CN112836789A - Ground connection wall deformation dynamic prediction method based on composite neural network algorithm - Google Patents

Ground connection wall deformation dynamic prediction method based on composite neural network algorithm Download PDF

Info

Publication number
CN112836789A
CN112836789A CN202011572061.7A CN202011572061A CN112836789A CN 112836789 A CN112836789 A CN 112836789A CN 202011572061 A CN202011572061 A CN 202011572061A CN 112836789 A CN112836789 A CN 112836789A
Authority
CN
China
Prior art keywords
deformation
prediction
neural network
diaphragm wall
monitoring
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
Application number
CN202011572061.7A
Other languages
Chinese (zh)
Inventor
刘维
赵华菁
史培新
管浩
唐强
付春青
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Suzhou University
Original Assignee
Suzhou University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Suzhou University filed Critical Suzhou University
Priority to CN202011572061.7A priority Critical patent/CN112836789A/en
Publication of CN112836789A publication Critical patent/CN112836789A/en
Pending legal-status Critical Current

Links

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
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/044—Recurrent networks, e.g. Hopfield 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
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
    • 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
    • G06N3/06—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons
    • 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

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • General Physics & Mathematics (AREA)
  • General Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Software Systems (AREA)
  • Evolutionary Computation (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Mathematical Physics (AREA)
  • Computational Linguistics (AREA)
  • Artificial Intelligence (AREA)
  • Business, Economics & Management (AREA)
  • Economics (AREA)
  • Human Resources & Organizations (AREA)
  • Strategic Management (AREA)
  • General Business, Economics & Management (AREA)
  • Marketing (AREA)
  • Tourism & Hospitality (AREA)
  • Game Theory and Decision Science (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Operations Research (AREA)
  • Quality & Reliability (AREA)
  • Development Economics (AREA)
  • Primary Health Care (AREA)
  • Neurology (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention discloses a ground connection wall deformation dynamic prediction method based on a composite neural network algorithm. Collecting horizontal displacement of the diaphragm wall by using a movable inclinometer, preprocessing monitoring data and forming a deformation database; dividing a deformation database into a training set and a testing set by adopting a CV-LSTM model, and learning training samples in the training set; and predicting the horizontal displacement of the diaphragm wall by using the test sample in the deformation database, and providing a basis for controlling the deformation risk. Compared with the traditional BP neural network, the CV-LSTM composite neural network algorithm prediction model provided by the invention has higher prediction precision, has better generalization capability than an independent LSTM deep network, is suitable for the dynamic prediction problem of diaphragm wall deformation, and can provide reference for realizing informatization management of a construction site.

Description

Ground connection wall deformation dynamic prediction method based on composite neural network algorithm
Technical Field
The invention relates to a diaphragm wall deformation prediction method, in particular to a diaphragm wall deformation dynamic prediction method based on a composite neural network.
Background
At present, China is in the rapid development stage of rail transit construction. In 2019, the number of newly increased stations in the country reaches 238, and in addition, rail transit projects of 59 lines in 12 regions are approved to be built, and the total project investment amount is about 9700 million yuan. Therefore, a large amount of station foundation pit engineering is generated, the excavation scale and the excavation depth of the foundation pit are continuously increased, for example, the maximum excavation depth of a certain station of a certain city subway line reaches 29 m. The foundation pit engineering is a space system comprising a soil body and a supporting structure, and is influenced by a plurality of internal and external factors such as geological conditions, construction quality, surrounding environment and the like. The field actual measurement deformation data is the embodiment of various influence comprehensive effects in the construction process, and the analysis and research of the field deformation monitoring data becomes an effective way for people to know the deformation characteristics of the foundation pit.
The prediction method mainly comprises three methods, namely an empirical method, a numerical simulation method and a machine learning method. The empirical method needs a large amount of monitoring data, is limited to establishing a discrete random model by using a differential equation, and is inconvenient for describing the essence and the internal law of the system change process; the numerical method has accuracy on the mathematical method, but due to the complexity of the deformation influence factors of the underground diaphragm wall of the foundation pit, the fuzziness of a physical mechanism and the changeability and uncertainty of parameters, the method is excessively generalized when in use, and the practical value is reduced. At present, most of machine learning methods adopt a classical BP neural network, the structure of the back propagation neural network is relatively simple, but the input time relevance cannot be accurately processed, and the error is large when long sequence data is processed. The other LSTM algorithm for processing sequence data has the advantages of convenience for sequence modeling, long-term memory capacity and poor generalization capacity.
Disclosure of Invention
Aiming at the defects of the existing machine learning method for predicting the deformation of the diaphragm wall, the invention provides the diaphragm wall deformation prediction method based on the composite neural network, which effectively improves the stability and the generalization of the prediction result, has higher precision in long-term prediction and realizes dynamic deformation monitoring of the foundation pit engineering construction site for diaphragm wall deformation.
In order to achieve the above object, the technical solution of the present invention is to provide a method for dynamically predicting wall-to-ground deformation based on a composite neural network algorithm, comprising the following steps:
(1) preprocessing deformation monitoring data of the foundation pit engineering diaphragm wall, recording the data as a monitoring daily report, and forming a deformation database;
(2) constructing a prediction model of a CV and LSTM composite neural network algorithm, dividing a deformation database into a training set and a test set by adopting a CV cross verification method, and learning a training set sample by adopting an LSTM neural network in a deep learning PyTorch frame package, wherein the set super parameters comprise the number of training rounds, the learning rate and the number of neurons in a hidden layer;
(3) and (3) using the optimal model obtained by training for underground continuous wall deformation prediction, comparing the obtained prediction value with the allowable deformation limit value of the underground continuous wall, and outputting a result.
The deformation history monitoring data in the step (1) comprises the following steps: monitoring the horizontal displacement of the diaphragm wall by a movable inclinometer buried in the concrete of the wall body, wherein the monitoring frequency is 1-3 times/day, and supplementing missing data by linear interpolation; the monitoring data preprocessing comprises the following steps: storing the monitoring daily newspaper in an Excel format, or directly transmitting data acquired by the inclinometer to a computer through a transmission cable for data storage; the deformation database includes: recording the deformation monitoring value acquired by each measuring point as
Figure RE-RE-DEST_PATH_IMAGE001
To indicate the measured pointiIn the first placetAnd (4) forming a deformation database with time series characteristics according to the deformation values of the days.
In the step (2), the prediction model reads data in the ground wall monitoring daily report by using an xlrd module in a PyTorch frame package, and stores the data into a tensor structure by using a tensor function; the input layer of the prediction model is
Figure RE-DEST_PATH_IMAGE002
The output layer of the prediction model is
Figure RE-RE-DEST_PATH_IMAGE003
Wherein N is the input information length, and M is the prediction time span; the prediction model divides a data set into K subsets by using K-CV, wherein K-1 subsets are taken as a training set in turn, and the remaining 1 subset is taken as a test set; the prediction model adopts an LSTM neural network to learn the training set samples, the set hyper-parameters comprise the number of training rounds, the learning rate and the number of neurons in a hidden layer, and the forward propagation formulas are respectively as follows:
the first module is 'forget gate' and is used for calculating forget proportion of neuron state information at a last moment:
Figure RE-RE-DEST_PATH_IMAGE005
the second module is the "input gate" for new information written into the ratio of neuron states:
Figure RE-RE-DEST_PATH_IMAGE007
Figure RE-RE-DEST_PATH_IMAGE009
Figure RE-RE-DEST_PATH_IMAGE011
the third module, the "output gate", will determine the information to be output as a hidden state:
Figure RE-RE-DEST_PATH_IMAGE013
Figure RE-RE-DEST_PATH_IMAGE015
wherein the content of the first and second substances,C t andh t respectively representtA temporal neuron state and a hidden state;f t ，i t and o t Respectively representtA forgetting gate, an input gate and an output gate at a moment;W，brespectively are a weight matrix and an offset vector in each gating module;srepresenting a sigmoid activation function; tanh represents a hyperbolic tangent activation function;
the prediction model compares the deformation prediction value with the deformation measured value, and calculates the precision evaluation index MSE of the deformation prediction value as an optimization target function; and (5) iteratively updating the parameters by adopting an Adam algorithm, and adjusting to obtain an optimal solution.
Performing deformation prediction on the test set sample by adopting an optimal model in K-CV in the step (3) to obtain a deformation prediction value of the diaphragm wall; and storing the prediction result by using an xlwr module.
The invention provides a ground connection wall deformation dynamic prediction method based on a composite neural network algorithm, which is characterized in that after new monitoring data are collected, the step (1) and the step (2) are repeated, and a CV-LSTM model is retrained for updating a deformation prediction result.
By adopting the ground connection wall deformation dynamic prediction method based on the composite neural network algorithm, provided by the technical scheme of the invention, when the output result is that the deformation prediction value is close to the warning value, the early warning information is pushed and corresponding deformation control measures are taken in time.
Compared with the prior art, the invention has the following advantages:
1. the invention aims at the characteristics of long period and nonlinearity of diaphragm wall deformation monitoring data, and overcomes the defects of overfitting and gradient explosion of the traditional neural network by adopting a CV-LSTM model prediction model.
2. The CV-LSTM model adopted by the invention has the advantage of processing long sequence data, and the accuracy of the long-term prediction of the deformation of the diaphragm wall can be effectively improved.
3. Compared with a single neural network model, the CV-LSTM model adopted by the invention has stronger stability and better generalization.
Drawings
FIG. 1 is a schematic diagram of a foundation pit plane and an arrangement of an inclinometer according to an embodiment of the invention;
FIG. 2 is a schematic cross-sectional design view of a standard section of a foundation pit according to an embodiment of the invention;
fig. 3 is a flowchart of deformation prediction according to an embodiment of the present invention.
In the figure, 1, filling soil with impurities; 2. powdery clay; 3. silt and silt are mixed in silt; 4. the silt is mixed with the powdery clay; 5. powdery clay; 6. the silt is mixed with silt.
Detailed Description
The invention is described in detail below with reference to the figures and specific embodiments.
Example 1
The invention selects the deep foundation pit engineering of a railway vehicle station in rail transit in a certain city as an example.
Referring to fig. 1, a schematic diagram of the arrangement of the plane of the foundation pit and the inclination measuring points in this embodiment is shown; the subway station foundation pit is in the east-west direction, the total length is about 103.0 m, the standard section width is 23.1 m, and the excavation depth is 24.16 m.
Referring to fig. 2, a schematic cross-sectional design diagram of a standard section of a foundation pit in this embodiment is shown; the foundation pit enclosure structure adopts an underground continuous wall with the thickness of 1.0 m, and the horizontal displacement of the underground continuous wall is monitored by a movable inclinometer buried in wall concrete. And (3) acquiring total 2329 groups of inclination measuring monitoring data from 8/8 days in 2018 to 7/9 days in 2019 of the monitoring points CX 10-CX 19 of the foundation pit B to establish a database.
Referring to fig. 3, a flowchart of the method for dynamically predicting deformation of the diaphragm wall based on the composite neural network provided in this embodiment includes steps of data preparation, model training, and model testing, and the flow is as follows: collecting horizontal displacement of the diaphragm wall by using a movable inclinometer, preprocessing monitoring data and forming a database; dividing a deformation database into a training set and a testing set by adopting a CV-LSTM model, and learning training samples in the training set; and predicting the horizontal displacement of the diaphragm wall by using the test sample in the deformation database, comparing the predicted value with the allowable deformation limit value of the diaphragm wall, and controlling the subsequent development of deformation by adjusting the construction technical parameters.
The method comprises the following steps:
and preprocessing the deformation monitoring data of the foundation pit engineering diaphragm wall, recording the data as a monitoring daily report, and completing missing data to form a deformation database.
S1.1, monitoring horizontal displacement of the diaphragm wall through a movable inclinometer buried in wall concrete. The monitoring frequency is 1 time per day in principle, the excavation depth is increased to 2 times per day after exceeding 20 m, 2-3 times per day when the deformation is abnormal, monitoring data are collated to form a monitoring daily report, and the monitoring daily report is stored in an Excel format; the data collected by the inclinometer can also be directly transmitted to the computer through a transmission cable for data storage;
s1.2, because the monitoring points shield the gland in the construction process, partial missing data is caused, and the linear interpolation is used for completion;
s1.3, recording the deformation monitoring value collected by each measuring point as
Figure RE-917401DEST_PATH_IMAGE001
To indicate the measured pointiIn the first placetAnd (4) forming a deformation database with time series characteristics according to the deformation values of the days.
Step two:
a prediction model of CV and LSTM neural network algorithms is constructed, a CV cross verification method is adopted to divide a deformation database into a training set and a test set, an LSTM neural network in a deep learning PyTorch frame package is adopted to learn samples in the training set, and set hyper-parameters comprise the number of training rounds, the learning rate and the number of neurons in a hidden layer.
S2.1, reading data in the ground wall monitoring daily report by using an xlrd module in a PyTorch frame package, and storing the data into a tensor structure by using a tensor function;
s2.2, the input layer of the prediction model is
Figure RE-DEST_PATH_IMAGE016
The output layer of the prediction model is
Figure RE-RE-DEST_PATH_IMAGE017
Wherein N is the length of the input information (3 days or 7 days or 15 days, etc.), M is the prediction time span (1 day or 3 days or 7 days, etc.);
s2.3, dividing the data set into K subsets by using K-CV, taking K-1 subsets as training sets in turn, and taking the remaining 1 subset as a test set;
s2.4, learning the training set sample by using an LSTM neural network, wherein the set hyper-parameters comprise training round number, learning rate and hidden layer neuron number, and forward propagation formulas are respectively as follows:
the first module is 'forget gate' and is used for calculating forget proportion of neuron state information at a last moment:
Figure RE-203849DEST_PATH_IMAGE005
the second module is the "input gate" for new information written into the ratio of neuron states:
Figure RE-186849DEST_PATH_IMAGE007
Figure RE-847637DEST_PATH_IMAGE009
Figure RE-662010DEST_PATH_IMAGE011
the third module, the "output gate", will determine the information to be output as a hidden state:
Figure RE-812368DEST_PATH_IMAGE013
Figure RE-344981DEST_PATH_IMAGE015
wherein the content of the first and second substances,C t andh t respectively representtA temporal neuron state and a hidden state;f t ，i t and o t Respectively representtA forgetting gate, an input gate and an output gate at a moment;W，brespectively are a weight matrix and an offset vector in each gating module;srepresenting a sigmoid activation function;tanhrepresenting a hyperbolic tangent activation function;
s2.5, comparing the deformation predicted value with the deformation measured value, calculating an accuracy evaluation index MSE of the deformation predicted value and the deformation measured value, and taking the accuracy evaluation index MSE as an optimization target function;
and S2.6, iteratively updating the network parameters by adopting an Adam algorithm, and adjusting to obtain an optimal solution.
Step three:
and using the optimal model obtained by training for predicting the deformation of the diaphragm wall, comparing the predicted value with the deformation limit value allowed by the diaphragm wall, and controlling the subsequent development of the deformation by adjusting the construction technical parameters.
S3.1, performing deformation prediction on the test set sample by using an optimal model in K-CV to obtain a deformation prediction value of the diaphragm wall;
s3.2, storing the prediction result by using an xlwr module;
s3.3, after new monitoring data are collected, repeating the first step and the second step, and retraining the CV-LSTM model to update the deformation prediction result
And S3.4, when the deformation prediction value is close to the warning value, pushing the early warning information and taking corresponding deformation control measures in time.
The specific implementation steps of this embodiment are as follows:
the method comprises the following steps:
the horizontal displacement of the diaphragm wall is monitored by a movable inclinometer buried in the concrete of the wall body. The monitoring frequency is 1 time per day in principle, the excavation depth is increased to 2 times per day after exceeding 20 m, 2-3 times per day when the deformation is abnormal, monitoring data are collated to form a monitoring daily report, and the monitoring daily report is stored in an Excel format; partial missing data is caused by the condition that a gland is shielded at a monitoring point in the construction process, and the linear interpolation is used for completion; recording the deformation monitoring value acquired by each measuring point as
Figure RE-747143DEST_PATH_IMAGE001
To indicate the measured pointiIn the first placetAnd (4) forming a deformation database with time series characteristics according to the deformation values of the days.
Step two:
and reading in data in the monitoring daily report by using an xlrd module in the PyTorch framework package, and storing the monitoring data into a tensor structure by using a tensor function. The input layer of the prediction model is
Figure RE-DEST_PATH_IMAGE018
The output layer of the prediction model is
Figure RE-9497DEST_PATH_IMAGE017
Where N inputs the information length and M represents the prediction time span. A CV cross verification method is adopted to divide a data set, a K value is 10, namely, a 10-fold cross verification method is adopted for CX 10-CX 19 slope measuring points in an original slope measuring deformation sample of the diaphragm wall, each measuring point is used as a test set and is predicted once, and therefore the generalization of a prediction model is relatively and objectively reflected. The LSTM neural network is adopted to learn the training set samples, three hyper-parameters of training round number, learning rate and hidden layer neuron number are set, and the Adam adaptive moment estimation algorithm is adopted to optimize because the LSTM network has a deeper structure and the parameters are difficult to update.
Step three:
and (3) performing deformation prediction on the test set sample by using the optimal model obtained by training, and calculating the precision evaluation index MSE of the test set sample.
In order to further verify the effectiveness of the CV-LSTM model, a classical BP neural network is selected for comparison, and 4 prediction tasks are designed to reflect the prediction effect of the prediction model under different input information lengths and prediction step lengths.
Task 1: predicting the deformation amount after 1 day according to the deformation monitoring values of the previous 3 days (N =3, M = 1);
task 2: predicting the deformation amount after 7 days according to the deformation monitoring value of the previous 3 days (N =3, M = 7);
task 3: predicting the deformation amount after 1 day according to the deformation monitoring value of the previous 15 days (N =15, M = 1);
and task 4: the deformation amount after 7 days (N =15, M = 7) was predicted from the deformation monitoring value of the previous 15 days.
Table 1 shows the MSE (unit/mm) values of the prediction errors of 10-fold cross validation2). As shown in Table 1, the CV-LSTM model performed well better than the BP model in all prediction tasks, resulting in smaller error values and better stability on the test set.
TABLE 110 prediction error MSE values (units/mm) for cross validation2）
Figure RE-DEST_PATH_IMAGE020
。

Claims (6)

1. A dynamic prediction method for deformation of a diaphragm wall based on a composite neural network algorithm is characterized by comprising the following steps:
(1) preprocessing deformation monitoring data of the foundation pit engineering diaphragm wall, recording the data as a monitoring daily report, and forming a deformation database;
(2) constructing a prediction model of a CV and LSTM composite neural network algorithm, dividing a deformation database into a training set and a test set by adopting a CV cross verification method, and learning a training set sample by adopting an LSTM neural network in a deep learning PyTorch frame package, wherein the set super parameters comprise the number of training rounds, the learning rate and the number of neurons in a hidden layer;
(3) and (3) using the optimal model obtained by training for underground continuous wall deformation prediction, comparing the obtained prediction value with the allowable deformation limit value of the underground continuous wall, and outputting a result.
2. The method for dynamically predicting the deformation of the diaphragm wall based on the composite neural network algorithm according to claim 1, wherein: the deformation history monitoring data in the step (1) comprises the following steps: monitoring the horizontal displacement of the diaphragm wall by a movable inclinometer buried in the concrete of the wall body, wherein the monitoring frequency is 1-3 times/day, and supplementing missing data by linear interpolation; the monitoring data preprocessing comprises the following steps: storing the monitoring daily newspaper in an Excel format, or directly transmitting data acquired by the inclinometer to a computer through a transmission cable for data storage; the deformation database includes: recording the deformation monitoring value acquired by each measuring point as
Figure RE-DEST_PATH_IMAGE001
To indicate the measured pointiIn the first placetAnd (4) forming a deformation database with time series characteristics according to the deformation values of the days.
3. The method for dynamically predicting the deformation of the diaphragm wall of the composite neural network algorithm according to claim 1, wherein the method comprises the following steps: in the step (2),
the prediction model reads data in the ground wall monitoring daily report by using an xlrd module in a PyTorch frame package, and stores the data into a tensor structure by using a tensor function;
the input layer of the prediction model is
Figure RE-RE-DEST_PATH_IMAGE002
The output layer of the prediction model is
Figure RE-DEST_PATH_IMAGE003
Wherein N is the input information length, and M is the prediction time span; the prediction model divides a data set into K subsets by using K-CV, wherein K-1 subsets are taken as a training set in turn, and the remaining 1 subset is taken as a test set;
the prediction model adopts an LSTM neural network to learn the training set samples, the set hyper-parameters comprise the number of training rounds, the learning rate and the number of neurons in a hidden layer, and the forward propagation formulas are respectively as follows:
the first module is 'forget gate' and is used for calculating forget proportion of neuron state information at a last moment:
Figure RE-DEST_PATH_IMAGE005
the second module is the "input gate" for new information written into the ratio of neuron states:
Figure RE-DEST_PATH_IMAGE007
Figure RE-DEST_PATH_IMAGE009
Figure RE-DEST_PATH_IMAGE011
the third module, the "output gate", will determine the information to be output as a hidden state:
Figure RE-DEST_PATH_IMAGE013
Figure RE-DEST_PATH_IMAGE015
wherein the content of the first and second substances,C t andh t respectively representtA temporal neuron state and a hidden state;f t ，i t and o t Respectively representtA forgetting gate, an input gate and an output gate at a moment;W，brespectively are a weight matrix and an offset vector in each gating module;srepresenting a sigmoid activation function; tanh represents a hyperbolic tangent activation function;
the prediction model compares the deformation prediction value with the deformation measured value, and calculates the precision evaluation index MSE of the deformation prediction value as an optimization target function; and (5) iteratively updating the parameters by adopting an Adam algorithm, and adjusting to obtain an optimal solution.
4. The method for dynamically predicting the deformation of the diaphragm wall based on the composite neural network algorithm according to claim 1, wherein: performing deformation prediction on the test set sample by adopting an optimal model in K-CV in the step (3) to obtain a deformation prediction value of the diaphragm wall; and storing the prediction result by using an xlwr module.
5. The method for dynamically predicting the deformation of the diaphragm wall based on the composite neural network algorithm according to claim 1, wherein: and (3) after new monitoring data are collected, repeating the step (1) and the step (2), and retraining the CV-LSTM model for updating the deformation prediction result.
6. The method for dynamically predicting the deformation of the diaphragm wall based on the composite neural network algorithm according to claim 1, wherein: and when the output result is that the deformation prediction value is close to the warning value, pushing the early warning information and taking corresponding deformation control measures in time.
CN202011572061.7A 2020-12-27 2020-12-27 Ground connection wall deformation dynamic prediction method based on composite neural network algorithm Pending CN112836789A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202011572061.7A CN112836789A (en) 2020-12-27 2020-12-27 Ground connection wall deformation dynamic prediction method based on composite neural network algorithm

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202011572061.7A CN112836789A (en) 2020-12-27 2020-12-27 Ground connection wall deformation dynamic prediction method based on composite neural network algorithm

Publications (1)

Publication Number Publication Date
CN112836789A true CN112836789A (en) 2021-05-25

Family

ID=75924749

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202011572061.7A Pending CN112836789A (en) 2020-12-27 2020-12-27 Ground connection wall deformation dynamic prediction method based on composite neural network algorithm

Country Status (1)

Country Link
CN (1) CN112836789A (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113792372A (en) * 2021-09-28 2021-12-14 苏州大学 Ground continuous wall deformation dynamic prediction method based on CV-LSTM combined model
CN113807031A (en) * 2021-11-18 2021-12-17 广东智云工程科技有限公司 Foundation pit disaster prediction early warning method based on LSTM and depth residual error neural network
CN117390381A (en) * 2023-09-11 2024-01-12 北京市政建设集团有限责任公司 Underground diaphragm wall joint seam leakage prediction method and device based on deep learning

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104616433A (en) * 2015-02-04 2015-05-13 清华大学 Real-time monitoring and early warning system for foundation pit engineering
CN107832568A (en) * 2017-12-12 2018-03-23 青岛理工大学 The assay method of deep pit digging soil layer dynamic level resistance coefficient
CN110470259A (en) * 2019-07-09 2019-11-19 西安工程大学 Landslide displacement dynamic prediction method based on LSTM
CN110608711A (en) * 2019-08-08 2019-12-24 广东毓秀科技有限公司 Method for predicting deformation of rail-crossing tunnel through big data
CN110704934A (en) * 2019-10-12 2020-01-17 上海勘察设计研究院(集团)有限公司 Web end lightweight display method based on BIM model
CN111475948A (en) * 2020-04-07 2020-07-31 青岛理工大学 Novel prediction method for settlement during subway tunnel construction based on L STM network
CN111486814A (en) * 2020-04-18 2020-08-04 福州大学 BIM three-dimensional visualization deep foundation pit deformation prediction system based on genetic algorithm

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104616433A (en) * 2015-02-04 2015-05-13 清华大学 Real-time monitoring and early warning system for foundation pit engineering
CN107832568A (en) * 2017-12-12 2018-03-23 青岛理工大学 The assay method of deep pit digging soil layer dynamic level resistance coefficient
CN110470259A (en) * 2019-07-09 2019-11-19 西安工程大学 Landslide displacement dynamic prediction method based on LSTM
CN110608711A (en) * 2019-08-08 2019-12-24 广东毓秀科技有限公司 Method for predicting deformation of rail-crossing tunnel through big data
CN110704934A (en) * 2019-10-12 2020-01-17 上海勘察设计研究院(集团)有限公司 Web end lightweight display method based on BIM model
CN111475948A (en) * 2020-04-07 2020-07-31 青岛理工大学 Novel prediction method for settlement during subway tunnel construction based on L STM network
CN111486814A (en) * 2020-04-18 2020-08-04 福州大学 BIM three-dimensional visualization deep foundation pit deformation prediction system based on genetic algorithm

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
卞恺慧: "基于神经网络的基坑形变预测研究及应用", 《中国优秀硕士学位论文全文数据库 基础科学辑》 *
王万通: "基于模糊神经网络的深基坑变形预测", 《重庆工学院学报(自然科学)》 *

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113792372A (en) * 2021-09-28 2021-12-14 苏州大学 Ground continuous wall deformation dynamic prediction method based on CV-LSTM combined model
CN113807031A (en) * 2021-11-18 2021-12-17 广东智云工程科技有限公司 Foundation pit disaster prediction early warning method based on LSTM and depth residual error neural network
CN117390381A (en) * 2023-09-11 2024-01-12 北京市政建设集团有限责任公司 Underground diaphragm wall joint seam leakage prediction method and device based on deep learning

Similar Documents

Publication Publication Date Title
CN107688871B (en) Water quality prediction method and device
CN112836789A (en) Ground connection wall deformation dynamic prediction method based on composite neural network algorithm
CN110427654B (en) Landslide prediction model construction method and system based on sensitive state
CN113792372A (en) Ground continuous wall deformation dynamic prediction method based on CV-LSTM combined model
KR102181966B1 (en) Soft survey method and system for hydraulic cylinder comprehensive test station
Zhang et al. Improving dam seepage prediction using Back‐propagation neural network and genetic algorithm
CN112036658B (en) Wind power real-time prediction method based on forgetting long-term and short-term memory with difference
CN114492211B (en) Residual oil distribution prediction method based on autoregressive network model
CN110533239B (en) Smart city air quality high-precision measurement method
Fei et al. Research on tunnel engineering monitoring technology based on BPNN neural network and MARS machine learning regression algorithm
CN115659749B (en) Foundation pit deformation prediction method and system, electronic equipment and storage medium
CN115077474A (en) Ground settlement trend prediction method and system based on machine learning
CN112307536B (en) Dam seepage parameter inversion method
CN115471679A (en) Method and intelligent system for synchronously assimilating water level and flow of natural river
CN114969932A (en) Foundation pit horizontal displacement prediction method based on TSNE-BP neural network
CN115935834A (en) History fitting method based on deep autoregressive network and continuous learning strategy
CN111428419A (en) Suspended sediment concentration prediction method and device, computer equipment and storage medium
CN113408190B (en) Surrounding rock deformation prediction method for highway tunnel construction period based on Bayes-LSTM model
Li et al. An improved cyclic multi model-eXtreme gradient boosting (CMM-XGBoost) forecasting algorithm on the GNSS vertical time series
CN110671092A (en) Oil gas productivity detection method and system
CN117113854B (en) Salt tide forecasting method based on ConvLSTM and three-dimensional numerical simulation
CN117408167A (en) Debris flow disaster vulnerability prediction method based on deep neural network
CN117252063A (en) Rock high slope excavation anchoring deformation prediction method and system based on machine learning
Xu et al. Dam settlement prediction based on random error extraction and multi-input LSTM network
CN115906669A (en) Dense residual error network landslide susceptibility evaluation method considering negative sample selection strategy

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