CN107587493B - Foundation pit excavation monitoring and early warning method based on neural network - Google Patents
Foundation pit excavation monitoring and early warning method based on neural network Download PDFInfo
- Publication number
- CN107587493B CN107587493B CN201710780941.5A CN201710780941A CN107587493B CN 107587493 B CN107587493 B CN 107587493B CN 201710780941 A CN201710780941 A CN 201710780941A CN 107587493 B CN107587493 B CN 107587493B
- Authority
- CN
- China
- Prior art keywords
- displacement
- foundation pit
- excavation
- monitoring
- value
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000009412 basement excavation Methods 0.000 title claims abstract description 100
- 238000012544 monitoring process Methods 0.000 title claims abstract description 96
- 238000000034 method Methods 0.000 title claims abstract description 51
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 29
- 238000006073 displacement reaction Methods 0.000 claims abstract description 177
- 238000003062 neural network model Methods 0.000 claims abstract description 20
- 238000004458 analytical method Methods 0.000 claims abstract description 16
- 238000011156 evaluation Methods 0.000 claims abstract description 4
- 239000013598 vector Substances 0.000 claims description 26
- 238000010276 construction Methods 0.000 claims description 23
- 210000002569 neuron Anatomy 0.000 claims description 18
- 230000008569 process Effects 0.000 claims description 10
- 230000002787 reinforcement Effects 0.000 claims description 7
- 238000012937 correction Methods 0.000 claims description 6
- 238000013461 design Methods 0.000 claims description 6
- 238000010586 diagram Methods 0.000 claims description 4
- 238000005728 strengthening Methods 0.000 claims description 4
- 230000000087 stabilizing effect Effects 0.000 claims description 3
- 230000009286 beneficial effect Effects 0.000 abstract description 2
- 238000001514 detection method Methods 0.000 abstract description 2
- 230000006870 function Effects 0.000 description 9
- 239000002689 soil Substances 0.000 description 6
- 238000012549 training Methods 0.000 description 6
- 238000004364 calculation method Methods 0.000 description 3
- 238000011161 development Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 2
- 230000003014 reinforcing effect Effects 0.000 description 2
- 238000009825 accumulation Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000004064 recycling Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
Images
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention relates to a detection early warning method, in particular to a foundation pit excavation monitoring early warning method based on a neural network, and belongs to the field of real-time monitoring and forecasting of deep foundation pit excavation deformation and stability analysis and evaluation of the deep foundation pit excavation deformation. The method mainly comprises the following steps: the method includes the steps of arranging foundation pit displacement monitoring points, determining horizontal and vertical displacements of different excavation depths of a foundation pit slope top, determining the mean values of the horizontal and vertical displacements of the different excavation depths of the foundation pit slope top, establishing a BP neural network model based on foundation pit displacement, compiling a neural network program in deep foundation pit deformation prediction, predicting the level and the mean value of the vertical displacement of the monitoring points excavated step by step in the foundation pit, and analyzing and evaluating stability of the foundation pit. The invention has the beneficial effects that: the method has the advantages of self-adaption, real-time performance, high precision and strong practicability, the prediction result is more consistent with the actual deformation, the real-time modeling prediction of the neural network is an effective method for displacement prediction, and the method has great practicability.
Description
Technical Field
The invention relates to a detection early warning method, in particular to a foundation pit excavation monitoring early warning method based on a neural network, and belongs to the field of real-time monitoring and forecasting of deep foundation pit excavation deformation and stability analysis and evaluation of the deep foundation pit excavation deformation.
Background
With the vigorous development of city basic construction and the progress of deep foundation pit excavation technology, the development and utilization ratio of urban underground space is continuously increased, and particularly, foundation pits of 10m, 20m and 30m are visible everywhere in a first-line city. The excavation depth of the foundation pit is continuously increased, the excavation difficulty is increased, the safety and stability of the foundation pit are considered, and the deformation caused by the foundation pit and the influence of the deformation on the surrounding environment are also considered. Heretofore, the design of strength control has been largely replaced by the design of structural deformation control. The key of the structural deformation control design is the deformation analysis and prediction of the foundation pit support structure, and the stability of the foundation pit slope is judged through the analysis and prediction of the deformation of the foundation pit support structure. Therefore, the method has important engineering application value on accurately predicting the displacement deformation of the foundation pit.
The existing methods for predicting the displacement deformation of the foundation pit comprise a finite element method and a stratum loss method. The finite element method is a powerful numerical calculation method, can not only calculate the internal force of the soil nail, the stress-strain relationship of the soil body, simulate the excavation process and the like, but also can consider the non-uniformity of the soil body and the complex state of anisotropy, but the theoretical research is still not perfect, a constitutive model which truly reflects the stress-strain relationship of the soil body is lacked, and the calculation parameters are difficult to accurately determine, so that the support and the displacement of the soil body cannot be accurately calculated, and the prediction effect and the actual deformation have larger deviation; the stratum loss method is to use the principle of the correlation between the horizontal displacement of the wall and the ground surface settlement and adopt the rod system finite element or the elastic beam method to calculate the vertical displacement of the ground surface, namely the ground settlement, however, the method is more applicable in the soft soil area and basically has no application value in other areas, and the two points restrict the application of the method in the engineering practice.
Disclosure of Invention
Aiming at the defects in the technology, the invention provides a foundation pit excavation monitoring and early warning method based on a neural network.
The technical scheme for solving the technical problems is as follows:
the invention aims to provide a method for analyzing stability of a side slope foundation pit by using a neural network principle. And then determining a BP neural network curve model according to the principle of the neural network in the deep foundation pit deformation prediction, further compiling a deep foundation pit deformation prediction program, and modeling and predicting according to the displacement data of the monitoring points of the foundation pit. And finally, evaluating the stability of the foundation pit slope through the reliability and credibility analysis of the BP neural network curve model.
Step one, arrangement of foundation pit displacement monitoring points
According to the technical specification for monitoring the construction foundation pit engineering GB50497-2009 and the monitoring regulations for the construction of the foundation pit engineering (DG/TJ08-2001 + 2006), M displacement monitoring datum points are arranged in a stable deformation-free area which is 3 times of the excavation depth (3H) outside a monitored foundation pit excavation surface, L displacement monitoring points are arranged at the top of a slope of the foundation pit or the top of a surrounding wall along the periphery of the foundation pit, the displacement monitoring datum points and the displacement monitoring points jointly form a displacement deformation monitoring control network of the excavation surface, and the horizontal displacement and the vertical displacement of the top of the slope of the foundation pit at different excavation depths are monitored along with the excavation of the foundation pit, which is shown in an attached figure 2.
Step two, determining horizontal and vertical displacements of different excavation depths of the slope top of the foundation pit
Along with the gradual increase of the excavation depth h of the foundation pit, the side wall can generate obvious deformation, wherein the displacement deformation of the top of the side wall is the largest, so the analysis is carried out by the displacement of the slope top of the foundation pit. Taking a section of the periphery of the top of the slope as an example, the number of the monitoring points on the section is N, see the attached figure 2.
Horizontal displacement data: xHi(hj)(i＝1,2,…,N；j＝1,2,…,n)
Vertical displacement data: xVi(hj)(i＝1,2,…,N；j＝1,2,…,n)
Wherein N represents the number of monitoring points arranged on the section of the foundation pit, and N represents the total times of the monitoring displacement taken in the monitoring process.
For convenient analysis, different excavation depths hjThe horizontal and vertical displacements for the N monitoring points are listed in table 1.
TABLE 1 horizontal and vertical Displacement of monitoring points
Step three, determining the average value of horizontal and vertical displacements of different excavation depths of the slope top of the foundation pit
The displacement data of a certain monitoring point on the top of the foundation pit cannot reflect the deformation of a certain section of the foundation pit, and for representativeness, the average value of the horizontal displacement and the vertical displacement of N monitoring points is used as the deformation of the section of the foundation pit in the excavation process, and can be represented by formulas (1) and (2):
mean of horizontal displacement
Mean value of vertical displacement
The data of the mean values of the horizontal displacement and the vertical displacement of the foundation pit monitoring points at different moments in the initial excavation stage are shown in a table 2, so that
Can be used as raw data for the following modeling program.
TABLE 2 data sheet of different excavation depths of each monitoring point of foundation pit, horizontal displacement and vertical displacement mean value
Step four, establishing a BP neural network model based on foundation pit displacement
Establishment of BP neural network model
The deformation of the foundation pit in the horizontal and vertical directions is reflected on the monitoring data, the displacement observation sequence of the foundation pit changes along with the excavation depth h, the front and the back are not independent any more and have a certain dependency relationship, the current state is the extension of the past state and is a dynamic change process, and therefore, according to the characteristic of the deformation of the deep foundation pit, the method for modeling and predicting the accumulation of the displacement data of the foundation pit along with different excavation depths by utilizing the neural network is provided. According to the modeling principle of the BP neural network prediction method, the BP network model is determined to be composed of 3 parts including an input layer, a hidden layer and an output layer, the number of units of the input layer is one, and the BP network model is composed of an excavation depth vector group [ h ]1,h2,…,hn]Represents; the number of output layer units is one, and the output layer units are composed of a foundation pit displacement vector group [ YH(h1),YH(h2),…，YH(hn)]Or [ YV(h1),YV(h2),…,YV(hn)]Represents; the weight and threshold from the input layer to the hidden layer are wjkAnd thetajkThe weight and threshold from the hidden layer to the output layer are vkjAnd gammakt. The established BP neural network model is shown in figure 3.
Algorithm for establishing (II) BP neural network model
1) According to the BP neural network model diagram, the net input and output of each neuron of the middle hidden layer are respectively represented by the following formulas (1) and (2):
bk＝f(S(k))＝1/(1+e-S(k))(k＝1,2,…,p) (2)
wherein j is the number of elements in the input vector group, and k is the number of hidden layer units.
2) The net input and actual output of each neuron in the output layer are represented by equations (3) and (4), respectively:
ct＝f(S(t))＝1/(1+e-S(t))(t＝1,2,…,n) (4)
wherein t is the number of elements in the output vector group.
3) According to a given desired output Y (h)j) Calculating the correction error d of each neuron of the output layertRepresented by formula (5):
dt＝(Y(hj)-ct)f′(S(t))(t＝1,2,…,n) (5)
4) error correction e for each neuron in the hidden layerkRepresented by formula (6):
5) modifying the implicit layer to output layer connection weights V and the output layer neuron threshold values gamma, where alpha is the learning rate, 0< alpha <1
△vkt＝α·dt·bk(k＝1,2,…,p；t＝1,2,…,n) (7)
△γkt＝α·dt(t＝1,2,…,n) (8)
6) Modifying the input layer to hidden layer connection weights W and the hidden layer neuron threshold θ, where β is the learning rate, 0< β <1
△wjk＝β·ek·hj(j＝1,2,…,n；k＝1,2,…,p) (9)
△θjk＝β·ek(k＝1,2,…,p) (10)
And randomly selecting the next learning mode pair and providing the next learning mode pair for the network until all the learning mode pairs are trained. And judging whether the network global error meets the precision requirement, if so, ending, otherwise, continuing.
Five-step programming of neural network program in deep foundation pit deformation prediction
According to the BP neural network model determined in the fourth step, a MATLAB neural network tool box is used for solving a predicted value of foundation pit displacement, and an excavation depth vector group [ h ]1,h2,…,hn]As input samples, set of displacement vectors [ Y [ ]H(h1),YH(h2),…，YH(hn)]Or [ YV(h1),YV(h2),…,YV(hn)]The vector is taken as a target vector and is transferred into a tool box to obtain network output, namely the next excavation depth hn+1The predicted displacement value of (2).
The programming process in the neural network tool box comprises the following steps:
defining an input sample as p ═ h1,h2,…,hn]Defining the target vector as t ═ YH(h1),YH(h2),…，YH(hn)]Or t ═ YV(h1),YV(h2),…,YV(hn)]；
Secondly, initializing the network by using a function initff, wherein the procedure of the step is as follows:
[w1,b1,w2,b2]＝initff(p,S1,t)；
thirdly, training the network by using a function trainbp:
df is 10; % display number of steps
me 10000; % maximum number of training steps
eg ═ 0.001; % target error
lr is 0.01; % learning rate
tp＝[df me eg lr]；
[w1,b1,w2,b2,ep,tr]＝trainbp(w1,b1,′tansig′,w2,b2,′purelin′,p,t,tp)；
After the training is completed, an approximation result of the function can be obtained.
Fourthly, the network output can be calculated by utilizing the function simulff
a＝simuff(p,w1,b1,′tansig′,w2,b2,′purelin′)
The final output result can be obtained by the four procedures
Or
I.e. respectively the excavation depth hn+1The predicted value of the horizontal displacement mean value and the predicted value of the vertical displacement mean value.
Average value prediction of horizontal and vertical displacement of monitoring points for step-by-step excavation of six foundation pits
The BP neural network program compiled in the fifth step is according to the excavation h1,h2,…,hnThe displacement measured value in depth obtains the excavation hn+1The predicted value of the depth displacement is calculated according to the excavation h1,h2,…,hn,hn+1The displacement measured value of the depth can obtain the excavation depth hn+2The modeling and the prediction are carried out simultaneously, and the prediction is carried out until the rule of the deformation of the foundation pit can be reflected. Excavation depth h of foundation pit monitoring points at different excavation depthsjThe mean horizontal and vertical displacement of time is predicted as follows:
1) excavation h by using foundation pit1,h2,…,hnEstablishing a neural network forecasting model for the monitoring displacement value in depth; the forecast model is used to obtain the opening of foundation pitDig hn+1Horizontal displacement mean prediction data of depth
And vertical displacement mean prediction data
The displacement can provide main reference data for information-based construction design;
2) continuing construction, monitoring and calculating to obtain foundation pit excavation hn+1Actual data Y of horizontal displacement mean value of depthH(hn+1) And vertical displacement mean value actual data YV(hn+1) And the horizontal displacement mean value prediction data obtained in 1)
And vertical displacement mean prediction data
Comparing to obtain relative error
And
when the relative error is within +/-5%, the prediction model is reliable, and the next prediction analysis is continued; and when the relative error is beyond +/-5%, analyzing the reason, adjusting the weight and the threshold value, and reestablishing the prediction model until the relative error meets the requirement.
3) Reuse of excavation h1,h2,…,hn+1Reestablishing a neural network model by monitoring displacement value in depth, and predicting excavation hn+2Horizontal displacement mean of depth
And vertical mean value of displacement
4) Continuing construction, monitoring and calculating to obtain foundation pit excavation hn+2Actual data Y of horizontal displacement mean value of depthH(hn+2) And vertical displacement mean value actual data YV(hn+2) And the horizontal displacement mean value prediction data obtained in 3)
And vertical displacement mean prediction data
Comparing to obtain relative error
And
when the relative error is out of +/-5%, analyzing the reason, adjusting the weight and the threshold value, and reestablishing the prediction model until the relative error meets the requirement; when the relative error is within +/-5%, the prediction model is reliable, and the next prediction analysis is continued.
In this way, the modeling process and the forecasting process are continually repeated as the construction progresses, with new data continually being supplemented. Therefore, the deformation of the foundation pit can be continuously tracked, the development trend of the foundation pit can be forecasted, and the construction service can be designed for information.
Seven steps of analyzing and evaluating the stability of the foundation pit
The foundation pit can generate large deformation in the excavation process, and surrounding buildings, underground pipelines, traffic roads and other municipal facilities are seriously affected, so that the control of the deformation of the foundation pit is particularly important. The deformation of the foundation pit is mainly reflected on horizontal displacement and vertical displacement, and monitoring and early warning grades (blue early warning, orange early warning and red early warning) are divided for the deformation of the foundation pit according to the technical specification for monitoring foundation pit engineering GB50497-2009, and are shown in Table 4. Adjusting the displacement predicted value of the monitoring point to meet the error requirement, comparing the displacement predicted value with early warning values of different levels, and if the displacement predicted value of the monitoring point is smaller than a blue early warning level warning value (a control value relative to the depth H of the foundation pit), stabilizing the foundation pit and excavating according to the original scheme; if the displacement predicted value of the monitoring point exceeds a blue alarm value (a control value relative to the depth H of the foundation pit) in the following table, indicating that the foundation pit enters an unstable primary stage, issuing a blue early warning at the moment to draw attention of constructors and strengthening a reinforcing measure in advance; if the displacement predicted value of the monitoring point exceeds an orange alarm value (a control value relative to the depth H of the foundation pit) in the following table, indicating that the foundation pit enters an unstable middle stage, issuing an orange early warning at the moment, adjusting an excavation scheme, and further optimizing a reinforcing measure by constructors; if the displacement predicted value of the monitoring point exceeds a red alarm value (a control value relative to the depth H of the foundation pit) in the following table, the foundation pit is about to be unstable, excessive displacement is generated after next excavation, so that the side wall of the foundation pit is unstable, the excavation is stopped at the moment, the reinforcement measures are further strengthened, and the preparation for evacuating personnel in time is well made.
TABLE 4 Foundation pit and supporting construction monitoring and early warning grade
The invention has the beneficial effects that: the method accurately predicts the next displacement of the foundation pit by establishing a foundation pit displacement-depth curve model, can implement strict tracking monitoring on the construction of the actual engineering, feeds back information according to the monitoring data, adjusts construction parameters along with the excavation depth, enables the construction and the prediction to be carried out simultaneously, and has the advantages of self-adaption, real-time performance, high precision, strong practicability and the like, and the prediction result is matched with the actual deformation. According to the learning modeling function of the neural network, along with the construction progress, the real-time modeling prediction of the neural network is an effective displacement prediction method, and the method is applied to predict the displacement of the foundation pit supporting structure and has great practicability.
Drawings
FIG. 1 is a flow chart of the present invention;
FIG. 2 is a schematic view of the arrangement of monitoring points of a foundation pit;
FIG. 3 is a model diagram of a BP neural network.
Detailed Description
The principles and features of this invention are described below in conjunction with examples which are set forth to illustrate, but are not to be construed to limit the scope of the invention.
A foundation pit excavation monitoring and early warning method based on a neural network comprises the following steps:
arrangement of foundation pit displacement monitoring points
According to the technical specification for monitoring the construction foundation pit engineering GB50497-2009 and the monitoring regulations for the construction of the foundation pit engineering (DG/TJ08-2001 + 2006), M displacement monitoring datum points are arranged in a stable deformation-free area which is 3 times of the excavation depth (3H) outside a monitored foundation pit excavation surface, L displacement monitoring points are arranged at the top of a slope of the foundation pit or the top of a retaining wall along the periphery of the foundation pit, the displacement monitoring datum points and the displacement monitoring points jointly form a displacement deformation monitoring control network of the excavation surface, and the horizontal displacement and the vertical displacement of the top of a slope of the foundation pit at different excavation depths are monitored along with the excavation of the foundation pit;
determination of horizontal and vertical displacement of different excavation depths of foundation pit slope top
Taking a section of the periphery of the top of the slope as an example, the number of the monitoring points on the section is N,
horizontal displacement data: xHi(hj)(i＝1,2,…,N；j＝1,2,…,n)，
Vertical displacement data: xVi(hj)(i＝1,2,…,N；j＝1,2,…,n)，
Wherein, N represents the number of monitoring points arranged on the section of the foundation pit, and N represents the total times of the monitoring displacement taken in the monitoring process;
determination of horizontal and vertical displacement mean values of different excavation depths of foundation pit slope top
The average values of the horizontal displacement and the vertical displacement of the N monitoring points are used as the deformation of the section of foundation pit in the excavation process, and can be represented by the formulas (1) and (2)
Mean of horizontal displacement
Mean value of vertical displacement
Order to
As raw data for a later modeling program;
establishment of BP (Back propagation) neural network model based on foundation pit displacement
Establishing BP neural network model
According to the modeling principle of the BP neural network prediction method, determining that a BP network model consists of three parts including an input layer, a hidden layer and an output layer, wherein the number of units of the input layer is one, and the input layer consists of an excavation depth vector group [ h ]1,h2,…,hn]Represents; the number of output layer units is one, and the output layer units are composed of a foundation pit displacement vector group [ YH(h1),YH(h2),…，YH(hn)]Or [ YV(h1),YV(h2),…,YV(hn)]Represents; the weight and threshold from the input layer to the hidden layer are wjkAnd thetajkThe weight and threshold from the hidden layer to the output layer are vkjAnd gammakt；
Algorithm for establishing BP neural network model
1) According to the BP neural network model diagram, the net input and output of each neuron of the middle hidden layer are respectively represented by the following formulas (1) and (2):
bk＝f(S(k))＝1/(1+e-S(k))(k＝1,2,…,p) (2)
wherein j is the number of elements in the input vector group, and k is the number of hidden layer units;
2) the net input and actual output of each neuron in the output layer are represented by equations (3) and (4), respectively:
ct＝f(S(t))＝1/(1+e-S(t))(t＝1,2,…,n) (4)
wherein t is the number of elements in the output vector group;
3) according to a given desired output Y (h)j) Calculating the correction error d of each neuron of the output layertRepresented by formula (5):
dt＝(Y(hj)-ct)f′(S(t))(t＝1,2,…,n) (5)
4) error correction e for each neuron in the hidden layerkRepresented by formula (6):
5) modifying the implicit layer to output layer connection weights V and the output layer neuron threshold values gamma, where alpha is the learning rate, 0< alpha <1
△vkt＝α·dt·bk(k＝1,2,…,p；t＝1,2,…,n) (7)
△γkt＝α·dt(t＝1,2,…,n) (8)
6) Modifying the input layer to hidden layer connection weights W and the hidden layer neuron threshold θ, where β is the learning rate, 0< β <1
△wjk＝β·ek·hj(j＝1,2,…,n；k＝1,2,…,p) (9)
△θjk＝β·ek(k＝1,2,…,p) (10)
Randomly selecting the next learning mode pair to provide for the network, judging whether the global error of the network meets the precision requirement or not until all the learning mode pairs are trained, if so, ending, otherwise, continuing;
compiling of neural network program in deformation prediction of deep foundation pit
According to the BP neural network model determined in the step four, a foundation pit displacement predicted value can be obtained by using an MATLAB neural network tool box, and an excavation depth vector group is formed[h1,h2,…,hn]As input samples, set of displacement vectors [ Y [ ]H(h1),YH(h2),…，YH(hn)]Or [ YV(h1),YV(h2),…,YV(hn)]The vector is taken as a target vector and is transferred into a tool box to obtain network output, namely the next excavation depth hn+1The displacement prediction value of (2);
the programming steps of the MATLAB neural network toolbox are as follows:
defining an input sample as p ═ h1,h2,…,hn]Defining the target vector as t ═ YH(h1),YH(h2),…，YH(hn)]Or t ═ YV(h1),YV(h2),…,YV(hn)]；
Secondly, initializing the network by using a function initff, wherein the procedure is as follows
[w1,b1,w2,b2]＝initff(p,S1,t)；
Thirdly, training the network by using a function trainbp:
df is 10; % display number of steps
me 10000; % maximum number of training steps
eg ═ 0.001; % target error
lr is 0.01; % learning rate
tp＝[df me eg lr]；
[w1,b1,w2,b2,ep,tr]＝trainbp(w1,b1,′tansig′,w2,b2,′purelin′,p,t,tp)；
After training is finished, an approximation result of the function can be obtained;
fourthly, the network output can be calculated by utilizing the function simulff
a＝simuff(p,w1,b1,′tansig′,w2,b2,′purelin′)；
The final output result can be obtained by the four procedures
Or
I.e. respectively the excavation depth hn+1The predicted value of the horizontal displacement mean value and the predicted value of the vertical displacement mean value.
Sixthly, predicting horizontal and vertical displacement mean values of monitoring points excavated by foundation pits step by step
Fifthly, according to excavation h, the programmed BP neural network program1,h2,…,hnThe displacement measured value in depth obtains the excavation hn+1The predicted value of the depth displacement is calculated according to the excavation h1,h2,…,hn,hn+1The displacement measured value of the depth can obtain the excavation depth hn+2The modeling and the prediction are carried out simultaneously, and the prediction is carried out until the rule of the deformation of the foundation pit can be reflected;
the average value of horizontal and vertical displacement of step-by-step excavation of the foundation pit monitoring points is predicted as follows:
digging a pit1,h2,…,hnEstablishing a neural network forecasting model for the monitoring displacement value in depth; the foundation pit excavation h is obtained by utilizing the forecasting modeln+1Horizontal displacement mean prediction data of depth
And vertical displacement mean prediction data
The displacement can provide main reference data for information-based construction design;
secondly, continuing construction, monitoring and calculating to obtain foundation pit excavation hn+1Actual data Y of horizontal displacement mean value of depthH(hn+1) And vertical displacement mean value actual data YV(hn+1) And the horizontal displacement mean value prediction data obtained in 1)
And vertical displacement mean prediction data
Comparing to obtain relative error
And
when the relative error is within +/-5%, the prediction model is reliable, and the next prediction analysis is continued; when the relative error is out of +/-5%, analyzing the reason, adjusting the weight and the threshold value, and reestablishing the prediction model until the relative error meets the requirement;
thirdly, recycling and excavating h1,h2,…,hn+1Reestablishing a neural network model by monitoring displacement value in depth, and predicting excavation hn+2Horizontal displacement mean of depth
And vertical mean value of displacement
Fourthly, continuing construction, monitoring and calculating to obtain the excavation h of the foundation pitn+2Actual data Y of horizontal displacement mean value of depthH(hn+2) And vertical displacement mean value actual data YV(hn+2) And the predicted number of horizontal shift means obtained in 3)
The weight value and the threshold value are adjusted, and the prediction model is reestablished until the relative error meets the requirement; when the relative error is within +/-5%, the prediction model is reliable, and the next prediction analysis is continued.
Analysis and evaluation of pit stability
Sixthly, adjusting the displacement predicted value of the monitoring point to meet the error requirement, comparing the adjusted displacement predicted value with early warning values of different levels, and if the displacement predicted value of the monitoring point is smaller than a blue early warning level warning value, stabilizing the foundation pit and excavating according to the original scheme; if the displacement predicted value of the monitoring point exceeds a blue alarm value in the following table, indicating that the foundation pit enters an unstable primary stage, issuing a blue early warning at the moment to attract the attention of constructors and strengthening reinforcement measures in advance; if the displacement predicted value of the monitoring point exceeds an orange alarm value in the following table, indicating that the foundation pit enters an unstable middle stage, issuing an orange early warning, adjusting an excavation scheme and further optimizing reinforcement measures by constructors; if the displacement predicted value of the monitoring point exceeds a red alarm value in the following table, the foundation pit is about to be unstable, excessive displacement is generated after next excavation, so that the side wall of the foundation pit is unstable, the excavation is stopped at the moment, the reinforcement measures are further strengthened, and the preparation for evacuating personnel in time is well made.
The invention uses the foundation pit excavation monitoring and early warning method based on the neural network to replace the traditional finite element method and the stratum loss method for predicting the displacement deformation of the foundation pit, can solve the problems that the theory of the finite element method is imperfect and the calculation value is difficult to determine, and simultaneously solves the problem that the use range of the stratum loss method is limited, and can evaluate the stability of the foundation pit slope through the reliable and credibility analysis of the BP neural network curve model in the implementation process, thereby laying a foundation for the excavation of the foundation pit.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents, improvements and the like that fall within the spirit and principle of the present invention are intended to be included therein.
Claims (1)
1. A foundation pit excavation monitoring and early warning method based on a neural network is characterized by comprising the following steps:
arrangement of foundation pit displacement monitoring points
According to the technical specification for monitoring the construction foundation pit engineering GB50497-2009 and the monitoring regulations for the construction of the foundation pit engineering (DG/TJ08-2001 + 2006), M displacement monitoring datum points are arranged in a stable deformation-free area which is 3 times of the excavation depth (3H) outside a monitored foundation pit excavation surface, L displacement monitoring points are arranged at the top of a slope of the foundation pit or the top of a retaining wall along the periphery of the foundation pit, the displacement monitoring datum points and the displacement monitoring points jointly form a displacement deformation monitoring control network of the excavation surface, and the horizontal displacement and the vertical displacement of the top of a slope of the foundation pit at different excavation depths are monitored along with the excavation of the foundation pit;
determination of horizontal and vertical displacement of different excavation depths of foundation pit slope top
Taking a section of the periphery of the top of the slope as an example, the number of the monitoring points on the section is N,
horizontal displacement data: xHi(hj)(i＝1,2,…,N；j＝1,2,…,n)，
Vertical displacement data: xVi(hj)(i＝1,2,…,N；j＝1,2,…,n)，
Wherein, N represents the number of monitoring points arranged on the section of the foundation pit, and N represents the total times of the monitoring displacement taken in the monitoring process;
determination of horizontal and vertical displacement mean values of different excavation depths of foundation pit slope top
The average values of the horizontal displacement and the vertical displacement of the N monitoring points are used as the deformation of the section of foundation pit in the excavation process, and can be represented by the formulas (1) and (2)
Mean of horizontal displacement
Mean value of vertical displacement
Order to
As raw data for a later modeling program;
establishment of BP (Back propagation) neural network model based on foundation pit displacement
Establishing BP neural network model
According to the modeling principle of the BP neural network prediction method, determining that a BP network model consists of three parts including an input layer, a hidden layer and an output layer, wherein the number of units of the input layer is one, and the input layer consists of an excavation depth vector group [ h ]1,h2,…,hn]Represents; the number of output layer units is one, and the output layer units are composed of a foundation pit displacement vector group [ YH(h1),YH(h2),…，YH(hn)]Or [ YV(h1),YV(h2),…,YV(hn)]Represents; the weight and threshold from the input layer to the hidden layer are wjkAnd thetajkThe weight and threshold from the hidden layer to the output layer are vkjAnd gammakt；
Algorithm for establishing BP neural network model
1) According to the BP neural network model diagram, the net input and output of each neuron of the middle hidden layer are respectively represented by the following formulas (1) and (2):
bk＝f(S(k))＝1/(1+e-S(k))(k＝1,2,…,p) (2)
wherein j is the number of elements in the input vector group, and k is the number of hidden layer units;
2) the net input and actual output of each neuron in the output layer are represented by equations (3) and (4), respectively:
ct＝f(S(t))＝1/(1+e-S(t))(t＝1,2,…,n) (4)
wherein t is the number of elements in the output vector group;
3) according to a given desired output Y (h)j) Calculating the correction error d of each neuron of the output layertRepresented by formula (5):
dt＝(Y(hj)-ct)f′(S(t))(t＝1,2,…,n) (5)
4) error correction e for each neuron in the hidden layerkRepresented by formula (6):
5) modifying the connection weight V from the hidden layer to the output layer and the threshold value gamma of the neuron in the output layer, wherein alpha is the learning rate, and alpha is more than 0 and less than 1
Δvkt＝α·dt·bk(k＝1,2,…,p；t＝1,2,…,n) (7)
Δγkt＝α·dt(t＝1,2,…,n) (8)
6) Modifying the connection weight W from the input layer to the hidden layer and the threshold value theta of the neuron in the hidden layer, wherein beta is the learning rate, and beta is more than 0 and less than 1
Δwjk＝β·ek·hj(j＝1,2,…,n；k＝1,2,…,p) (9)
Δθjk＝β·ek(k＝1,2,…,p) (10)
Randomly selecting the next learning mode pair to provide for the network, judging whether the global error of the network meets the precision requirement or not until all the learning mode pairs are trained, if so, ending, otherwise, continuing;
compiling of neural network program in deformation prediction of deep foundation pit
According to the BP neural network model determined in the step four, a foundation pit displacement predicted value can be obtained by using an MATLAB neural network tool box, and an excavation depth vector group [ h ]1,h2,…,hn]As input samples, set of displacement vectors [ Y [ ]H(h1),YH(h2),…，YH(hn)]Or [ YV(h1),YV(h2),…,YV(hn)]The vector is taken as a target vector and is transferred into a tool box to obtain network output, namely the next excavation depth hn+1The displacement prediction value of (2);
sixthly, predicting horizontal and vertical displacement mean values of monitoring points excavated by foundation pits step by step
Fifthly, according to excavation h, the programmed BP neural network program1,h2,…,hnThe displacement measured value in depth obtains the excavation hn+1The predicted value of the depth displacement is calculated according to the excavation h1,h2,…,hn,hn+1The displacement measured value of the depth can obtain the excavation depth hn+2The modeling and the prediction are carried out simultaneously, and the prediction is carried out until the rule of the deformation of the foundation pit can be reflected; specifically, the method comprises the following steps:
(S1) excavating a foundation pit h1,h2,…,hnEstablishing a neural network forecasting model for the monitoring displacement value in depth; the foundation pit excavation h is obtained by utilizing the forecasting modeln+1Horizontal displacement mean prediction data of depth
And vertical displacement mean prediction data
The displacement can provide main reference data for information-based construction design;
(S2) continuing construction, monitoring and calculating to obtain foundation pit excavation hn+1Actual data Y of horizontal displacement mean value of depthH(hn+1) And vertical displacement mean value actual data YV(hn+1) And the horizontal displacement average value prediction data obtained in (S1)
And vertical displacement mean prediction data
Comparing to obtain relative error
And
when the relative error is within +/-5%, the prediction model is reliable, and the next prediction analysis is continued; when the relative error is out of +/-5%, analyzing the reason, adjusting the weight and the threshold value, and reestablishing the prediction model until the relative error meets the requirement;
(S3) RebateBy digging h1,h2,…,hn+1Reestablishing a neural network model by monitoring displacement value in depth, and predicting excavation hn+2Horizontal displacement mean of depth
And vertical mean value of displacement
(S4) continuing construction, monitoring and calculating to obtain foundation pit excavation hn+2Actual data Y of horizontal displacement mean value of depthH(hn+2) And vertical displacement mean value actual data YV(hn+2) And the horizontal displacement average value prediction data obtained in (S3)
And vertical displacement mean prediction data
Comparing to obtain relative error
And
when the relative error is out of +/-5%, analyzing the reason, adjusting the weight and the threshold value, and reestablishing the prediction model until the relative error meets the requirement; when the relative error is within +/-5%, the prediction model is reliable, and the next prediction analysis is continued;
analysis and evaluation of pit stability
Sixthly, adjusting the displacement predicted value of the monitoring point to meet the error requirement, comparing the adjusted displacement predicted value with early warning values of different levels, and if the displacement predicted value of the monitoring point is smaller than a blue early warning level warning value, stabilizing the foundation pit and excavating according to the original scheme; if the displacement predicted value of the monitoring point exceeds the blue alarm value, indicating that the foundation pit enters an unstable primary stage, and issuing a blue early warning to attract attention of constructors and strengthening reinforcement measures in advance; if the displacement predicted value of the monitoring point exceeds the orange alarm value, indicating that the foundation pit enters an unstable middle stage, issuing an orange early warning at the moment, adjusting an excavation scheme, and further optimizing reinforcement measures by constructors; and if the displacement predicted value of the monitoring point exceeds the red alarm value, indicating that the foundation pit is about to be unstable, generating excessive displacement after the next excavation to cause instability of the side wall of the foundation pit, stopping excavation at the moment, further strengthening the reinforcement measures and preparing evacuation personnel in time.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710780941.5A CN107587493B (en) | 2017-09-01 | 2017-09-01 | Foundation pit excavation monitoring and early warning method based on neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710780941.5A CN107587493B (en) | 2017-09-01 | 2017-09-01 | Foundation pit excavation monitoring and early warning method based on neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107587493A CN107587493A (en) | 2018-01-16 |
| CN107587493B true CN107587493B (en) | 2021-07-09 |
Family
ID=61050897
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710780941.5A Active CN107587493B (en) | 2017-09-01 | 2017-09-01 | Foundation pit excavation monitoring and early warning method based on neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107587493B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109543237B (en) * | 2018-10-29 | 2023-09-15 | 苏州科技大学 | Foundation pit displacement prediction method based on GA-BP neural network |
| CN112459080B (en) * | 2020-11-16 | 2022-04-15 | 上海市机械施工集团有限公司 | Intelligent adjusting method for steel support axial force based on bp neural network |
| CN112505033B (en) * | 2020-12-28 | 2024-05-28 | 中科朗劢技术有限公司 | Efficient super-digging exploratory pit detection method for construction process |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103116804A (en) * | 2013-01-31 | 2013-05-22 | 苏州科技学院 | Fuzzy neural network model and intelligent prediction method for deep excavation deformation |
| CN104501766A (en) * | 2014-12-25 | 2015-04-08 | 青岛理工大学 | Deep foundation pit excavation slope vertical displacement vector angle monitoring parameter and early warning method |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8577830B2 (en) * | 2010-11-24 | 2013-11-05 | Technion Research And Development Foundation Ltd. | Method for detection of tunnel excavation by brillouin optical time domain reflectometry using multiple underground optical fibers |
-
2017
- 2017-09-01 CN CN201710780941.5A patent/CN107587493B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103116804A (en) * | 2013-01-31 | 2013-05-22 | 苏州科技学院 | Fuzzy neural network model and intelligent prediction method for deep excavation deformation |
| CN104501766A (en) * | 2014-12-25 | 2015-04-08 | 青岛理工大学 | Deep foundation pit excavation slope vertical displacement vector angle monitoring parameter and early warning method |
Non-Patent Citations (4)
| Title |
|---|
| BP神经网络在深基坑工程中的应用;齐干等;《建筑技术开发》;20041231(第05期);全文 * |
| 实测数据分析法在深基坑监测预测中的应用;李国满;《中国优秀硕士学位论文全文数据库工程科技Ⅱ辑（月刊 ）》;20090415(第4期);第3-4节、第6节 * |
| 某深基坑监测及变形预测模型研究;董文宝;《中国优秀硕士学位论文全文数据库工程科技Ⅱ辑（月刊）》;20131215(第S2期);第4节 * |
| 深基坑支护结构安全预警系统研究;李惠强等;《华中科技大学学报(城市科学版)》;20020331(第01期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107587493A (en) | 2018-01-16 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Lin et al. | A new quantitative method for risk assessment of water inrush in karst tunnels based on variable weight function and improved cloud model | |
| Cheng et al. | Developing a risk assessment method for complex pipe jacking construction projects | |
| CN103743441B (en) | The polynary of a kind of side slope safety is coupling in line monitoring system and method | |
| CN104501766B (en) | Deep foundation pit excavation slope vertical displacement vector angle monitoring parameter and early warning method | |
| Jan et al. | Neural network forecast model in deep excavation | |
| CN107587493B (en) | Foundation pit excavation monitoring and early warning method based on neural network | |
| CN111259605A (en) | Soil landslide monitoring, early warning and evaluating method | |
| CN113779835A (en) | AI and intelligent monitoring system based deep and large foundation pit safety early warning method | |
| CN105095679A (en) | Security risk early warning measurement and judgment method of foundation pit tunnel engineering | |
| CN110705795A (en) | Tunnel gush risk grade fuzzy prediction method based on comprehensive advanced geological prediction technology | |
| KR20190061302A (en) | Apparatus for predicting advance rate of shield tunnel boring machine and method thereof | |
| CN111119902B (en) | Tunnel dynamic construction method based on BP neural network | |
| CN115495956B (en) | Safety evaluation method for unloading deformation of deep and large rock foundation pit | |
| JP7186472B1 (en) | Risk assessment method for large-diameter cast-in-place pile construction in deep water based on a two-dimensional cloud model | |
| CN109241627A (en) | The dynamic shoring method of probability hierarchical and the device of Automated Design supporting scheme | |
| CN117195507B (en) | Simulation prediction method and system for landslide river-plugging impact dam area building process | |
| CN113408184B (en) | Anchoring slope safety evaluation method based on genetic algorithm and discrete element analysis method | |
| CN109781044A (en) | Comprehensive gradual approximation early warning method for slope instability | |
| CN109948198A (en) | A kind of fender graded method for evaluating reliability based on nonlinear function | |
| CN112035941A (en) | Prediction method for surface subsidence of deep foundation pit excavation based on BAS-BP model | |
| CN112926110A (en) | Real-time visual early warning method for risks in construction process of subway station | |
| CN111551438A (en) | Method for evaluating large deformation anchoring control effect of soft rock of large buried depth tunnel | |
| CN114997671A (en) | Foundation pit deformation safety risk assessment method based on artificial neural network and entropy method | |
| CN109711063B (en) | Station underground excavation footage adjusting method and device suitable for soft upper and hard lower strata | |
| Yin et al. | Practice of optimisation theory in geotechnical engineering |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| CB03 | Change of inventor or designer information | ||
| CB03 | Change of inventor or designer information |
Inventor after: Guo Lu Inventor after: He Keqiang Inventor after: Niu Xiao Inventor before: He Keqiang Inventor before: Niu Xiao Inventor before: Guo Lu |
|
| GR01 | Patent grant | ||
| GR01 | Patent grant |