CN109978226B - Shield construction ground settlement prediction method based on cyclic neural network - Google Patents
Shield construction ground settlement prediction method based on cyclic neural network Download PDFInfo
- Publication number
- CN109978226B CN109978226B CN201910068462.XA CN201910068462A CN109978226B CN 109978226 B CN109978226 B CN 109978226B CN 201910068462 A CN201910068462 A CN 201910068462A CN 109978226 B CN109978226 B CN 109978226B
- Authority
- CN
- China
- Prior art keywords
- layer
- neural network
- settlement
- shield
- model
- 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
- 238000000034 method Methods 0.000 title claims abstract description 48
- 238000010276 construction Methods 0.000 title claims abstract description 45
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 42
- 125000004122 cyclic group Chemical group 0.000 title description 4
- 238000012549 training Methods 0.000 claims abstract description 19
- 230000000306 recurrent effect Effects 0.000 claims description 30
- 230000008569 process Effects 0.000 claims description 15
- 238000005259 measurement Methods 0.000 claims description 9
- 238000012360 testing method Methods 0.000 claims description 8
- 210000002569 neuron Anatomy 0.000 claims description 7
- 230000005641 tunneling Effects 0.000 claims description 5
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims description 5
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 claims description 4
- 230000000694 effects Effects 0.000 claims description 4
- 238000005457 optimization Methods 0.000 claims description 4
- 230000004913 activation Effects 0.000 claims description 2
- 239000000284 extract Substances 0.000 claims 1
- 230000001191 orthodromic effect Effects 0.000 claims 1
- 230000000644 propagated effect Effects 0.000 claims 1
- 230000002123 temporal effect Effects 0.000 claims 1
- 238000009412 basement excavation Methods 0.000 abstract description 7
- 239000011435 rock Substances 0.000 description 6
- 238000004364 calculation method Methods 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 3
- 230000006870 function Effects 0.000 description 3
- 238000007596 consolidation process Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000001537 neural effect Effects 0.000 description 2
- 238000007781 pre-processing Methods 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 239000002689 soil Substances 0.000 description 2
- 238000012795 verification Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000005520 cutting process Methods 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 238000013480 data collection Methods 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000009977 dual effect Effects 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000001125 extrusion Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000004806 packaging method and process Methods 0.000 description 1
- 239000011178 precast concrete Substances 0.000 description 1
- 238000003825 pressing Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000004062 sedimentation Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/08—Construction
-
- 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
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A10/00—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE at coastal zones; at river basins
- Y02A10/40—Controlling or monitoring, e.g. of flood or hurricane; Forecasting, e.g. risk assessment or mapping
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Economics (AREA)
- General Health & Medical Sciences (AREA)
- Human Resources & Organizations (AREA)
- Strategic Management (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Tourism & Hospitality (AREA)
- Software Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Marketing (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- General Business, Economics & Management (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Primary Health Care (AREA)
- Development Economics (AREA)
- Game Theory and Decision Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Excavating Of Shafts Or Tunnels (AREA)
Abstract
The invention relates to a method for predicting ground settlement in shield construction based on a circulating neural network, which selects main factors causing ground settlement in shield construction, fuses the settlement conditions at the current moment and uses the factors as input data together, and the ground settlement is effectively predicted through a deep neural network based on the circulating neural network established in the invention. The invention provides a detailed model establishing and training method, firstly, the settlement condition data is transmitted through a time channel of a circulating neural network, then main influencing factors are organically combined with the settlement condition data, and the characteristics are adaptively extracted through a deep network layer, and finally, a settlement prediction result is obtained. Therefore, the method has high prediction precision and strong generalization capability, and can be put into practical application of shield tunnel excavation ground settlement prediction.
Description
Technical Field
The invention relates to the field of intelligent monitoring of shield construction, in particular to a method for predicting shield construction ground settlement based on a recurrent neural network.
Background
The shield method is a fully mechanical construction method in the construction of the subsurface excavation method, which is a mechanical construction method for pushing a shield machine in the ground, preventing collapse into a tunnel by using a shield shell and duct pieces to support surrounding rocks around, excavating a soil body in front of an excavation surface by using a cutting device, transporting out of the tunnel by using an unearthing machine, pressing and jacking at the rear part by using a jack, and assembling precast concrete duct pieces to form a tunnel structure.
The surface subsidence refers to the surface subsidence phenomenon caused by extrusion of surrounding rocks on an excavated surface due to excavation, or relaxation of the surrounding rocks due to collapse, consolidation of gaps between the excavated surrounding rocks and supports, and between the surrounding rocks and lining due to the reduction of underground water level, subsidence of the supports in weak surrounding rocks, and the like.
The ground subsidence caused by shield propulsion includes five stages: initial settlement, settlement in front of the excavation face, settlement of the shield machine during passing, settlement of the shield tail gap and final consolidation settlement. In the shield method construction process, the soil body around the construction tunnel is inevitably loosened and sunk, so that the stratum is lost, and the ground surface is sunk. When the settlement exceeds a certain range, the stability of buildings along the line can be influenced, and more seriously, the tunnel can collapse, so that the serious result which cannot be estimated is caused. Therefore, the development and research of the ground settlement real-time prediction technology in the shield tunnel excavation process are very meaningful.
The invention CN107239599A of China discloses a prediction method of ground surface subsidence caused by shield construction based on a neural fuzzy inference system, which selects main influence factors of the ground surface subsidence caused by the shield construction, forms the main influence factors into an effective data set, and calculates the effective data set by using a ground surface subsidence prediction model built by the neural fuzzy inference system, thereby obtaining a prediction result and carrying out model verification according to the prediction result. But the invention only considers the influence factors and does not consider the influence of the generated surface subsidence on the change trend of the subsequent surface subsidence.
The invention CN107092990A of China discloses a shield construction ground settlement prediction system and method based on big data analysis, and a big data platform module is adopted to provide mass data storage and parallel computing service for modules such as data collection, data preprocessing, feature extraction, prediction model establishment, model application and the like, so as to form a ground settlement prediction system. The prediction method comprises the following steps: building a big data platform; collecting shield construction ground settlement data; preprocessing data; extracting characteristics; establishing a ground settlement prediction model; packaging a functional interface of the prediction model; and predicting the ground settlement in the construction process. But the invention only proposes the steps of building a prediction system and does not delve into the algorithms of the specific implementation.
Disclosure of Invention
The invention aims to provide a method for predicting the ground settlement of shield construction based on a recurrent neural network, which has high practicability by predicting the ground settlement based on the recurrent neural network technology through real-time data and construction parameters generated in the shield construction process so as to quickly process a large amount of data and obtain a prediction result.
In order to achieve the above object, the present invention provides the following technical solutions:
a shield construction ground settlement prediction method based on a recurrent neural network is summarized as follows:
(1) Collecting real data and establishing a data set;
(2) Establishing a shield construction ground settlement model based on a recurrent neural network;
(3) Training the model to obtain all hyper-parameters;
(4) And verifying the model and putting the model into practical use.
In the actual construction process, the ground settlement has many influence factors, and the tunnel position, the geological condition and the shield parameters can influence the ground settlement to a great extent, which is a result of the comprehensive effect of various complex factors. The recurrent neural network is one of algorithms which are prominent in the deep learning technology, has very strong nonlinear fitting capacity, and has incomparable processing capacity to time series data compared with other models. Moreover, the training process of the model is the determining process of the hyperparameter, and through the learning of a large amount of real data, each hyperparameter of the model can be continuously optimized, so that a more accurate ground settlement predicted value is given.
Further, the overall scheme of the present invention is specifically described as follows:
(1) And collecting real data and establishing a data set.
The method is a basis of a training model of a large amount of complete real data, and the shield construction ground settlement prediction method based on the recurrent neural network, which is provided by the method, not only collects settlement values changing along with time, but also comprises tunnel positions, geological conditions and shield parameters of each time point. The tunnel position comprises the tunnel depth and the tunneling distance; the geological conditions comprise tunnel top geology, tunnel bottom geology and underground water level; the shield parameters comprise working face pressure, propulsion speed, pitch angle, shield tail grouting pressure and shield tail grouting filling rate. After the data are collected, the data set is divided according to the requirements input by the model, and each sample data comprises ten items of attributes, and also comprises ten latest settlement measurement data at the current moment, which are used as input data of the model together.
(2) And establishing a shield construction ground settlement model based on a circulating neural network.
The model has five layers, and the output of each layer is the input of the next layer. The first layer is a recurrent neural network layer, which inputs the ten-time settlement measurement data in the sample data in the step (1) and obtains the result after the time axis effect; the second layer realizes the dual functions of the direct connection layer and the input layer, combines the output of the first layer with ten attribute values in sample data after output processing, and takes the combination as the output of the layer; the third layer to the fifth layer are directly connected layers, the number of nodes is 20, 10 and 3 respectively, the characteristic quantities are gradually compressed through the three layers until the output of the last fifth layer is the predicted value of the next three ground settlement values.
(3) And training the model to obtain all hyper-parameters.
Training the shield construction ground settlement model based on the recurrent neural network in the step (2), wherein a back propagation algorithm based on time and a back propagation algorithm are required; the training process is a process of continuously updating the hyper-parameters to obtain higher prediction precision, and comprises the following steps:
(1) inputting a sample data, performing forward propagation once according to a network established by a shield construction ground settlement model based on a recurrent neural network, and calculating to obtain an output value;
(2) the method comprises the steps that network back propagation is established once according to a shield construction ground settlement model based on a cyclic neural network, wherein a direct connection layer uses a back propagation algorithm, and a first layer of the cyclic neural network layer uses a time-based back propagation algorithm to calculate gradient values of all hyper-parameters;
(3) updating the values of the hyper-parameters, wherein an Adam optimization algorithm is adopted in the updating process;
(4) and (4) judging whether the model precision meets the requirement, if so, finishing training, and otherwise, returning to the step (1).
(4) And (5) verifying the model and putting the model into practical use.
After training, the model is tested with the partitioned test set to determine whether the resulting model still performs satisfactorily on the training set. After the model is confirmed to have certain generalization, the model can be put into an actual shield construction ground settlement prediction task.
Compared with the prior art, the invention has the beneficial effects that:
the invention designs a shield construction ground settlement prediction method based on a recurrent neural network, which selects a tunnel position, geological conditions and shield parameters in shield tunnel excavation and a settlement measurement value up to the current moment as input data, automatically learns characteristics through the structures of a recurrent layer and a plurality of direct connection layers, continuously optimizes model parameters, and finally obtains a ground settlement prediction model which can be used in actual engineering after verification of a test set. Compared with the traditional machine learning method, the method is based on larger-scale real data, integrates important influence parameters in shield construction, and is higher in prediction precision and higher in generalization capability.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, and it is obvious that the drawings in the following description are the embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to these drawings without creative efforts.
Fig. 1 is a flowchart of a shield construction ground settlement prediction method based on a recurrent neural network according to an embodiment of the present invention;
fig. 2 is a schematic diagram of a shield construction ground settlement model based on a recurrent neural network according to an embodiment of the present invention.
Detailed Description
In order to make the technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention.
As shown in fig. 1, a method for predicting shield construction ground settlement based on a recurrent neural network includes the following steps:
(1) And collecting real data and establishing a data set.
Selecting key factors causing ground settlement in construction and actual settlement values to jointly form a more effective data set, wherein the key factors comprise tunnel depth, tunneling distance, tunnel top geology, tunnel bottom geology, underground water level, working face pressure, propulsion speed, pitch angle, shield tail grouting pressure and shield tail grouting filling rate, and the actual settlement values are measured values of the latest ten times of settlement at the current time point.
Suppose that at time t, the last ten settlement measurements by the current time are used separatelyThe ten key factors influencing the ground settlement are respectively shown asShowing that three sedimentation measurements after time t are respectively usedThat means, then the sample set X and label set Y containing n samples is equal to:
(2) And establishing a shield construction ground settlement model based on a recurrent neural network.
As shown in fig. 2, the model has five layers, and the output of each layer is the input of the next layer. The first layer is a circulating neural network structure, the input is a settlement measurement value for nearly ten times, the output is a result transmitted by a time shaft, and the calculation formula is as follows:
wherein,
o 10 =0
in the formula,
o 1i -the output value of the ith neuron of the first layer;
x i -input values for the ith neuron;
w 11 -weights along the time channels;
w 12 -the weight of the first layer network.
The second layer network structure applies the output of the first layer circular neural network structure to a ReLU activation function, and simultaneously combines ten key factors influencing ground settlement as the output of the layer, namely:
in the formula,
o 2i -the output value of the ith neuron of the second layer;
w 2 -weights of the layer two network.
The third layer to the fifth layer are direct connection layers, main features are gradually extracted through the three layers of networks, the number of neurons is reduced, and finally the predicted value of the subsequent three times of ground subsidence is obtained. The three layers are calculated as follows:
in the formula,
o 3 -the output value of the third layer;
o 4 -an output value of the fourth layer;
o 5 -an output value of the fifth layer;
w 3 -weights for layer three networks;
b 3 -bias of layer three network.
Wherein,
in the formula,
x ij -the ith row and the jth column element of the matrix x.
(3) And training the model to obtain all hyper-parameters.
(1) After obtaining the output value o of forward propagation, evaluating the prediction effect of the model by using a mean square error loss function, wherein the calculation formula is as follows:
in the formula,
l-mean square error loss value.
(2) And respectively calculating the gradient size of each layer of hyper-parameters by using a back propagation algorithm and a time-based back propagation algorithm. The back propagation method is applied to the direct connection layer, and the time-based back propagation algorithm is applied to the recurrent neural network layer. The gradient value of each parameter can be obtained by the direct connection layer through a chain ruleAndthe gradient calculation formula of the circulating layer is as follows:
the first layer gradient is:
wherein,
(3) after obtaining all gradient values of each network layer, parameter updating is carried out by using an Adam optimization algorithm, wherein a direct connection layer needs to be updated and has a weight value and a bias item, a recurrent neural network layer needs to be updated and is the weight value, and the calculation process of the Adam optimization algorithm is as follows:
V dw =β 1 V dw +(1-β 1 )dw,V db =β 1 V db +(1-β 1 )db
S dw =β 2 S dw +(1-β 2 )(dw) 2 ,S db =β 2 S db +(1-β 2 )(db) 2
wherein,
β 1 =0.9
β 2 =0.999
ε=10 -8
in the formula,
dw — gradient value of weight w;
db — the gradient value of the offset b;
V dw ,S dw -an exponentially weighted average of dw;
V db ,S db -an exponentially weighted average of db;
thus, a complete forward propagation, backward propagation and parameter updating process is completed, and then the next set of training data is used for continuously iterating to update all hyper-parameters until one iteration of all training data is completed. The training data is about 90% of the total number of samples, and the remaining 10% is used as a test set to test the model.
(4) And after the training data is used for repeatedly updating the parameters until the preset precision requirement is met, stopping iteration, wherein the obtained parameters are the model parameters after the training is finished.
(4) And (5) verifying the model and putting the model into practical use.
And then testing the model by using a test set, wherein parameters are not updated in the testing process, and only whether the precision of the model meets the requirements is judged, so that the model has good generalization, and can be put into the actual shield tunneling engineering for predicting the ground settlement.
The above description is only illustrative of the preferred embodiments of the present invention and should not be taken as limiting the scope of the invention in any way. Any changes or modifications made by those skilled in the art based on the above disclosure should be considered as equivalent effective embodiments, and all the changes or modifications should fall within the protection scope of the technical solution of the present invention.
Claims (8)
1. A shield construction ground settlement prediction method based on a recurrent neural network is characterized in that comprehensive consideration of tunnel position, geological conditions, shield parameters and settlement conditions is adopted, characteristics are extracted in a self-adaptive mode by utilizing a deep network model, and prediction of ground settlement is achieved, and the method comprises the following processes:
(1) Collecting real data and establishing a data set;
(2) Establishing a shield construction ground settlement model based on a recurrent neural network;
(3) Training the model to obtain all hyper-parameters;
(4) Verifying the model and putting the model into practical use;
the shield construction ground settlement model based on the recurrent neural network in the step (2) comprises the following steps:
the first layer is a circulating neural network layer, inputs the latest ten-time settlement measurement data at the current moment, and extracts the characteristic value of the settlement measurement data through the propagation of a time channel;
the second layer is used for combining the output of the first layer with ten attribute values of the tunnel depth, the tunneling distance, the tunnel top geology, the tunnel bottom geology, the underground water level, the working face pressure, the propulsion speed, the pitch angle, the shield tail grouting pressure and the shield tail grouting filling rate to be used as the output of the layer;
the third layer is a direct connection layer, the number of nodes is 20, and the feature values which are completely input after combination are extracted;
the fourth layer is a direct connection layer, the number of nodes is 10, and the characteristic quantity is compressed;
and the fifth layer is a direct connection layer, the number of the nodes is 3, and the characteristic quantity is further compressed to obtain the required ground settlement predicted value.
2. The circular neural network-based shield construction ground settlement prediction method of claim 1, wherein: the data set in the step (1) starts from main factors influencing ground settlement, including settlement conditions changing along with time, and tunnel positions, geological conditions and shield parameters of each time point;
the tunnel position comprises the depth and the tunneling distance of the tunnel;
the geological conditions comprise tunnel top geology, tunnel bottom geology and underground water level;
the shield parameters comprise working face pressure, propulsion speed, pitch angle, shield tail grouting pressure and shield tail grouting filling rate;
the settlement condition comprises the latest ten settlement measurement data at the current moment.
3. The method for predicting the ground subsidence in shield construction based on the recurrent neural network of claim 1, wherein the recurrent neural network layer of the first layer is referred to the data of the processed subsidence condition and propagated along the time path, and the function of the method is as follows:
in the formula,
o 1i is the output value of the ith neuron of the first layer,
x i -input values for the ith neuron;
w 11 -weights along the temporal paths;
w 12 -the weight of the first layer network.
4. The method for predicting the ground subsidence in the shield construction based on the recurrent neural network as claimed in claim 1, wherein the subsidence data processed by the recurrent neural network layer in the second layer is combined with ten attributes of main influencing factors of the ground subsidence to form new input data, and the function of the new input data is as follows:
in the formula,
o 2i -the output value of the ith neuron of the second layer;
w 2 -weight of the layer two network;
x i -ithgodAn input value via the primitive.
5. The method for predicting the ground subsidence in the shield construction based on the recurrent neural network as claimed in claim 1, wherein the three last-layer network structures are all straight-connected layers, the number of features is gradually reduced, and the predicted value of the ground subsidence of the last three times is obtained, and the function is as follows:
in the formula,
o 3 -an output value of the third layer;
o 4 -an output value of the fourth layer;
o 5 -an output value of the fifth layer;
w 3 -weights of the layer three network;
w 4 -weight of the layer four network;
w 5 -weight of fifth layer network;
b 3 -bias of layer three network;
b 4 -biasing of the fourth layer network;
b 5 -biasing of the fifth layer network.
6. The method for predicting the ground settlement in the shield construction based on the recurrent neural network as claimed in claim 1, wherein during model training, a mean square error loss function is used to judge the prediction effect, a back propagation algorithm is used for the orthodromic layer, a time-based back propagation algorithm is used for the recurrent neural network layer, and the mean square error loss function is:
in the formula,
l-mean square error loss value.
7. The method for predicting the ground subsidence of the shield construction based on the recurrent neural network as claimed in claim 6, wherein Adam optimization algorithm is adopted during parameter updating, and the parameter β is 1 =0.9,β 2 =0.999,ε=10 -8 。
8. The method for predicting the ground subsidence in the shield construction based on the recurrent neural network as claimed in claim 7, wherein the model obtained is tested by a test set after the training of the model is completed, and the model is put into practical engineering for use after the model is determined to have good generalization.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910068462.XA CN109978226B (en) | 2019-01-24 | 2019-01-24 | Shield construction ground settlement prediction method based on cyclic neural network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910068462.XA CN109978226B (en) | 2019-01-24 | 2019-01-24 | Shield construction ground settlement prediction method based on cyclic neural network |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109978226A CN109978226A (en) | 2019-07-05 |
CN109978226B true CN109978226B (en) | 2023-02-28 |
Family
ID=67076686
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910068462.XA Active CN109978226B (en) | 2019-01-24 | 2019-01-24 | Shield construction ground settlement prediction method based on cyclic neural network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109978226B (en) |
Families Citing this family (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110617074B (en) * | 2019-09-20 | 2020-10-09 | 西安电子科技大学 | Incidence relation method for ground settlement and tunneling parameters in shield construction |
CN111119902B (en) * | 2019-12-16 | 2021-04-06 | 北京科技大学 | Tunnel dynamic construction method based on BP neural network |
CN111460737B (en) * | 2020-04-09 | 2023-12-29 | 昆山阳翎机器人科技有限公司 | Intelligent settlement prediction method and system for slurry air pressure balance shield |
CN111832223A (en) * | 2020-06-29 | 2020-10-27 | 上海隧道工程有限公司 | Neural network-based shield construction surface subsidence prediction method |
CN112364422B (en) * | 2020-11-13 | 2023-06-13 | 中铁二十局集团有限公司 | MIC-LSTM-based dynamic prediction method for shield construction earth surface deformation |
CN113128106B (en) * | 2021-04-06 | 2023-03-21 | 汕头大学 | Method for determining surface subsidence caused by shield construction of karst stratum |
CN113204824B (en) * | 2021-05-21 | 2023-04-07 | 上海大学 | Multi-model fusion shield construction settlement prediction method and system |
CN113239439B (en) * | 2021-05-21 | 2022-04-05 | 上海大学 | Shield construction ground surface settlement prediction system and method |
CN113298220B (en) * | 2021-05-31 | 2023-08-04 | 中铁十六局集团北京轨道交通工程建设有限公司 | Neural network optimization-based shield tunneling speed prediction method |
CN113486818B (en) * | 2021-07-09 | 2022-05-20 | 吉林大学 | Full fighting rate prediction system and method based on machine vision |
CN113931636B (en) * | 2021-10-22 | 2024-05-07 | 广州地铁设计研究院股份有限公司 | Tunnel underpass existing operation subway line grouting construction parameter selection method and application thereof |
CN115130734A (en) * | 2022-06-06 | 2022-09-30 | 北京城建设计发展集团股份有限公司 | Method and system for predicting construction influence of penetration project based on LightGBM and deep learning algorithm |
CN118070567B (en) * | 2024-04-18 | 2024-07-23 | 安徽省交通控股集团有限公司 | Shield speed control system and method based on quantitative simulation of tunnel construction environment |
Family Cites Families (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107239599A (en) * | 2017-05-16 | 2017-10-10 | 五邑大学 | Based on Ground surface settlement method caused by the shield-tunneling construction of neural fuzzy inference system |
CN109242171A (en) * | 2018-08-28 | 2019-01-18 | 河南省豫晋高速公路建设有限公司 | A kind of shield-tunneling construction Ground surface settlement method based on BIM and RS-SVR |
-
2019
- 2019-01-24 CN CN201910068462.XA patent/CN109978226B/en active Active
Also Published As
Publication number | Publication date |
---|---|
CN109978226A (en) | 2019-07-05 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109978226B (en) | Shield construction ground settlement prediction method based on cyclic neural network | |
Li et al. | Evolution characteristics and displacement forecasting model of landslides with stair-step sliding surface along the Xiangxi River, three Gorges Reservoir region, China | |
US20230144184A1 (en) | Advanced geological prediction method and system based on perception while drilling | |
Yang et al. | Improved PLS and PSO methods-based back analysis for elastic modulus of dam | |
Feng et al. | Estimating mechanical rock mass parameters relating to the Three Gorges Project permanent shiplock using an intelligent displacement back analysis method | |
Mohanty et al. | Artificial neural network modeling for groundwater level forecasting in a river island of eastern India | |
Zhang et al. | Improved coupled Markov chain method for simulating geological uncertainty | |
Reddy | An empirical study on the best prediction model of water inflow into drill and blast tunnels among several machine learning techniques | |
Jan et al. | Neural network forecast model in deep excavation | |
KR101967978B1 (en) | Apparatus for predicting net penetration rate of shield tunnel boring machine and method thereof | |
CN113204824B (en) | Multi-model fusion shield construction settlement prediction method and system | |
Fattahi et al. | Hybrid Monte Carlo simulation and ANFIS-subtractive clustering method for reliability analysis of the excavation damaged zone in underground spaces | |
CN111967079A (en) | Foundation pit deformation prediction method based on improved artificial bee colony algorithm and BP neural network | |
Chen et al. | Application of group decision-making AHP of confidence index and cloud model for rock slope stability evaluation | |
CN112765791B (en) | TBM card-sticking risk prediction method based on numerical value sample and random forest | |
CN105678417A (en) | Prediction method and device for tunnel face water inflow of construction tunnel | |
CN113255990A (en) | Real-time prediction system and method for soil quality of excavation surface in tunnel construction by shield method | |
CN115481565A (en) | Earth pressure balance shield tunneling parameter prediction method based on LSTM and ant colony algorithm | |
Qiu et al. | TBM tunnel surrounding rock classification method and real-time identification model based on tunneling performance | |
Li et al. | Research and application of deformation prediction model for deep foundation pit based on LsTM | |
CN115293316A (en) | Prediction method for deep-buried thick coal seam water flowing fractured zone based on SSA-ELMAN | |
CN105160111A (en) | Simulation method for loess tunnel numerical model by considering surrounding rock structural characteristics | |
CN118094694A (en) | Training method and device for large-diameter shield tunneling earth surface subsidence intelligent prediction model | |
CN114564886A (en) | Shield tunneling parameter prediction method based on geological parameter quantification | |
CN114091339A (en) | Method for predicting bow blowing instantaneous yield of drag suction dredger based on GRU |
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 | ||
GR01 | Patent grant | ||
GR01 | Patent grant |