Data for having 66 days in data.txt, on June 1st, 2010, the data form of correspondence was " 1,0,0 ";

2, data fitting: by data.txt file typing landslide time medium-short term prediction multinomial model. Data select: x displacement; Data item in data file uses: use total data item; Preserve result of calculation: be, Model Selection: ��=P₀+ P₁T+P₂t². Click " starting to calculate ". Generating the file of " output+ current time " by name under data.txt catalogue, the inside includes three the large scale png pictures being decorated with curvilinear figure, fitting data file outputdata.txt, matching journal file output.txt; As shown in Figure 4,5, 6.

3, fitting result and displacement versus time relation curve are checked;

4, data select: y displacement; Data item in data file uses: use total data item; Preserve result of calculation: be, Model Selection: ��=P₀+ P₁T+P₂t². Click " starting to calculate "; Check fitting result and displacement versus time relation curve;

5, data select: resultant displacement; Data item in data file uses: use total data item; Preserve result of calculation: be, Model Selection: ��=P₀+ P₁T+P₂t². Click " starting to calculate ". Check fitting result and displacement versus time relation curve. Presented acceleration mode by the known landslide of Fig. 5��Figure 10, be, continue operation.

6, data select: x displacement; Data item in data file uses: use total data item; Preserve result of calculation: be; Multinomial model:, click " starting to calculate ". The file of " output+ current time " by name is generated under data.txt catalogue, the inside includes three the large scale png pictures (displacement versus time relation curve, Velocity-time relation curve and acceleration-time curve) being decorated with curvilinear figure, fitting data file outputdata.txt, matching journal file output.txt. As shown in Figure 11,12.

Horizontal displacement is: ${\mathrm{\Δ}}_h=\frac{-19.0161-20.4666t+{0.5316t}^2}{1458.87-t}+\frac{19.0161}{1458.87}$

Horizontal velocity is: $v_h=\frac{{0.5316t}^2-20.4666t-19.0161}{{(1458.87-t)}^2}+\frac{1.0632t-20.4666}{1458.87-t}$

Horizontal acceleration is:

$a_h=\frac{{1.0632t}^2-40.9332t-30.0322}{{(1458.87-t)}^3}+\frac{2.1264t-40.9332}{{(1458.87-t)}^2}+\frac{1.0632}{1458.87-t}$

Asymptote value is: 1459

7, data select: y displacement; Data item in data file uses: use total data item; Preserve result of calculation: be; Multinomial model:, click " starting to calculate ". The file of " output+ current time " by name is generated under data.txt catalogue, the inside includes three the large scale png pictures (displacement versus time relation curve, Velocity-time relation curve and acceleration-time curve) being decorated with curvilinear figure, fitting data file outputdata.txt, matching journal file output.txt. Such as Figure 13, shown in 14.

Vertical displacement is: ${\mathrm{\Δ}}_v=\frac{-20.2882+97.9027t+{0.2647t}^2}{1672.3-t}+\frac{20.2882}{1672.3}$

Vertical speed is: $v_v=\frac{{0.2647t}^2+97.9027t-20.2882}{{(1672.3-t)}^2}+\frac{0.5294t+97.9027}{1672.3-t}$

Normal acceleration is:

$a_v=\frac{{0.5294t}^2+195.8054t-40.5764}{{(1672.3-t)}^3}+\frac{1.0588t+195.8054}{{(1672.3-t)}^2}+\frac{0.5294}{1672.3-t}$

Asymptote value is: 1672

8, data select: resultant displacement;Data item in data file uses: use total data item; Preserve result of calculation: be; Multinomial model:, click " starting to calculate "; The file of " output+ current time " by name is generated under data.txt catalogue, the inside includes three the large scale png pictures (displacement versus time relation curve, Velocity-time relation curve and acceleration-time curve) being decorated with curvilinear figure, fitting data file outputdata.txt, matching journal file output.txt. As shown in Figure 15,16.

Resultant displacement is:

Sum velocity is:

Resultant acceleration is:

Asymptote value is: 1478

Finally asymptote value (natural law) is converted into the date of corresponding monitoring initial time, the acute sliding time of coming down can be obtained, see attached list 1. This forecasting model fitting result table is in order to illustrate that fitting degree is high, and correlation coefficient levels off to 1, and remaining difference is little, and forecast result is reliable; Asymptote value correspondence calls time the i.e. landslide play sliding time in advance simultaneously.

Visible, correlation coefficient more than 0.997, the match value obtained by the present invention is very close to measured value, and average remaining difference is less than 52, and model is high to displacement fitting degree, then just high according to the accuracy of built model prediction landslide time of origin. Therefore, the forecast result that model obtains is accurately and reliably.

Under square one, forecast adopts the value that asymptote is less, so asymptote value selects 1459, its corresponding date is 2014.05.29. Namely coming down according to the present invention, estimating the landslide play sliding time is on May 29th, 2014.

Claims (1)

1. a landslide time medium-short term prediction method, it is characterised in that the method comprises the following steps:

A, landslide parameter processing, by Slip moinitoring data inputting to data.txt file, first row: natural law; Secondary series: x displacement, x displacement is horizontal displacement, unit cm; 3rd row: y displacement, y displacement is vertical displacement, unit cm;

B, data fitting: by data.txt file typing landslide time medium-short term prediction multinomial model;

C, data select, including X displacement, Y displacement or resultant displacement;

Data item in d, data file uses, and including using total data item or only using top n data item, N is arithmetic number;

E, preservation result of calculation;

F, Model Selection: ��=P₀+P₁t+P₂t²

P in formula₀For initial displacement, P₁For initial velocity, 2P₂For acceleration, �� is displacement, and t monitors the time;

G, whether present acceleration mode, be;

H, multinomial model:

In formula: P₃��P₄��P₅��P₆It is undetermined coefficient, changes along with the change of data.txt file;

I, start calculate;

J, obtain forecast result.