CN118094694A - Training method and device for large-diameter shield tunneling earth surface subsidence intelligent prediction model - Google Patents
Training method and device for large-diameter shield tunneling earth surface subsidence intelligent prediction model Download PDFInfo
- Publication number
- CN118094694A CN118094694A CN202410058408.8A CN202410058408A CN118094694A CN 118094694 A CN118094694 A CN 118094694A CN 202410058408 A CN202410058408 A CN 202410058408A CN 118094694 A CN118094694 A CN 118094694A
- Authority
- CN
- China
- Prior art keywords
- training data
- initial
- prediction model
- data set
- determining
- 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.)
- Pending
Links
- 238000012549 training Methods 0.000 title claims abstract description 190
- 230000005641 tunneling Effects 0.000 title claims abstract description 75
- 238000000034 method Methods 0.000 title claims abstract description 42
- 238000004062 sedimentation Methods 0.000 claims abstract description 71
- 210000002569 neuron Anatomy 0.000 claims abstract description 43
- 238000007781 pre-processing Methods 0.000 claims abstract description 12
- 238000012360 testing method Methods 0.000 claims description 36
- 239000011159 matrix material Substances 0.000 claims description 26
- 238000005259 measurement Methods 0.000 claims description 17
- 238000010276 construction Methods 0.000 claims description 14
- 239000003550 marker Substances 0.000 claims description 8
- 238000004422 calculation algorithm Methods 0.000 description 18
- 238000005457 optimization Methods 0.000 description 18
- 238000010586 diagram Methods 0.000 description 14
- 238000013461 design Methods 0.000 description 12
- 230000008569 process Effects 0.000 description 11
- 238000004364 calculation method Methods 0.000 description 10
- 238000013528 artificial neural network Methods 0.000 description 7
- 238000004590 computer program Methods 0.000 description 7
- 230000006870 function Effects 0.000 description 7
- 239000002689 soil Substances 0.000 description 7
- 238000011156 evaluation Methods 0.000 description 6
- 230000000694 effects Effects 0.000 description 5
- 238000010801 machine learning Methods 0.000 description 5
- 238000012545 processing Methods 0.000 description 5
- 210000004027 cell Anatomy 0.000 description 4
- 238000007637 random forest analysis Methods 0.000 description 4
- 239000011435 rock Substances 0.000 description 4
- 238000009412 basement excavation Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 238000012544 monitoring process Methods 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 230000015572 biosynthetic process Effects 0.000 description 2
- 238000004891 communication Methods 0.000 description 2
- 125000004122 cyclic group Chemical group 0.000 description 2
- 230000002068 genetic effect Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 239000002245 particle Substances 0.000 description 2
- 230000001360 synchronised effect Effects 0.000 description 2
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000003066 decision tree Methods 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000008034 disappearance Effects 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 238000011022 operating instruction Methods 0.000 description 1
- 230000000306 recurrent effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 239000002893 slag Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000008093 supporting effect Effects 0.000 description 1
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 1
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention provides a training method and training equipment for an intelligent prediction model of ground subsidence of a large-diameter shield tunneling ground, which can improve the accuracy of the prediction model in predicting ground subsidence. The method comprises the following steps: determining an influence range of influence factor time sequence data corresponding to a target area; determining an initial training data set from the shield tunneling parameter time sequence data based on the influence range; preprocessing an initial training data set; constructing an initial sedimentation prediction model based on a long-short-term memory network LSTM, wherein super parameters of the initial sedimentation prediction model comprise the number of hidden neurons, the learning rate and the number of LSTM layers; inputting target training data into an initial sedimentation prediction model to obtain a predicted sedimentation value; optimizing the number of hidden neurons, the learning rate and the LSTM layer number based on the predicted sedimentation value and the marking information; and performing iterative training based on the optimized hidden layer neuron number, the optimized learning rate and the optimized LSTM layer number to obtain an earth surface subsidence intelligent prediction model.
Description
Technical Field
The invention relates to the field of tunnel tunneling, in particular to a training method and equipment for an intelligent prediction model of ground subsidence of large-diameter shield tunneling.
Background
In the shield construction process of urban tunnels, shield tunneling construction inevitably causes stratum disturbance, thereby generating additional load on surface soil and adjacent structures, and then generating the processes of surface subsidence and deformation, thereby threatening ground construction and traffic roads. The artificial intelligence technology is greatly developed, and meanwhile, the requirement of development in an increasingly intelligent direction is also provided for the shield tunnel construction technology. The machine learning technology is used for predicting the earth surface subsidence in the shield construction process in real time, which is important for formulating construction strategies and ensuring construction safety.
Currently, a large number of researchers deeply discuss the problem of ground surface subsidence prediction caused by shield tunnel construction, wherein the traditional research means comprise an empirical formula method, a theoretical method, a numerical simulation method, a similar model analysis method and the like.
However, these methods all have certain limitations, such as 1) the difficulty in predicting due to the extensive uncertainty in the discrete strength of the rock-soil mass material; 2) The complexity of the model is due to the multitude of independent variables, which makes the model difficult to apply widely; 3) The dimension of the input parameters is limited, and the nonlinear relation analysis is not strong.
Disclosure of Invention
The embodiment of the invention provides a training method and training equipment for an intelligent prediction model of ground subsidence of a large-diameter shield tunneling ground surface, which can improve the accuracy of the prediction model in predicting ground subsidence.
The first aspect of the invention provides a training method of an intelligent prediction model for ground subsidence of large-diameter shield tunneling, which comprises the following steps:
determining an influence range of influence factor time sequence data corresponding to a target area;
determining an initial training data set from the shield tunneling parameter time sequence data based on the influence range, wherein each initial training data set is a tunneling machine operation parameter and a stratum characteristic parameter at different moments;
preprocessing the initial training data set to obtain a model training data set, wherein the model training data set comprises a training data set and a test data set;
Constructing an initial sedimentation prediction model based on a long-short-term memory network LSTM, wherein super parameters of the initial sedimentation prediction model comprise the number of hidden neurons, the learning rate and the number of LSTM layers;
inputting target training data into the initial sedimentation prediction model to obtain a predicted sedimentation value corresponding to the target training data, wherein the target training data is any training data in the training data set;
optimizing the number of hidden neurons, the learning rate and the LSTM layer number based on the predicted sedimentation value and the marking information corresponding to the target training data;
And performing iterative training based on the optimized hidden layer neuron number, the optimized learning rate and the optimized LSTM layer number until a preset iteration termination condition is reached, so as to obtain an intelligent prediction model of the earth surface subsidence.
In one possible design, the determining the influence range of the influence factor time sequence data corresponding to the target area includes:
acquiring a position coordinate set of all measuring points which meet the sedimentation rate requirement in the target area;
constructing a covariance matrix corresponding to the time sequence data of the influence factors based on the measuring point position coordinate set;
determining a first eigenvalue of a maximum eigenvector and a second eigenvalue of a minimum eigenvector corresponding to the covariance matrix;
and determining the influence range according to the first characteristic value and the second characteristic value.
In one possible design, the determining the influence range according to the first feature value and the second feature value includes:
determining an initial range corresponding to the target area according to the first characteristic value and the second characteristic value;
determining an adjustment long half shaft corresponding to the initial range and an inclination angle corresponding to the adjustment long half shaft;
And adjusting the initial range according to the adjustment long half shaft and the inclination angle to obtain the influence range.
In one possible design, the determining the initial range corresponding to the target area according to the first feature value and the second feature value includes:
determining an initial range corresponding to the target area through the following formula;
Wherein s is the initial range, lambda 2 is the first eigenvalue, lambda 2 is the second eigenvalue, (x, y) is the measurement point position coordinate in the measurement point position coordinate set, and (x p,yp) is the estimated central position coordinate of the initial range.
In one possible design, the determining the adjustment major axis corresponding to the initial range and the inclination angle corresponding to the adjustment major axis includes:
The adjustment major half axis is determined by the following formula:
Wherein a is the adjustment long half shaft, s is the initial range, and lambda 1 is the first characteristic value;
The tilt angle is determined by the following formula:
Wherein α is the tilt angle, v 1 is the eigenvector of the maximum eigenvalue corresponding to the covariance matrix, and (x, y) is the measurement point position coordinate in the measurement point position coordinate set.
In one possible design, the constructing the covariance matrix corresponding to the time-series data of the influencing factors based on the position coordinates of the measuring points includes:
The covariance of any one test position coordinate in the test point position coordinate set is calculated through the following formula:
Sigma (x, y) is the covariance of any one test position coordinate in the test point position coordinate set, (x i,yi) is any one test position coordinate in the test point position coordinate set, n is the number of test position coordinates included in the test point position coordinate set, And/>The sample mean value corresponding to the two random variables;
constructing the covariance matrix based on the covariance by the following formula:
x 1 is the first independent variable and x d is the d-th independent variable.
In one possible design, the optimizing the number of hidden neurons, the learning rate, and the LSTM layer number based on the predicted sedimentation value and the label information corresponding to the target training data includes:
based on the predicted sedimentation value and the marking information corresponding to the target training data, optimizing the hidden neuron number, the learning rate and the LSTM layer number according to the following formula:
Wherein f is the hidden neuron number, the learning rate and the LSTM number of layers, p (D 1:t |f) is likelihood distribution of y, p (f) is prior probability distribution of f, p (f|d 1:t) is posterior probability distribution of f, y is the predicted sedimentation value, D 1:t is a set of the predicted sedimentation value and the marker information, and D 1:t is determined by the following formula:
D1:t＝{(x1,y1),(x2,y2),(x3,y3),...,(xt,yt)};
x t is the t-th marker information, and y t is the t-th predicted sedimentation value;
Y t is determined by the following formula:
yt＝f(xt)+εt；
Epsilon t is the observed error.
The second aspect of the invention provides a training device of an intelligent prediction model of ground subsidence of large-diameter shield tunneling, which comprises the following components:
The first determining module is used for determining the influence range of the influence factor time sequence data corresponding to the target area;
The second determining module is used for determining an initial training data set from the shield tunneling parameter time sequence data based on the influence range, wherein each initial training data set in the initial training data set is a tunneling machine operation parameter and a stratum characteristic parameter at different moments;
The preprocessing module is used for preprocessing the initial training data set to obtain a model training data set, wherein the model training data set comprises a training data set and a test data set;
The construction module is used for constructing an initial sedimentation prediction model based on a long-short-term memory network LSTM, wherein super parameters of the initial sedimentation prediction model comprise the number of hidden neurons, the learning rate and the number of LSTM layers;
The input module is used for inputting target training data into the initial sedimentation prediction model to obtain a predicted sedimentation value corresponding to the target training data, wherein the target training data is any one training data in the training data set;
The optimizing module is used for optimizing the number of hidden neurons, the learning rate and the LSTM layer number based on the predicted sedimentation value and the marking information corresponding to the target training data;
And the training module is used for carrying out iterative training based on the optimized hidden layer neuron number, the optimized learning rate and the optimized LSTM layer number until a preset iteration termination condition is reached, so as to obtain an intelligent prediction model of the earth surface subsidence.
The third aspect of the invention provides an electronic device, which comprises a memory and a processor, wherein the processor is used for realizing the steps of the training method of the large-diameter shield tunneling surface subsidence intelligent prediction model according to the first aspect when executing a computer management program stored in the memory.
A fourth aspect of the present invention provides a computer readable storage medium having stored thereon a computer management program which when executed by a processor implements the steps of the training method of the large diameter shield tunneling surface subsidence intelligent prediction model of the first aspect described above.
In summary, it can be seen that in the embodiment provided by the invention, the influence range of the surface subsidence is determined by constructing the calculation method of the covariance matrix influence range ellipse, and the operation parameter time sequence data in the influence range is used for training the surface subsidence intelligent prediction model, so that the accuracy of the surface subsidence intelligent prediction model prediction is obviously improved. Meanwhile, key information of the earth surface subsidence is revealed in time sequence data statistics characteristics of earth surface subsidence influence factors, and the key information is fully mined through the LSTM model, so that the prediction capability of a prediction algorithm model is improved, the prediction effect is better, and the engineering requirements are met.
Drawings
FIG. 1 is a schematic diagram of RNN and LSTM model structures according to an embodiment of the present invention;
FIG. 2 is a flow chart of a training method of an intelligent prediction model of large-diameter shield tunneling earth surface subsidence, which is provided by the embodiment of the invention;
FIG. 3 is a schematic diagram of an influence range provided by an embodiment of the present invention;
FIG. 4 is a schematic diagram of parameter selection according to an embodiment of the present invention;
FIG. 5 is a schematic diagram of a virtual structure of a training device of an intelligent prediction model of large-diameter shield tunneling earth surface subsidence, which is provided by the embodiment of the invention;
FIG. 6 is a schematic hardware structure diagram of a training device of an intelligent prediction model of large-diameter shield tunneling surface subsidence, which is provided by the embodiment of the invention;
Fig. 7 is a schematic diagram of an embodiment of an electronic device according to an embodiment of the present invention;
Fig. 8 is a schematic diagram of an embodiment of a computer readable storage medium according to an embodiment of the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to fall within the scope of the invention.
The Recurrent Neural Network (RNN) is superior to BPNN because of its timing of hidden layer input and output. By the cyclic operation, information is continuously stored, and the unfolding self-cyclic structure can be regarded as the same network to be duplicated and connected for a plurality of times. While RNNs can handle long-term dependency problems in theory, they cannot be realized in practice. Therefore, a special network structure LSTM (long-short-term memory neural network) is developed, and the problem of gradient disappearance in the training process is effectively solved.
Referring to fig. 1, fig. 1 is a schematic diagram of RNN and LSTM model structures that can be provided by the present invention, where the RNN repeating unit includes only one network layer, and the LSTM cell structure includes four network layers. Sigma and tanh are activation functions, sigma represents a sigmoid function (outputs 0 to 1), and tanh is a hyperbolic tangent function (outputs-1 to 1). Each time t has a hidden state C (cell state) and the contents are controlled by three gates: the amnestic gate (Forget Gate) retains cell states C t-1 to C t; an Input Gate (Input Gate) controls the Input x to be saved to C t; an Output Gate (Output Gate) control C t outputs to h t. Three control gates and one cell state vector are i (input gate), f t (forget gate), g t (input replenishment), o t (output gate), respectively. The forget gate, the input gate and the output gate are calculated as follows:
ft＝σ(Wf·[ht-1,xt]+bf)
it＝σ(Wi·[ht-1,xt]+bi)
Ct＝tanh(WC·[ht-1,xt]+bC)
ot＝σ(Wo·[ht-1,xt]+bo)。
The training method of the large-diameter shield tunneling surface subsidence intelligent prediction model is described below from the training device of the large-diameter shield tunneling surface subsidence intelligent prediction model, and the training device of the large-diameter shield tunneling surface subsidence intelligent prediction model can be a server or a service unit in the server, and is not particularly limited.
Referring to fig. 2, fig. 2 is a flow chart of a training method of an intelligent prediction model of ground subsidence in large-diameter shield tunneling according to an embodiment of the present invention, including:
201. And determining the influence range of the influence factor time sequence data corresponding to the target area.
In this embodiment, all the measuring point position coordinates meeting the sedimentation rate requirement in the target area are obtained, that is, all the measuring point position coordinates meeting the sedimentation rate requirement are obtained based on engineering field monitoring, and a calculation strategy of covariance matrix influence range ellipses is adopted to estimate potential influence areas of surface subsidence in the target area under different confidence levels, so that the influence range of influence factor time sequence data in the target area can be determined.
In one embodiment, determining the influence range of the influence factor time sequence data by the training device of the intelligent prediction model for the earth surface subsidence caused by the large-diameter shield tunneling comprises:
acquiring a position coordinate set of all measuring points which meet the sedimentation rate requirement in a target area;
constructing a covariance matrix corresponding to the time sequence data of the influence factors based on the measuring point position coordinate set;
determining a first eigenvalue of a maximum eigenvector and a second eigenvalue of a minimum eigenvector corresponding to the covariance matrix;
and determining the influence range according to the first characteristic value and the second characteristic value.
In this embodiment, the training device of the large-diameter shield tunneling earth surface subsidence intelligent prediction model may first obtain all measurement point position coordinate sets meeting the requirement of the subsidence rate in the target area, and construct a covariance matrix corresponding to the time sequence data of the influencing factors based on the measurement point position coordinate sets, where elements of the covariance matrix are formed by covariances among vector elements, and are used to measure the overall change condition of two random variables, and for the random variables, the covariance formula is as follows:
Where n is the total sample volume (i.e., the number of test position coordinates included in the set of test point position coordinates), AndFor the sample mean value corresponding to the two random variables, d independent variables x k (k=1, 2, …, d) and n monitoring samples are set, according to the definition of covariance, covariance between each pair of variables is calculated to obtain a calculation result, and the calculation results are orderly arranged to construct a covariance matrix, wherein the covariance matrix can be represented by the following formula:
Thereafter, the first eigenvalue λ 1 of the maximum eigenvector and the second eigenvalue λ 2 of the minimum eigenvector of the covariance matrix are acquired, and the manner of acquiring the first eigenvalue λ 1 and the second eigenvalue λ 2 is not particularly limited here.
And finally, determining an influence range according to the first characteristic value and the second characteristic value, wherein the influence range is specifically as follows:
In performing the interval estimation (i.e. determining the range of possible values of the range of influence) on the scale S, if an interval can be found for a given small probability β (S 1,s2), such that:
p(s1＜s＜s2)＝1-β
Wherein, (S 1,s2) is the confidence interval of the scale S, S 1 and S 2 are the critical values of the scale S, and if S is less than or equal to S 1 and S is more than or equal to S 2, determining whether the scale S is located in a domain; the probability β is defined as the significance level and 1- β is defined as the confidence level.
That is, an initial range S corresponding to the target area is first determined according to the first feature value and the second feature value, where the range of the ellipse is to include all coordinates in the coordinate set of the measuring point position, and the initial range S is determined according to the following formula:
Wherein s is an initial range, lambda 2 is a first characteristic value, lambda 2 is a second characteristic value, (x, y) is a measuring point position coordinate in a measuring point position coordinate set, and (x p,yp) is a central position coordinate estimated in the initial range;
Then, an adjusting long half shaft corresponding to the initial range and an inclination angle corresponding to the adjusting long half shaft are determined, and specifically, the adjusting long half shaft can be determined through the following formula:
Wherein a is an adjustment long half shaft, s is an initial range, and lambda 1 is a first characteristic value;
The tilt angle is determined by the following formula:
Wherein alpha is the inclination angle, v 1 is the eigenvector of the maximum eigenvalue corresponding to the covariance matrix, and (x, y) is the measuring point position coordinate in the measuring point position coordinate set.
Finally, after knowing the initial range and knowing the adjustment of the long half axis and the inclination angle alpha of the long half axis and the X axis, the initial range can be adjusted by adjusting the long half axis a and the inclination angle alpha to obtain the influence range.
Referring to fig. 3, fig. 3 is a schematic diagram of an influence range provided by an embodiment of the present invention, a two-dimensional plane data set is constructed based on a measurement point position coordinate set of time-series data of influence factors in a target area, then, based on a confidence interval length of the data set, we estimate an influence range of surface subsidence under different confidence levels, taking the progress of mining a target area to 167 th ring as an example, the total number of engineering measurement points of the position is 44 (i.e. a measurement point position coordinate set), and then, using 44 position coordinates, a covariance matrix of the two-dimensional coordinate data is calculated, and a final influence range 2a=66.96 m is calculated.
202. An initial training data set is determined from the shield tunneling parameter time series data based on the range of influence.
In this embodiment, the training device of the large-diameter shield tunneling earth surface subsidence intelligent prediction model may determine an initial training data set from the shield tunneling parameter time sequence data based on the influence range, perform statistical calculation on the average value of the influence ranges of each group of subsidence points under three confidence levels to obtain a calculation result, convert the calculation result into a shield tunneling ring number, and determine shield tunneling data corresponding to the shield tunneling ring number as initial training data, where each initial training data in the initial training data set is a tunneling machine operation parameter and a stratum characteristic parameter at different moments.
The degree of surface subsidence is affected by tunneling parameters, formation lithology parameters, abnormal conditions, and the like. In the tunneling process of the shield machine, a plurality of tunneling parameters such as the state of the shield machine, the oil temperature, the electromechanical parameters, the grouting pressure, the cutterhead state and the like are recorded according to different rock stratum traversed. During the construction process, the sensor automatically acquires various parameters, and the average value acquired in each ring is used as an input data set. Based on the mechanism that the earth surface subsidence is affected, the tunneling parameters and the stratum geological parameters are mainly considered as input parameters of earth surface subsidence prediction:
(1) Heading machine operating parameters
1) The pressure in the excavation cabin is kept slightly higher than the sum of the water and soil pressure and the lateral pressure of the face in the sealed cabin all the time in the shield tunneling process, so that the disturbance to the excavated earth surface soil body is reduced.
2) The control of the pushing speed of the shield can influence the amount of excavated soil, and the too fast speed can cause land loss so as to increase the ground subsidence.
3) Synchronous grouting pressure and grouting quantity, a gap exists between soil body and duct pieces, the safety of a ground building and an underlying pipeline is considered, shield tail synchronous grouting should be filled in time to generate supporting effect on stratum, stratum settlement is effectively controlled,
4) Cutter torque and cutter rotational speed. The torque of the cutter head is reasonably controlled in the shield construction process, so that disturbance of excavation to soil mass can be effectively reduced, and earth surface subsidence is controlled. Excessive rotational speed of the cutterhead can influence slag discharge, and further influence surface subsidence.
(2) Formation characteristic parameters
The invention uses the tunnel depth span ratio and surrounding rock grade as stratum characteristic parameters, wherein the tunnel depth span ratio is used for describing the influence of large-diameter shield tunneling on the ground subsidence, the ratio of the tunnel buried depth H measured by each ring in an interval longitudinal section geological map to the section diameter D of a large-diameter shield tunneling machine is used as geometric parameters, as shown in fig. 4, fig. 4 is a schematic diagram for parameter selection provided by the invention, the surrounding rock grade parameters are derived from engineering geological survey data, and the construction stratum types are graded in the engineering survey data. In the shield tunneling process, the penetrated stratum is different, the intensity of the penetrated rock mass is also different, tunneling parameters are influenced, and the influence on the earth surface subsidence is unavoidable.
203. The initial training data set is preprocessed to obtain the model training data set.
In this embodiment, after determining the initial training data set, the training device of the large-diameter shield tunneling earth surface subsidence intelligent prediction model may perform preprocessing on the initial training data set to obtain model training data, where the model training data includes a training data set and a test data set.
It should be noted that the pretreatment includes, but is not limited to, the following operations: extracting dependent variables and independent variables, processing missing data, encoding classification data, data normalization processing and the like.
204. And constructing an initial settlement prediction model based on the long-short-term memory network LSTM.
In this embodiment, the training device of the large-diameter shield tunneling earth surface subsidence intelligent prediction model may construct an initial subsidence model through LSTM, and set the super parameters of the initial Chen Jiange prediction model as hidden neurons, learning rate and LSTM layer number.
205. And inputting the target training data into an initial sedimentation prediction model to obtain a predicted sedimentation value corresponding to the target training data.
In this embodiment, after the initial subsidence prediction model is built, the training device of the large-diameter shield tunneling earth surface subsidence intelligent prediction model may arbitrarily select one training data from the training data set as the target training data, and input the target training data into the initial subsidence prediction model to obtain the corresponding predicted subsidence value.
206. And optimizing the number of hidden neurons, the learning rate and the LSTM layer number based on the predicted sedimentation value and the marking information corresponding to the target training data.
In this embodiment, after the training device of the large-diameter shield tunneling earth surface subsidence intelligent prediction model obtains the predicted subsidence value, the predicted subsidence value and the actual subsidence value may be compared, the distance between the original data and the predicted data may be determined, and the number of LSTM hidden layer neurons, the learning rate and the LSTM layer number may be optimized according to the distance, so as to shorten the distance between the predicted result and the actual result. When running LSTM, it is important to perform a hyper-parametric optimization, and the best hyper-parametric combination that can improve the model performance can be found. By selecting a proper hyper-parameter combination, the generalization capability of the model can be improved, so that better performance is obtained on the test set.
In the process of optimizing the super-parameters in the initial settlement prediction model, a bayesian Optimization algorithm (Bayesian Optimization, BO) may be used as the super-parameter Optimization algorithm, and of course, other Optimization algorithms may also be used as the super-parameter Optimization algorithm, for example, a genetic algorithm (Genetic Algorithm, GA), a differential evolution algorithm (DIFFERENTIAL EVOLUTION, DE), and a particle swarm Optimization algorithm (PARTICLE SWARM Optimization, PSO) are not particularly limited, so long as super-parameter Optimization can be achieved. The following is an example of bayesian optimization:
The Bayesian optimization algorithm is an efficient global optimization method, obtains an ideal solution through less-times objective function evaluation, is particularly suitable for solving complex optimization problems of multimodal objective function, missing value and high estimation cost, and particularly optimizes the number of hidden neurons, learning rate and LSTM layer number through the following formula:
Wherein f is the hidden neuron number, the learning rate and the LSTM layer number, p (D 1:t |f) is the likelihood distribution of y, p (f) is the prior probability distribution of f, p (f|d 1:t) is the posterior probability distribution of f, y is the predicted sedimentation value, D 1:t is the set of the predicted sedimentation value and the marker information, and D 1:t is determined by the following formula:
D1:t＝{(x1,y1),(x2,y2),(x3,y3),...,(xt,yt)};
x t is the t-th marker information, and y t is the t-th predicted sedimentation value;
Y t is determined by the following formula:
yt＝f(xt)+εt；
Epsilon t is the observed error.
The Bayesian optimization algorithm consists of posterior probability distribution and a maximized acquisition function, the calculation process is repeated until the difference between the predicted value and the optimal value is smaller than a threshold value, the acquisition function ensures that the total loss is minimized by the selected evaluation point sequence, and then the optimal super-parameters required by optimizing the neural network are obtained.
207. And performing iterative training based on the optimized hidden layer neuron number, the optimized learning rate and the optimized LSTM layer number until a preset iteration termination condition is reached, so as to obtain the intelligent prediction model of the earth surface subsidence.
In this embodiment, the training device of the large-diameter shield tunneling earth surface subsidence intelligent prediction model may perform iterative training based on the optimized LSTM hidden layer neuron number, the optimized learning rate and the optimized LSTM layer number until a preset iteration termination condition is reached, so as to obtain the earth surface subsidence intelligent prediction model, that is, after each training is completed, whether the current iteration number reaches a preset value or whether the super-parameter converges is judged, if the current iteration number reaches the preset value or the super-parameter converges, it is determined that a preset iteration termination condition is reached, otherwise, it is determined that the preset iteration termination condition is not reached, and training is continued.
It should be noted that, different evaluation indexes exist in the prediction model, the directions of emphasis are different, and a single index lacks comprehensiveness in model evaluation. After the intelligent prediction model of the earth surface subsidence is obtained by training, in order to evaluate the quality of the earth surface subsidence model, the following evaluation indexes are respectively determined coefficients (R 2), mean Absolute Error (MAE), root Mean Square Error (RMSE) and Mean Absolute Percentage Error (MAPE), and are defined as follows:
MAE reflects how close the actual value is to the predicted value, and RMSE is also used for error rate calculation. For a model, the closer R 2 is to 1, the smaller the MAE and RMSE, and the more optimal the model predicts. MAPE is very sensitive to extreme values, which makes it not very good for handling outliers, but it converts errors into a percentage form, which makes it somewhat interpretable and comparable.
It is to be noted that, according to engineering monitoring data, 880 sets of original settlement measuring point settlement data are obtained altogether, a total of 200 loops of shield machine data are obtained, and the average characteristics of the operation parameter time series data in the earth surface settlement influence range caused by ultra-large diameter shield excavation are used as input data of a machine learning prediction model to respectively establish large-diameter shield settlement prediction models based on Bayesian optimization LSTM, BP neural network and random forest algorithm. Considering that the three algorithms are superior to each other under Bayesian optimization and are limited by the performance of the existing computer, the super parameters of the key of the optimization model should be selected as much as possible according to the previous experience. For the LSTM model, the hidden neurons, the learning rate and the LSTM layer number of the LSTM model are selected and optimized; for the BP neural network model, selecting and optimizing the input batch size and the learning rate; for the random forest model, an optimal decision tree is selected. Model specific parameter presets are shown in table 1 below.
TABLE 1
And substituting tunneling parameters and geological parameters in the ranges under 3 different confidence intervals as input parameters into LSTM, BP neural network and random forest settlement prediction model based on Bayesian optimization for training, wherein the effect of the table settlement prediction model is closely related to the selection of the time sequence data acquisition range of the influencing factors. At 90% confidence level, both MAE and RMSE were at a lower level, while the larger R 2 fit performed well, and the three algorithm models all showed the best prediction. The tunneling parameters in the 90% confidence interval range have more compact influence on the surface subsidence, and the overall prediction accuracy shows a trend of decreasing along with the increase of the number of loops of input data along with the increase of the selected interval range. Meanwhile, the integral prediction effect of the three models is compared, so that the LSTM model prediction effect is obviously better than that of the traditional machine learning algorithm (BP neural network and random forest), the MAE and RMSE evaluation indexes are obviously smaller than that of the traditional machine learning algorithm, and the MAPE can reach 8.91% at the minimum. The method also shows that the shield tunneling data has time sequence value, the traditional machine learning algorithm cannot mine the time sequence value of the shield, and a shield tunneling data prediction model is built through a cyclic neural network, so that the earth surface subsidence rule changing along with time can be reflected.
In summary, it can be seen that in the embodiment provided by the invention, the influence range of the surface subsidence is determined by constructing the calculation method of the covariance matrix influence range ellipse, and the operation parameter time sequence data in the influence range is used for training the surface subsidence intelligent prediction model, so that the accuracy of the surface subsidence intelligent prediction model prediction is obviously improved. Meanwhile, key information of the earth surface subsidence is revealed in time sequence data statistics characteristics of earth surface subsidence influence factors, and the key information is fully mined through the LSTM model, so that the prediction capability of a prediction algorithm model is improved, the prediction effect is better, and the engineering requirements are met.
The embodiment of the invention is described above by the training method of the large-diameter shield tunneling surface subsidence intelligent prediction model, and the embodiment of the invention is described below by the training device of the large-diameter shield tunneling surface subsidence intelligent prediction model.
Referring to fig. 5, in an embodiment of the present invention, a virtual structure diagram of a training device for a large-diameter shield tunneling surface subsidence intelligent prediction model, where the training device 500 for a large-diameter shield tunneling surface subsidence intelligent prediction model includes:
a first determining module 501, configured to determine an influence range of influence factor time sequence data corresponding to a target area;
A second determining module 502, configured to determine an initial training data set from the shield tunneling parameter time sequence data based on the influence range, where each initial training data in the initial training data set is a tunneling machine operation parameter and a stratum characteristic parameter at different moments;
A preprocessing module 503, configured to preprocess the initial training data set to obtain a model training data set, where the model training data set includes a training data set and a test data set;
the construction module 504 is configured to construct an initial sedimentation prediction model based on the long-short-term memory network LSTM, where super parameters of the initial sedimentation prediction model include the number of hidden neurons, the learning rate, and the LSTM layer number;
The input module 505 is configured to input target training data into the initial sedimentation prediction model to obtain a predicted sedimentation value corresponding to the target training data, where the target training data is any one training data in the training data set;
An optimizing module 506, configured to optimize the number of hidden neurons, the learning rate, and the LSTM layer number based on the predicted sedimentation value and the label information corresponding to the target training data;
The training module 507 is configured to perform iterative training based on the optimized number of neurons in the hidden layer, the optimized learning rate, and the optimized LSTM layer number, until a preset iteration termination condition is reached, so as to obtain an intelligent prediction model of surface subsidence.
In one possible design, the first determining module 501 is specifically configured to:
acquiring a position coordinate set of all measuring points which meet the sedimentation rate requirement in a target area;
constructing a covariance matrix corresponding to the time sequence data of the influence factors based on the measuring point position coordinate set;
determining a first eigenvalue of a maximum eigenvector and a second eigenvalue of a minimum eigenvector corresponding to the covariance matrix;
and determining the influence range according to the first characteristic value and the second characteristic value.
In one possible design, the determining, by the first determining module 501, the influence range according to the first feature value and the second feature value includes:
determining an initial range corresponding to the target area according to the first characteristic value and the second characteristic value;
determining an adjustment long half shaft corresponding to the initial range and an inclination angle corresponding to the adjustment long half shaft;
And adjusting the initial range according to the adjustment long half shaft and the inclination angle to obtain the influence range.
In one possible design, the first determining module 501 is further specifically configured to:
determining an initial range corresponding to the target area through the following formula;
Wherein s is the initial range, lambda 2 is the first eigenvalue, lambda 2 is the second eigenvalue, (x, y) is the measurement point position coordinate in the measurement point position coordinate set, and (x p,yp) is the estimated central position coordinate of the initial range.
In one possible design, the first determining module 501 is further specifically configured to:
The adjustment major half axis is determined by the following formula:
Wherein a is the adjustment long half shaft, s is the initial range, and lambda 1 is the first characteristic value;
The tilt angle is determined by the following formula:
Wherein α is the tilt angle, v 1 is the eigenvector of the maximum eigenvalue corresponding to the covariance matrix, and (x, y) is the measurement point position coordinate in the measurement point position coordinate set.
In one possible design, the first determining module 501 is further specifically configured to:
The covariance of any one test position coordinate in the test point position coordinate set is calculated through the following formula:
Sigma (x, y) is the covariance of any one test position coordinate in the test point position coordinate set, (x i,yi) is any one test position coordinate in the test point position coordinate set, n is the number of test position coordinates included in the test point position coordinate set, And y is the sample mean value corresponding to the two random variables;
constructing the covariance matrix based on the covariance by the following formula:
x 1 is the first independent variable and x d is the d-th independent variable.
In one possible design, the optimizing module 506 is specifically configured to:
based on the predicted sedimentation value and the marking information corresponding to the target training data, optimizing the hidden neuron number, the learning rate and the LSTM layer number according to the following formula:
Wherein f is the hidden neuron number, the learning rate and the LSTM number of layers, p (D 1:t |f) is likelihood distribution of y, p (f) is prior probability distribution of f, p (f|d 1:t) is posterior probability distribution of f, y is the predicted sedimentation value, D 1:t is a set of the predicted sedimentation value and the marker information, and D 1:t is determined by the following formula:
D1:t＝{(x1,y1),(x2,y2),(x3,y3),...,(xt,yt)};
x t is the t-th marker information, and y t is the t-th predicted sedimentation value;
Y t is determined by the following formula:
yt＝f(xt)+εt；
Epsilon t is the observed error.
Fig. 5 above describes the training device for the intelligent prediction model of subsidence of the earth surface in the large-diameter shield tunneling in the embodiment of the present invention from the perspective of the modularized functional entity, and the following describes the training device for the intelligent prediction model of subsidence of the earth surface in the large-diameter shield tunneling in the embodiment of the present invention from the perspective of hardware processing, referring to fig. 6, an embodiment of the training device 600 for the intelligent prediction model of subsidence of the earth surface in the large-diameter shield tunneling in the embodiment of the present invention is shown in a schematic diagram, and the training device 600 for the intelligent prediction model of subsidence of the earth surface in the large-diameter shield tunneling comprises:
Input device 601, output device 602, processor 603, and memory 604 (where the number of processors 603 may be one or more, one processor 603 is illustrated in fig. 6). In some embodiments of the invention, the input device 601, output device 602, processor 603, and memory 604 may be connected by a communication bus or other means, where a communication bus connection is exemplified in fig. 6.
The processor 603 is configured to execute the following steps by calling the operation instructions stored in the memory 604:
determining an influence range of influence factor time sequence data corresponding to a target area;
Determining an initial training data set from the shield tunneling parameter time sequence data based on the influence range, wherein each initial training data in the initial training data set is a tunneling machine operation parameter and a stratum characteristic parameter at different moments;
preprocessing the initial training data set to obtain a model training data set, wherein the model training data set comprises a training data set and a test data set;
Constructing an initial sedimentation prediction model based on a long-short-term memory network LSTM, wherein super parameters of the initial sedimentation prediction model comprise the number of hidden neurons, the learning rate and the number of LSTM layers;
inputting target training data into the initial sedimentation prediction model to obtain a predicted sedimentation value corresponding to the target training data, wherein the target training data is any training data in the training data set;
optimizing the number of hidden neurons, the learning rate and the LSTM layer number based on the predicted sedimentation value and the marking information corresponding to the target training data;
And performing iterative training based on the optimized hidden layer neuron number, the optimized learning rate and the optimized LSTM layer number until a preset iteration termination condition is reached, so as to obtain an intelligent prediction model of the earth surface subsidence.
The processor 603 is further configured to execute any of the embodiments corresponding to fig. 2 by invoking the operating instructions stored in the memory 604.
Referring to fig. 7, fig. 7 is a schematic diagram of an embodiment of an electronic device according to an embodiment of the invention.
As shown in fig. 7, an embodiment of the present invention provides an electronic device including a memory 710, a processor 720, and a computer program 711 stored on the memory 710 and executable on the processor 720, the processor 720 implementing the following steps when executing the computer program 711:
determining an influence range of influence factor time sequence data corresponding to a target area;
Determining an initial training data set from the shield tunneling parameter time sequence data based on the influence range, wherein each initial training data in the initial training data set is a tunneling machine operation parameter and a stratum characteristic parameter at different moments;
preprocessing the initial training data set to obtain a model training data set, wherein the model training data set comprises a training data set and a test data set;
Constructing an initial sedimentation prediction model based on a long-short-term memory network LSTM, wherein super parameters of the initial sedimentation prediction model comprise the number of hidden neurons, the learning rate and the number of LSTM layers;
inputting target training data into the initial sedimentation prediction model to obtain a predicted sedimentation value corresponding to the target training data, wherein the target training data is any training data in the training data set;
optimizing the number of hidden neurons, the learning rate and the LSTM layer number based on the predicted sedimentation value and the marking information corresponding to the target training data;
And performing iterative training based on the optimized hidden layer neuron number, the optimized learning rate and the optimized LSTM layer number until a preset iteration termination condition is reached, so as to obtain an intelligent prediction model of the earth surface subsidence.
In a specific implementation, when the processor 720 executes the computer program 711, any implementation of the embodiment corresponding to fig. 2 may be implemented.
Because the electronic device described in this embodiment is a device used for implementing the training apparatus for the large-diameter shield tunneling earth surface subsidence intelligent prediction model in this embodiment of the present invention, based on the method described in this embodiment of the present invention, those skilled in the art can understand the specific implementation manner of the electronic device in this embodiment and various modifications thereof, so how the electronic device implements the method in this embodiment of the present invention will not be described in detail herein, and only the device used by those skilled in the art to implement the method in this embodiment of the present invention is within the scope of the present invention to be protected.
Referring to fig. 8, fig. 8 is a schematic diagram of an embodiment of a computer readable storage medium according to an embodiment of the invention.
As shown in fig. 8, an embodiment of the present invention further provides a computer-readable storage medium 800 having stored thereon a computer program 811, which computer program 811 when executed by a processor performs the steps of:
determining an influence range of influence factor time sequence data corresponding to a target area;
Determining an initial training data set from the shield tunneling parameter time sequence data based on the influence range, wherein each initial training data in the initial training data set is a tunneling machine operation parameter and a stratum characteristic parameter at different moments;
preprocessing the initial training data set to obtain a model training data set, wherein the model training data set comprises a training data set and a test data set;
Constructing an initial sedimentation prediction model based on a long-short-term memory network LSTM, wherein super parameters of the initial sedimentation prediction model comprise the number of hidden neurons, the learning rate and the number of LSTM layers;
inputting target training data into the initial sedimentation prediction model to obtain a predicted sedimentation value corresponding to the target training data, wherein the target training data is any training data in the training data set;
optimizing the number of hidden neurons, the learning rate and the LSTM layer number based on the predicted sedimentation value and the marking information corresponding to the target training data;
And performing iterative training based on the optimized hidden layer neuron number, the optimized learning rate and the optimized LSTM layer number until a preset iteration termination condition is reached, so as to obtain an intelligent prediction model of the earth surface subsidence.
In a specific implementation, the computer program 811 is executed by a processor at a time to implement any of the embodiments corresponding to fig. 2.
In the foregoing embodiments, the descriptions of the embodiments are focused on, and for those portions of one embodiment that are not described in detail, reference may be made to the related descriptions of other embodiments.
Embodiments of the present invention also provide a computer program product comprising computer software instructions which, when run on a processing device, cause the processing device to perform the flow as in the corresponding embodiment of fig. 2.
The above embodiments are only for illustrating the technical solution of the present invention, and not for limiting the same; although the invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some of the technical features thereof can be replaced equivalently; such modifications and substitutions do not depart from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims (10)
1. The training method of the intelligent prediction model for the earth surface subsidence of the large-diameter shield tunneling is characterized by comprising the following steps of:
determining an influence range of influence factor time sequence data corresponding to a target area;
Determining an initial training data set from the shield tunneling parameter time sequence data based on the influence range, wherein each initial training data in the initial training data set is a tunneling machine operation parameter and a stratum characteristic parameter at different moments;
preprocessing the initial training data set to obtain a model training data set, wherein the model training data set comprises a training data set and a test data set;
Constructing an initial sedimentation prediction model based on a long-short-term memory network LSTM, wherein super parameters of the initial sedimentation prediction model comprise the number of hidden neurons, the learning rate and the number of LSTM layers;
inputting target training data into the initial sedimentation prediction model to obtain a predicted sedimentation value corresponding to the target training data, wherein the target training data is any training data in the training data set;
optimizing the number of hidden neurons, the learning rate and the LSTM layer number based on the predicted sedimentation value and the marking information corresponding to the target training data;
And performing iterative training based on the optimized hidden layer neuron number, the optimized learning rate and the optimized LSTM layer number until a preset iteration termination condition is reached, so as to obtain an intelligent prediction model of the earth surface subsidence.
2. The method of claim 1, wherein determining the range of influence of the time series data of the influencing factor corresponding to the target region comprises:
acquiring a position coordinate set of all measuring points which meet the sedimentation rate requirement in the target area;
constructing a covariance matrix corresponding to the time sequence data of the influence factors based on the measuring point position coordinate set;
determining a first eigenvalue of a maximum eigenvector and a second eigenvalue of a minimum eigenvector corresponding to the covariance matrix;
and determining the influence range according to the first characteristic value and the second characteristic value.
3. The method of claim 2, wherein the determining the range of influence from the first characteristic value and the second characteristic value comprises:
determining an initial range corresponding to the target area according to the first characteristic value and the second characteristic value;
determining an adjustment long half shaft corresponding to the initial range and an inclination angle corresponding to the adjustment long half shaft;
And adjusting the initial range according to the adjustment long half shaft and the inclination angle to obtain the influence range.
4. The method of claim 3, wherein determining the initial range corresponding to the target area according to the first feature value and the second feature value comprises:
determining an initial range corresponding to the target area through the following formula;
Wherein s is the initial range, lambda 2 is the first eigenvalue, lambda 2 is the second eigenvalue, (x, y) is the measurement point position coordinate in the measurement point position coordinate set, and (x p,yp) is the estimated central position coordinate of the initial range.
5. The method of claim 3, wherein determining the adjustment major axis corresponding to the initial range and the tilt angle corresponding to the adjustment major axis comprises:
The adjustment major half axis is determined by the following formula:
Wherein a is the adjustment long half shaft, s is the initial range, and lambda 1 is the first characteristic value;
The tilt angle is determined by the following formula:
Wherein α is the tilt angle, v 1 is the eigenvector of the maximum eigenvalue corresponding to the covariance matrix, and (x, y) is the measurement point position coordinate in the measurement point position coordinate set.
6. The method of claim 2, wherein constructing a covariance matrix corresponding to the time-series data of the influencing factors based on the position coordinates of the measuring points comprises:
The covariance of any one test position coordinate in the test point position coordinate set is calculated through the following formula:
Sigma (x, y) is the covariance of any one test position coordinate in the test point position coordinate set, (x i,yi) is any one test position coordinate in the test point position coordinate set, n is the number of test position coordinates included in the test point position coordinate set, And/>The sample mean value corresponding to the two random variables;
constructing the covariance matrix based on the covariance by the following formula:
x 1 is the first independent variable and x d is the d-th independent variable.
7. The method of any one of claims 1 to 6, wherein optimizing the number of hidden neurons, the learning rate, and the LSTM layer number based on the predicted sedimentation value and the label information corresponding to the target training data comprises:
based on the predicted sedimentation value and the marking information corresponding to the target training data, optimizing the hidden neuron number, the learning rate and the LSTM layer number according to the following formula:
Wherein f is the hidden neuron number, the learning rate and the LSTM number of layers, p (D 1:t |f) is likelihood distribution of y, p (f) is prior probability distribution of f, p (f|d 1:t) is posterior probability distribution of f, y is the predicted sedimentation value, D 1:t is a set of the predicted sedimentation value and the marker information, and D 1:t is determined by the following formula:
D1:t＝{(x1,y1),(x2,y2),(x3,y3),...,(xt,yt)};
x t is the t-th marker information, and y t is the t-th predicted sedimentation value;
Y t is determined by the following formula:
yt＝f(xt)+εt；
Epsilon t is the observed error.
8. The utility model provides a training device of intelligent prediction model of major diameter shield tunneling earth's surface subsidence which characterized in that includes:
The first determining module is used for determining the influence range of the influence factor time sequence data corresponding to the target area;
The second determining module is used for determining an initial training data set from the shield tunneling parameter time sequence data based on the influence range, wherein each initial training data set in the initial training data set is a tunneling machine operation parameter and a stratum characteristic parameter at different moments;
The preprocessing module is used for preprocessing the initial training data set to obtain a model training data set, wherein the model training data set comprises a training data set and a test data set;
The construction module is used for constructing an initial sedimentation prediction model based on a long-short-term memory network LSTM, wherein super parameters of the initial sedimentation prediction model comprise the number of hidden neurons, the learning rate and the number of LSTM layers;
The input module is used for inputting target training data into the initial sedimentation prediction model to obtain a predicted sedimentation value corresponding to the target training data, wherein the target training data is any one training data in the training data set;
The optimizing module is used for optimizing the number of hidden neurons, the learning rate and the LSTM layer number based on the predicted sedimentation value and the marking information corresponding to the target training data;
And the training module is used for carrying out iterative training based on the optimized hidden layer neuron number, the optimized learning rate and the optimized LSTM layer number until a preset iteration termination condition is reached, so as to obtain an intelligent prediction model of the earth surface subsidence.
9. An electronic device, comprising:
Memory, processor, the processor is used for realizing the training method of the intelligent prediction model of large-diameter shield tunneling earth surface subsidence according to any one of claims 1 to 8 when executing the computer management class program stored in the memory.
10. A computer-readable storage medium having stored thereon a computer management class program, characterized in that: the training method of the intelligent prediction model for large-diameter shield tunneling surface subsidence is realized when the computer management program is executed by a processor.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410058408.8A CN118094694A (en) | 2024-01-15 | 2024-01-15 | Training method and device for large-diameter shield tunneling earth surface subsidence intelligent prediction model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410058408.8A CN118094694A (en) | 2024-01-15 | 2024-01-15 | Training method and device for large-diameter shield tunneling earth surface subsidence intelligent prediction model |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN118094694A true CN118094694A (en) | 2024-05-28 |
Family
ID=91151624
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410058408.8A Pending CN118094694A (en) | 2024-01-15 | 2024-01-15 | Training method and device for large-diameter shield tunneling earth surface subsidence intelligent prediction model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN118094694A (en) |
-
2024
- 2024-01-15 CN CN202410058408.8A patent/CN118094694A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Zhu et al. | Prediction of rockhead using a hybrid N-XGBoost machine learning framework | |
| Wang et al. | A hybrid intelligent approach for constructing landslide displacement prediction intervals | |
| Shahin | Artificial intelligence in geotechnical engineering: applications, modeling aspects, and future directions | |
| CN109978226B (en) | Shield construction ground settlement prediction method based on cyclic neural network | |
| Li et al. | An approach using random forest intelligent algorithm to construct a monitoring model for dam safety | |
| Fattahi | Applying soft computing methods to predict the uniaxial compressive strength of rocks from schmidt hammer rebound values | |
| Njock et al. | Artificial neural network optimized by differential evolution for predicting diameters of jet grouted columns | |
| KR101967978B1 (en) | Apparatus for predicting net penetration rate of shield tunnel boring machine and method thereof | |
| Khan et al. | Prediction of the residual strength of clay using functional networks | |
| CN113204824B (en) | Multi-model fusion shield construction settlement prediction method and system | |
| Xue et al. | Seismic liquefaction potential assessed by neural networks | |
| Mahmoodzadeh et al. | Optimized machine learning modelling for predicting the construction cost and duration of tunnelling projects | |
| Jin et al. | Improved radial movement optimization to determine the critical failure surface for slope stability analysis | |
| Wang et al. | Direct interval prediction of landslide displacements using least squares support vector machines | |
| CN115952590A (en) | Shield tunnel optimization design method and device based on BO-RF-MOMPA hybrid algorithm | |
| CN115293316A (en) | Prediction method for deep-buried thick coal seam water flowing fractured zone based on SSA-ELMAN | |
| Wang et al. | Estimation of parameters of probability integral method model based on improved fireworks algorithm | |
| Mamat et al. | Artificial neural networks in slope of road embankment stability applications: a review and future perspectives | |
| Zhou et al. | Microseismic location in hardrock metal mines by machine learning models based on hyperparameter optimization using bayesian optimizer | |
| Zhou et al. | Hybrid random forest-based models for predicting shear strength of structural surfaces based on surface morphology parameters and metaheuristic algorithms | |
| Zhang et al. | Time series prediction of chimney foundation settlement by neural networks | |
| Shen et al. | Uncertainty analysis method of slope safety factor based on quantile-based ensemble learning | |
| CN117786794A (en) | Shield tunneling existing tunnel deformation optimization control method and system | |
| Sepahvand et al. | Evaluation of deep machine learning-based models of soil cumulative infiltration | |
| CN116861347A (en) | Magnetic force abnormal data calculation method based on deep learning model |
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 |