CN116739176A - Tunnel mechanized construction risk prediction method based on deep belief network - Google Patents
Tunnel mechanized construction risk prediction method based on deep belief network Download PDFInfo
- Publication number
- CN116739176A CN116739176A CN202310743584.0A CN202310743584A CN116739176A CN 116739176 A CN116739176 A CN 116739176A CN 202310743584 A CN202310743584 A CN 202310743584A CN 116739176 A CN116739176 A CN 116739176A
- Authority
- CN
- China
- Prior art keywords
- mechanized construction
- tunnel
- risk
- model
- tunnel mechanized
- 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
- 238000010276 construction Methods 0.000 title claims abstract description 152
- 238000000034 method Methods 0.000 title claims abstract description 54
- 239000002245 particle Substances 0.000 claims abstract description 51
- 238000012549 training Methods 0.000 claims abstract description 49
- 238000013058 risk prediction model Methods 0.000 claims abstract description 44
- 238000011156 evaluation Methods 0.000 claims abstract description 32
- 238000012360 testing method Methods 0.000 claims abstract description 22
- 238000005457 optimization Methods 0.000 claims abstract description 17
- 238000007781 pre-processing Methods 0.000 claims abstract description 11
- 238000012795 verification Methods 0.000 claims abstract description 4
- 210000002569 neuron Anatomy 0.000 claims description 27
- 239000013598 vector Substances 0.000 claims description 27
- 230000008569 process Effects 0.000 claims description 17
- 230000006870 function Effects 0.000 claims description 16
- 239000011435 rock Substances 0.000 claims description 15
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims description 10
- 230000002159 abnormal effect Effects 0.000 claims description 9
- 230000008859 change Effects 0.000 claims description 9
- 230000000694 effects Effects 0.000 claims description 5
- 238000010200 validation analysis Methods 0.000 claims description 5
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 claims description 4
- 238000009827 uniform distribution Methods 0.000 claims description 4
- 230000004913 activation Effects 0.000 claims description 3
- 238000012937 correction Methods 0.000 claims description 3
- 238000002790 cross-validation Methods 0.000 claims description 3
- 238000012804 iterative process Methods 0.000 claims description 3
- 238000012545 processing Methods 0.000 claims description 3
- 238000013480 data collection Methods 0.000 abstract description 4
- 230000009286 beneficial effect Effects 0.000 description 3
- 230000015572 biosynthetic process Effects 0.000 description 3
- 238000009430 construction management Methods 0.000 description 3
- 230000006378 damage Effects 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 239000003673 groundwater Substances 0.000 description 3
- 238000005381 potential energy Methods 0.000 description 3
- 230000000903 blocking effect Effects 0.000 description 2
- 238000004140 cleaning Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 239000000428 dust Substances 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000001603 reducing effect Effects 0.000 description 2
- 239000004576 sand Substances 0.000 description 2
- 239000002893 slag Substances 0.000 description 2
- 239000002689 soil Substances 0.000 description 2
- 239000002352 surface water Substances 0.000 description 2
- 230000005641 tunneling Effects 0.000 description 2
- 238000009423 ventilation Methods 0.000 description 2
- 238000009825 accumulation Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000009172 bursting Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 238000009966 trimming Methods 0.000 description 1
Classifications
-
- 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
- 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
- G06N3/047—Probabilistic or stochastic 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
- G06N3/0475—Generative 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
- G06N3/088—Non-supervised learning, e.g. competitive learning
-
- 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
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0635—Risk analysis of enterprise or organisation activities
-
- 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)
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- General Physics & Mathematics (AREA)
- Economics (AREA)
- Strategic Management (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Evolutionary Computation (AREA)
- Molecular Biology (AREA)
- Data Mining & Analysis (AREA)
- Biomedical Technology (AREA)
- Software Systems (AREA)
- Computational Linguistics (AREA)
- Mathematical Physics (AREA)
- General Engineering & Computer Science (AREA)
- Biophysics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Marketing (AREA)
- General Business, Economics & Management (AREA)
- Tourism & Hospitality (AREA)
- Entrepreneurship & Innovation (AREA)
- Quality & Reliability (AREA)
- Operations Research (AREA)
- Game Theory and Decision Science (AREA)
- Development Economics (AREA)
- Primary Health Care (AREA)
- Educational Administration (AREA)
- Probability & Statistics with Applications (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The application discloses a tunnel mechanized construction risk prediction method based on a deep confidence network, which comprises the steps of establishing a tunnel mechanized construction risk evaluation system, establishing a tunnel mechanized construction risk prediction database, dividing a training set and a testing set, then carrying out learning training and verification on the training set, testing the performance of a tunnel mechanized construction risk prediction model by the testing set, and simultaneously adjusting the super parameters of a DBN model by using an improved particle swarm optimization algorithm, so that model training is accelerated and accuracy is improved. The method integrates data collection, data preprocessing, model building and risk prediction, can predict the tunnel mechanized construction risk according to the survey condition before construction, and can reduce the risk by adopting related measures by a designer.
Description
Technical Field
The application relates to a tunnel mechanized construction risk prediction method based on a deep belief network, and belongs to the technical field of tunnel mechanized construction.
Background
Along with the rapid development of tunnel construction, the tunnel boring machine is widely applied to tunnel and underground engineering construction in the fields of traffic, water conservancy and the like because of the characteristics of high boring speed, safe and reliable construction, environmental protection in universities, reduced labor intensity of workers and the like. If sudden unexpected situations occur in the tunnel mechanized construction process, the sudden unexpected situations can cause intolerable huge losses to constructors and equipment. It is necessary to predict the risk of the tunnel mechanized construction by analyzing the site risk source before construction. Before construction, the possible value of the occurrence risk is obtained by predicting the mechanical construction risk of the tunnel. When the risk is too large, the designer can reduce the risk by taking relevant measures, so that a construction scheme for minimizing the construction risk is adopted, and the life safety of the constructor is guaranteed and the property loss is reduced.
Currently, more traditional methods are to subjectively judge construction risks by expert experience, and risk identification is carried out by only expert experience, so that risk omission and incomplete identification are likely to be caused. And the tunnel mechanized construction has the characteristics of complexity, dynamic property and nonlinearity, and the problems of fuzziness and randomness of tunnel mechanized construction risk evaluation cannot be solved only by using an expert evaluation method, which is shown by lower accuracy of a prediction result and cannot meet the requirement of construction reference.
Disclosure of Invention
According to one aspect of the application, a tunnel mechanized construction risk prediction method based on a deep confidence network is provided, the prediction method integrates data collection, data preprocessing, model building and risk prediction, the tunnel mechanized construction risk can be predicted according to the survey condition before construction, and a designer can reduce the risk by taking relevant measures.
The tunnel mechanized construction risk prediction method based on the deep belief network is characterized by comprising the following steps of:
s1, establishing a tunnel mechanized construction risk factor evaluation standard;
s2, establishing a tunnel mechanized construction risk index system according to a risk factor evaluation standard;
s3, collecting experts to establish risk factor evaluation vectors and risk prediction results according to risk factor evaluation criteria, wherein each group of risk factor evaluation vectors corresponds to one group of risk prediction results;
s4, preprocessing the data collected in the S3, wherein the preprocessing comprises the steps of processing the missing value and the abnormal value in the data set, so that the error influence of the abnormal data on model training is avoided;
s5, constructing a tunnel mechanized construction risk prediction database according to the preprocessed data;
s6, dividing the data in the prediction database in the S5 into a training set and a testing set, wherein the training set data are used for training the performance of the model and the verification model, and the testing set is used for testing the prediction accuracy of the optimal tunnel mechanized construction risk prediction model;
s7, initializing super parameters of a tunnel mechanized construction risk prediction model;
s8, a K-fold cross validation method is used, the training set is divided into a new training set and a validation set again, model parameters are initialized randomly, the new training set data is used for training a model, and the performance of the data test model in the validation set is used;
s9, inputting feature quantities of the tunnel mechanized construction risk factors by using a limited Boltzmann machine RBM (radial basis function) unsupervised forward learning, and training layer by layer to enable shallow original features to obtain high-level expression;
s10, reversely fine-tuning the super-parameters of the tunnel mechanized construction risk prediction model by using an error reverse propagation algorithm to enable the super-parameters to be converged to a global optimal point, and optimizing the identification capacity of the model on risk factors;
s11, adjusting super parameters of a tunnel mechanized construction risk prediction model by using an improved particle swarm optimization algorithm;
s12, judging whether the tunnel mechanized construction risk prediction model meets the optimizing condition of the particle swarm optimization algorithm, if so, obtaining the tunnel mechanized construction risk prediction model with optimal performance, if not, constructing a new tunnel mechanized construction risk prediction model by using the super parameters determined in the S11, and repeating the steps S8-S11;
s13, using the tunnel mechanized construction risk prediction model obtained in the S12, inputting test set data to predict tunnel mechanized construction risk, and testing the prediction performance of the tunnel mechanized construction risk prediction model;
s14, according to the site construction condition and the risk factor evaluation standard, the site personnel inputs the risk factor membership vector into the tunnel mechanized construction risk prediction model to obtain the tunnel mechanized construction risk probability.
Further, the super parameters comprise the number of input nodes, the number of output nodes, the number of hidden layers nodes, the number of training iterations, the batch size, the learning rate and the momentum;
the super parameters to be initialized include the number of input nodes, the number of output nodes, the number of hidden layers nodes, the number of training iterations, the batch size, the learning rate and the momentum.
Further, the initialization model parameters include neuron connection weights w of the visible layer unit i and the hidden layer unit j ij Bias a of visible layer element neuron i i And bias b of hidden layer element j j ；
Wherein, the absolute average error MAE and the relative average error MRE are used as performance evaluation indexes of the model, and are defined as follows:
wherein: y is output,i And t fact,i Respectively representing a predicted value and a true value of the input of the ith sample;
the smaller the MRE and MAE values are, the more accurate the prediction effect of the tunnel mechanized construction risk prediction model is.
Further, gaussian noise is added to the visible layer where the visible layer unit is located and the hidden layer where the hidden layer unit is located.
Further, the unsupervised forward learning process of the limited boltzmann machine RBM in S9 includes the following steps:
parameter initialization: initializing RBM model parameters, namely connecting weights among layers and offsets of all layers, and randomly extracting samples from uniform distribution in [ -1,1] to serve as parameters of the RBM model;
the joint energy function between the visible layer and the hidden layer is:
wherein v is i Is the i-th neuron state in the visible layer;
h j is the j-th neuron state in the hidden layer;
a i and b j Bias of the ith neuron in the visible layer and the jth neuron in the hidden layer, respectively;
w ij is the weight between the ith neuron in the visible layer and the jth neuron in the hidden layer;
θ＝[w ij ,a i ,b j ]the table type requires a parameter space which is solved through training;
c is the standard deviation in the gaussian function;
calculate hidden layer element h j Probability of being activated:
wherein: σ () represents using a Sigmoid function as an activation function;
calculating visible layer element v i Probability of being activated:
using the CD-1 algorithm, weights and biases for the model are calculated:
wherein Deltaw is ij Is the value of the change in weight;
Δa i a change value of the bias vector in the visible layer;
Δb j is the value of the change in the bias vector in the hidden layer.
Further, the reverse trimming process using the error back propagation algorithm includes the following steps:
the gradient descent process is designed by using an Adam algorithm, after bias correction, each iteration learning rate has a certain range, and the supervised learning of the tunnel mechanized construction risk prediction model is accelerated;
the update rule is as follows:
wherein: m is m t And v t Respectively representing first moment estimation and second moment estimation of the t-th iteration parameter;
epsilon is a very small constant for preventing the denominator from being 0;
alpha represents the step factor of the network weight update.
Further, in S11, the step of adjusting the super parameter of the tunnel mechanized construction risk prediction model by using the particle swarm optimization algorithm is as follows:
11-1 initializing a particle swarm;
wherein the position of the particle is initializedAnd speed->
11-2, calculating a fitness value;
calculating the fitness value of each particle, and finding the optimal position of the particle in the round of particle swarmAnd search for the position of the optimal particle in history +.>
11-3 update the speed and position of the particles:
in the method, in the process of the application,is the particle velocity; />Is the particle position;
ω represents inertial weight, the value of which is between [0,1], typically ω=0.9;
c 1 、c 2 is a learning factor;
r 1 、r 2 is in [0,1]]Random numbers in between;
is the optimal position of the ith particle;
is a global optimal particle location;
11-4 if the misclassification rate of the training sample meets the set condition or the iteration number reaches a preset value, the particle swarm optimization is finished, otherwise, the process jumps to 11-2, k=k+1, and 11-3 and 11-4 are repeatedly executed until the discrimination condition is met;
dynamically adjusting the inertial weight omega in the iterative process:
wherein k is the iteration number;
k max the maximum iteration number of the algorithm is calculated;
ω max is the maximum value of the inertia weight;
ω min is the minimum of the inertial weights.
Further, the prediction method further includes the steps of modifying input factors of risk prediction during construction, and modifying the input factors as follows:
15.1 taking 7 factors of underground water condition, rock stratum property, rock property, ground stress influence, underground pipeline condition and distance from surrounding buildings as risk evaluation vectors in construction;
15.2 repeating S1-S12, and retraining a tunnel mechanized construction risk prediction model used in construction;
and 15.3, deploying the trained tunnel mechanized construction risk prediction model on a tunnel boring machine to realize timely tunnel mechanized construction risk prediction.
The application has the beneficial effects that:
1) According to the tunnel mechanized construction risk prediction method of the deep belief network, disclosed by the application, the super parameters of the DBN model are adjusted by using an improved particle swarm optimization algorithm, so that model training is accelerated and accuracy is improved; the method integrates data collection, data preprocessing, model building and risk prediction, and can predict the tunnel mechanized construction risk according to the survey condition before construction; the construction reference can be provided for on-site workers, and the designer can reduce the risk by taking relevant measures, so that the construction scheme for minimizing the construction risk is adopted, thereby being beneficial to guaranteeing the life safety of the constructor and reducing the property loss.
Drawings
FIG. 1 is a schematic flow chart of the present application;
FIG. 2 is a diagram of a deep belief network model incorporating the present application;
fig. 3 is a block diagram of an RBM model included in the present application.
Detailed Description
The present application is described in detail below with reference to examples, but the present application is not limited to these examples.
Referring to fig. 1-3, a method for predicting the risk of tunnel mechanized construction based on a deep belief network, the method for predicting the risk of construction comprises the following steps:
s1, establishing a tunnel mechanized construction risk factor evaluation standard;
specifically, factors that may cause risk of construction include: groundwater conditions, rock formation characteristics, rock characteristics, ground stress effects, underground pipeline conditions, distances from surrounding buildings, working sites, construction operation non-norms, construction technology shortages and construction management confusion;
wherein, groundwater condition: the underground water or partial surface water contained in the surrounding rock of the tunnel mainly comes from the damage caused by the fact that the underground water or partial surface water enters the tunnel in a seepage or gushing mode during construction. When the moisture content in the air is too high, construction mechanical equipment can be corroded, and serious tunnel face water burst even has a certain probability to directly cause surrounding rock collapse, so that the tunnel mechanized construction is threatened;
formation properties: the influence of different rock strata on the mechanized construction of the tunnel is different;
ground stress influence: ground stress is the stress that exists inside the crust. In the mechanical construction process of the tunnel, the stress of surrounding rock is changed, the structural stress is also caused, the risk of rock burst occurs, and the safety of rock mass is greatly influenced;
underground pipeline condition: when an underground pipeline exists in a region needing construction, if the underground pipeline is damaged due to misoperation, the life of construction equipment and surrounding residents can be influenced;
distance from surrounding building: when the surrounding building is too close to the construction site, the construction is affected to a certain extent;
work site: it is also important whether the operation site has safety guarantee measures, and when the safety measures are prepared fully, the probability of risk occurrence is lower;
the risk of the mechanized construction of the tunnel is increased due to the fact that construction operation is not standard, construction technology is insufficient and construction management is disordered.
Given that some risk factors are too abstract, it is difficult to directly use them to train a neural network model, and the risk factors must be digitized. The influence of the risk factors on the construction is classified into 5 grades of no influence, small influence, medium influence, large influence and serious influence; accordingly, these five classes are denoted by the corresponding numbers 1, 2, 3, 4, 5, respectively, and a small number may be used when the impact is between the two classes.
S2, establishing a tunnel mechanized construction risk index system according to a risk factor evaluation standard;
specifically, in consideration of risk factors in the step S1, a tunnel mechanized construction risk index system is established by combining experience of relevant experts.
The first-level risks include geological risks, ambient environment risks, equipment risks and risks of the tunnel.
The geological risks include tunnel face collapse, soil body leakage and water bursting sand burst, the surrounding environment risks include stratum ground settlement, surrounding building deformation and underground pipeline damage, the equipment risks include tunnel boring machine blocking machine, tunnel boring machine assembly and debugging failure, bracket deformation instability, unsmooth ventilation and dust collection and discontinuous slag discharge, and the tunnel risks include tunneling route deviation, supporting structure deformation, lining leakage and tunnel ponding, and total 16 secondary risks.
S3, collecting risk factor evaluation vectors and risk prediction results established by experts according to risk factor evaluation standards;
the risk factor evaluation vectors are shown in table 1:
TABLE 1
The vector format is as follows:
x＝[x 1 ,x 2 ,…,x 9 ,x 10 ] x i ∈R[1,5]
wherein, the risk prediction result vector is shown in table 2:
TABLE 2
The vector format is as follows:
y＝[y 1 ,y 2 ,y 3 ,y 4 ,y 5 ]，y i ∈R[0,1]and is also provided with
It should be noted that each set of risk factor evaluation vectors corresponds to a set of risk prediction results.
S4, preprocessing the data collected in the S3;
the method comprises the steps of preprocessing collected data before dividing the data, wherein the data preprocessing method is data cleaning;
the data cleaning is mainly to avoid the error influence of abnormal data on model training by processing the missing value and the abnormal value in the data set.
The missing value is expressed as Null in the data storage library, and the processing method is discarding, namely directly deleting the row record with the missing value;
wherein the values in the risk factor evaluation vector satisfy x i ∈R[1,5]When x is i If not, the value is abnormal;
the values in the risk prediction result vector satisfy
y i ∈R[0,1]And is also provided with
When y is i If the condition is not satisfied, the value is an outlier.
Discarding the abnormal value, namely directly deleting the row record containing the abnormal value;
s5, constructing a tunnel mechanized construction risk prediction database according to the preprocessed data;
and (3) establishing a tunnel mechanized construction risk prediction database by using the data processed by the S4.
S6, dividing the data in the S5 into a training set and a testing set, taking 80% of the original data set as the training set and the rest 20% as the testing set;
the training set data are used for training the performance of the model and verifying the performance of the model so as to select a tunnel mechanized construction risk prediction model with the best prediction effect from the training set data, and the testing set is used for testing the prediction accuracy of the optimal tunnel mechanized construction risk prediction model.
And S7, randomly initializing super parameters of a tunnel mechanized construction risk prediction model.
The super parameters to be initialized include the number of input nodes, the number of output nodes, the number of hidden layers, the number of hidden layer nodes, the number of training iterations, the batch size, the learning rate and the momentum;
wherein the number of input nodes is equal to the number of input variables in the data to be processed. According to the risk factor evaluation vector, the system is divided into 10 nodes, which respectively correspond to underground water conditions, rock layer characteristics, rock characteristics, ground stress influence, underground pipeline conditions, distances from surrounding buildings, operation sites, construction operation non-normative, construction technology shortage and construction management confusion;
it should be pointed out that the next hidden layer at the input node should be divided into 16 nodes, which correspond to the face collapse, soil leakage, water burst and sand burst, stratum ground subsidence, surrounding building deformation, underground pipeline destruction, tunnel boring machine blocking, tunnel boring machine assembly and debugging failure, bracket deformation instability, ventilation and dust collection unsmooth, slag discharge discontinuity, tunneling route deviation, support structure deformation, lining leakage and tunnel water accumulation.
The number of output nodes is equal to the number of outputs associated with each input, 1 node is known according to the prediction result vector, and the tunnel mechanized construction risk is corresponding to the number of the output nodes;
it should be noted that the last hidden layer at the output node should be divided into 4 nodes, corresponding to geological risk, ambient risk, equipment risk and tunnel own risk, respectively;
hidden layer number: the meaning of the hidden layer is to abstract the characteristics of the input data to another dimension space to show the more abstract characteristics of the input data, and the characteristics can be better divided linearly. When the number of layers is too small, the features which can be extracted are limited, and when the problem is too complex, the accuracy of the model may not meet the requirement, and thus the model is under-fitted. The deeper the layer number is, the better the ability of fitting the function is, but the problem of over fitting can be brought, and the training difficulty is increased, so that the model is difficult to converge;
number of hidden nodes: the number of neurons used in the hidden layer. Too few neurons in the hidden layer will result in a under fit; when the neural network has too many nodes, the limited amount of information contained in the training set is insufficient to train all neurons in the hidden layer, resulting in overfitting;
batch size: sample size used in one iteration. When the batch is too small, the training data is very difficult to converge;
iteration number: each time a batch of data is trained, it is an iteration number. When the iteration number is too large, overfitting can be caused; when the number of iterations is too small, a under-fit may result.
Learning rate: amplitude size of each parameter update. When the learning rate is too large, the parameters to be optimized can be caused to fluctuate near the minimum value and cannot be converged; when the learning rate is too small, the parameters to be optimized can be slowly converged;
momentum: is proposed according to the principle of energy conversion between physical potential energy and kinetic energy. When the momentum is larger, the energy converted into potential energy is larger, so that the potential energy is more likely to get rid of the constraint of the local concave domain, enter the global concave domain and further obtain the global optimal solution; momentum is mainly used when the weights are updated.
S8, a K-fold cross validation method is used to re-divide the training set into a new training set and a new validation set; randomly initializing model parameters, training a model by using new training set data, and testing the performance of the model by using data in a verification set;
wherein the randomly initialized model parameters are parameters of randomly extracting samples from the uniform distribution of [ -1,1] as a model;
wherein, the parameters of the model are the neuron connection weights w of the visible layer unit i and the hidden layer unit j ij The bias a of layer element neuron i is seen i Bias b of hidden layer unit j j ；
Wherein, the absolute average error MAE and the relative average error MRE are used as performance evaluation indexes of the model, and are defined as follows:
wherein: y is output,i And y fact,i Representing the predicted and actual values of the i-th sample input, respectively. The smaller the MRE and MAE values are, the more accurate the prediction effect of the tunnel mechanized construction risk prediction model is.
S9, inputting risk factor characteristic quantity by using a limited Boltzmann machine RBM (radial basis function) unsupervised forward learning, training layer by layer, enabling shallow original characteristics to obtain high-level expression, and adding Gaussian noise into a visible layer and a hidden layer in order to enhance generalization capability of a model;
the RBM unsupervised forward learning process comprises the steps of:
parameter initialization: initializing RBM model parameters, namely the connection weight between layers and the bias of each layer, and randomly extracting samples from uniform distribution in [ -1,1] as parameters of the model;
the joint energy function between the visible layer and the hidden layer is:
wherein v is i Is the i-th neuron state in the visible layer; h is a j Is the j-th neuron state in the hidden layer; a, a i And b j Bias of the ith neuron in the visible layer and the jth neuron in the hidden layer, respectively; w (w) ij Is the weight between the ith neuron in the visible layer and the jth neuron in the hidden layer; θ= [ w ] ij ,a i ,b j ]The table type requires a parameter space which is solved through training; c is the standard deviation in the gaussian function;
calculate hidden layer element h j Probability of being activated:
wherein: σ () represents using a Sigmoid function as an activation function;
calculating visible layer element v i Probability of being activated:
using the CD-1 algorithm (contrast divergence algorithm), the weights and biases of the model are calculated:
wherein Deltaw is ij Is the value of the change in weight;
Δa i a change value of the bias vector in the visible layer;
Δb j is the value of the change in the bias vector in the hidden layer.
S10, reversely fine-tuning the super-parameters of the tunnel mechanized construction risk prediction model by using a BP algorithm (error back propagation algorithm) to enable the super-parameters to be converged to a global optimal point, optimizing the identification capability of the model to the risk factors,
preferably, the back-tuning process using the BP algorithm includes the steps of:
the gradient descent process is designed by using an Adam algorithm, and the method has the advantages that after bias correction, each iteration learning rate has a determined range, and the supervised learning of the tunnel mechanized construction risk prediction model is accelerated; the update rule is as follows:
wherein: m is m t And v t Respectively representing first moment estimation and second moment estimation of the t-th iteration parameter; epsilon is a very small constant for preventing the denominator from being 0; alpha represents a step factor of updating the network weight;
adam algorithm is a random optimization method of self-adaptive momentum, and is an optimizer algorithm in deep learning.
S11, using an improved particle swarm optimization algorithm to adjust the super parameters of the DBN model;
wherein, the step of adjusting the super parameter of the DBN model by using the particle swarm optimization algorithm is as follows:
11-1, initializing a particle swarm;
wherein the position of the particle is initializedAnd speed->
11-2, calculating a fitness value;
calculating the fitness value of each particle, and finding the optimal position of the particle in the round of particle swarmAnd search for the position of the optimal particle in history +.>
11-3, updating the speed and the position of the particles;
wherein the velocities and positions of all particles are updated:
in the method, in the process of the application,is the particle velocity; />Is the particle position; omega represents inertia weight and has a value between 0 and 1]ω=0.9 is generally taken; c 1 、c 2 Is a learning factor; r is (r) 1 、r 2 Is in [0,1]]Random numbers in between; />Is the optimal position of the ith particle;is a global optimal particle location;
11-4 if the misclassification rate of the training sample meets the set condition or the iteration number reaches the preset value, the particle swarm PSO optimization is finished, otherwise, the process jumps to 11-2, k=k+1, and 11-3 and 11-4 are repeatedly executed until the discrimination condition is met.
In the conventional particle swarm optimization algorithm, the inertia weight ω is a constant value in the iterative process, which easily causes the algorithm to fall into a local optimum. In order to avoid the initial particle falling into the local optimum, ω can be dynamically adjusted during the iteration:
wherein k is the iteration number; k (k) max The maximum iteration number of the algorithm is calculated; omega max Is the maximum value of the inertia weight; omega min Is the minimum of the inertial weights.
And S12, judging whether the tunnel mechanized construction risk prediction model meets the optimizing condition of the particle swarm optimization algorithm, if so, obtaining the tunnel mechanized construction risk prediction model with optimal performance, and if not, constructing a new tunnel mechanized construction risk prediction model by using the super parameters determined in the step S11, and repeating the steps S8-S11.
S13, using the tunnel mechanized construction risk prediction model obtained in S12, inputting test set data to predict tunnel mechanized construction risk, and testing the prediction performance of the tunnel mechanized construction risk prediction model.
S14, according to the site construction condition and the risk factor evaluation standard, the site personnel inputs the risk factor membership vector into the tunnel mechanized construction risk prediction model, so that the tunnel mechanized construction risk probability can be obtained.
In actual use, the risk factor evaluation vector is obtained by field personnel according to the field construction condition and the risk factor evaluation standard.
S15, considering that the risk prediction can only be used for risk prediction before construction, input factors need to be modified to be used for risk prediction during construction;
the modification steps are as follows:
15-1 taking into account the importance of 7 factors of groundwater condition, formation property, rock property, ground stress influence, underground pipeline condition and distance from surrounding building at the time of construction, these 7 factors should be taken as risk evaluation vectors at the time of construction;
15-2 repeating the steps S1-S12, and retraining a tunnel mechanized construction risk prediction model used in construction;
15-3, the trained tunnel mechanized construction risk prediction model can be deployed on a tunnel boring machine, so that timely tunnel mechanized construction risk prediction is realized.
Compared with the traditional DBN, the tunnel mechanized construction risk prediction method of the deep belief network provided by the application has the advantages that the super-parameters of the DBN model are adjusted by using an improved particle swarm optimization algorithm, so that model training is accelerated and accuracy is improved; compared with the traditional tunnel mechanized construction risk prediction method, the method integrates data collection, data preprocessing, model establishment and risk prediction, and can predict tunnel mechanized construction risk according to the survey condition before construction; in order to improve the universality, the input parameters are adjusted so that the input parameters can be deployed on a tunnel boring machine for risk prediction in construction; the method can provide construction reference for field staff, and designers can reduce risks by taking relevant measures, so that the construction scheme for minimizing construction risks is adopted, thereby being beneficial to guaranteeing the life safety of the constructors and reducing property loss.
While the application has been described in terms of preferred embodiments, it will be understood by those skilled in the art that various changes and modifications can be made without departing from the scope of the application, and it is intended that the application is not limited to the specific embodiments disclosed.
Claims (8)
1. The tunnel mechanized construction risk prediction method based on the deep belief network is characterized by comprising the following steps of:
s1, establishing a tunnel mechanized construction risk factor evaluation standard;
s2, establishing a tunnel mechanized construction risk index system according to a risk factor evaluation standard;
s3, collecting experts to establish risk factor evaluation vectors and risk prediction results according to risk factor evaluation criteria, wherein each group of risk factor evaluation vectors corresponds to one group of risk prediction results;
s4, preprocessing the data collected in the S3, wherein the preprocessing comprises the steps of processing the missing value and the abnormal value in the data set, so that the error influence of the abnormal data on model training is avoided;
s5, constructing a tunnel mechanized construction risk prediction database according to the preprocessed data;
s6, dividing the data in the prediction database in the S5 into a training set and a testing set, wherein the training set data are used for training the performance of the model and the verification model, and the testing set is used for testing the prediction accuracy of the optimal tunnel mechanized construction risk prediction model;
s7, initializing super parameters of a tunnel mechanized construction risk prediction model;
s8, a K-fold cross validation method is used, the training set is divided into a new training set and a validation set again, model parameters are initialized randomly, the new training set data is used for training a model, and the performance of the data test model in the validation set is used;
s9, inputting feature quantities of the tunnel mechanized construction risk factors by using a limited Boltzmann machine RBM (radial basis function) unsupervised forward learning, and training layer by layer to enable shallow original features to obtain high-level expression;
s10, reversely fine-tuning the super-parameters of the tunnel mechanized construction risk prediction model by using an error reverse propagation algorithm to enable the super-parameters to be converged to a global optimal point, and optimizing the identification capacity of the model on risk factors;
s11, adjusting super parameters of a tunnel mechanized construction risk prediction model by using an improved particle swarm optimization algorithm;
s12, judging whether the tunnel mechanized construction risk prediction model meets the optimizing condition of the particle swarm optimization algorithm, if so, obtaining the tunnel mechanized construction risk prediction model with optimal performance, if not, constructing a new tunnel mechanized construction risk prediction model by using the super parameters determined in the S11, and repeating the steps S8-S11;
s13, using the tunnel mechanized construction risk prediction model obtained in the S12, inputting test set data to predict tunnel mechanized construction risk, and testing the prediction performance of the tunnel mechanized construction risk prediction model;
s14, according to the site construction condition and the risk factor evaluation standard, the site personnel inputs the risk factor membership vector into the tunnel mechanized construction risk prediction model to obtain the tunnel mechanized construction risk probability.
2. The tunnel mechanized construction risk prediction method based on the deep belief network according to claim 1, wherein the super parameters include the number of input nodes, the number of output nodes, the number of hidden layers nodes, the number of training iterations, the batch size, the learning rate and the momentum;
the super parameters to be initialized include the number of input nodes, the number of output nodes, the number of hidden layers nodes, the number of training iterations, the batch size, the learning rate and the momentum.
3. The method for predicting the risk of tunnel mechanized construction based on a deep belief network according to claim 1, wherein the initialization model parameters include neuron connection weights w of a visible layer unit i and a hidden layer unit j ij Bias a of visible layer element neuron i i And bias b of hidden layer element j j ；
Wherein, the absolute average error MAE and the relative average error MRE are used as performance evaluation indexes of the model, and are defined as follows:
wherein: y is output,i And y fact,i Respectively representing a predicted value and a true value of the input of the ith sample;
the smaller the MRE and MAE values are, the more accurate the prediction effect of the tunnel mechanized construction risk prediction model is.
4. A method for predicting risk of tunnel mechanized construction based on deep belief network according to claim 3, wherein gaussian noise is added to a visible layer where the visible layer unit is located and a hidden layer where the hidden layer unit is located.
5. The method for predicting the risk of tunnel mechanized construction based on deep belief network according to claim 4, wherein the limited boltzmann machine RBM unsupervised forward learning process in S9 comprises the steps of:
parameter initialization: initializing RBM model parameters, namely connecting weights among layers and offsets of all layers, and randomly extracting samples from uniform distribution in [ -1,1] to serve as parameters of the RBM model;
the joint energy function between the visible layer and the hidden layer is:
wherein v is i Is the i-th neuron state in the visible layer;
h j is the j-th neuron state in the hidden layer;
a i and b j Bias of the ith neuron in the visible layer and the jth neuron in the hidden layer, respectively;
w ij is the weight between the ith neuron in the visible layer and the jth neuron in the hidden layer;
θ＝[w ij ,a i ,b j ]the table type requires a parameter space which is solved through training;
c is the standard deviation in the gaussian function;
calculate hidden layer element h j Probability of being activated:
wherein: σ () represents using a Sigmoid function as an activation function;
calculating visible layer element v i Probability of being activated:
using the CD-1 algorithm, weights and biases for the model are calculated:
wherein Deltaw is ij Is the value of the change in weight;
Δa i a change value of the bias vector in the visible layer;
Δb j is the value of the change in the bias vector in the hidden layer.
6. The method for predicting the risk of tunnel mechanized construction based on a deep belief network according to claim 1, wherein the performing the reverse fine tuning process using the error back propagation algorithm comprises the steps of:
the gradient descent process is designed by using an Adam algorithm, after bias correction, each iteration learning rate has a certain range, and the supervised learning of the tunnel mechanized construction risk prediction model is accelerated;
the update rule is as follows:
wherein: m is m t And v t Respectively representing first moment estimation and second moment estimation of the t-th iteration parameter;
epsilon is a very small constant for preventing the denominator from being 0;
alpha represents the step factor of the network weight update.
7. The method for predicting the risk of tunnel mechanized construction based on the deep belief network according to claim 1, wherein the step of adjusting the super parameter of the model for predicting the risk of tunnel mechanized construction by using the particle swarm optimization algorithm in S11 comprises the following steps:
11-1 initializing a particle swarm;
wherein the position of the particle is initializedAnd speed->
11-2, calculating a fitness value;
calculating the fitness value of each particle, and finding the optimal position of the particle in the round of particle swarmAnd search for the position of the optimal particle in history +.>
11-3 update the speed and position of the particles:
in the method, in the process of the application,is the particle velocity; />Is the particle position;
ω represents inertial weight, the value of which is between [0,1], typically ω=0.9;
c 1 、c 2 is a learning factor;
r 1 、r 2 is in [0,1]]Random numbers in between;
is the optimal position of the ith particle;
is a global optimal particle location;
11-4 if the misclassification rate of the training sample meets the set condition or the iteration number reaches a preset value, the particle swarm optimization is finished, otherwise, the process jumps to 11-2, k=k+1, and 11-3 and 11-4 are repeatedly executed until the discrimination condition is met;
dynamically adjusting the inertial weight omega in the iterative process:
wherein k is the iteration number;
k max the maximum iteration number of the algorithm is calculated;
ω max is the maximum value of the inertia weight;
ω min is the minimum of the inertial weights.
8. The method for predicting the risk of tunnel mechanized construction based on a deep belief network according to claim 1, wherein the predicting method further comprises the steps of modifying input factors of risk prediction during construction, and the modifying the input factors comprises the following steps:
15.1 taking 7 factors of underground water condition, rock stratum property, rock property, ground stress influence, underground pipeline condition and distance from surrounding buildings as risk evaluation vectors in construction;
15.2 repeating S1-S12, and retraining a tunnel mechanized construction risk prediction model used in construction;
and 15.3, deploying the trained tunnel mechanized construction risk prediction model on a tunnel boring machine to realize timely tunnel mechanized construction risk prediction.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310743584.0A CN116739176A (en) | 2023-06-21 | 2023-06-21 | Tunnel mechanized construction risk prediction method based on deep belief network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310743584.0A CN116739176A (en) | 2023-06-21 | 2023-06-21 | Tunnel mechanized construction risk prediction method based on deep belief network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116739176A true CN116739176A (en) | 2023-09-12 |
Family
ID=87918253
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310743584.0A Pending CN116739176A (en) | 2023-06-21 | 2023-06-21 | Tunnel mechanized construction risk prediction method based on deep belief network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116739176A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117236137A (en) * | 2023-11-01 | 2023-12-15 | 龙建路桥股份有限公司 | Winter continuous construction control system for deep tunnel in high and cold area |
| CN117252236A (en) * | 2023-10-25 | 2023-12-19 | 中国路桥工程有限责任公司 | Tunnel blasting vibration peak prediction method, system, equipment and medium based on DBN-LSTM-BWAA |
| CN118133635A (en) * | 2024-05-07 | 2024-06-04 | 深地科学与工程云龙湖实验室 | Tunnel contact channel deformation control method, system, equipment and storage medium |
| CN118332521A (en) * | 2024-06-17 | 2024-07-12 | 山东建筑大学 | Crust deformation time sequence simulation method based on particle swarm optimization random forest |
-
2023
- 2023-06-21 CN CN202310743584.0A patent/CN116739176A/en active Pending
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117252236A (en) * | 2023-10-25 | 2023-12-19 | 中国路桥工程有限责任公司 | Tunnel blasting vibration peak prediction method, system, equipment and medium based on DBN-LSTM-BWAA |
| CN117252236B (en) * | 2023-10-25 | 2024-10-08 | 中国路桥工程有限责任公司 | Tunnel blasting vibration peak prediction method, system, equipment and medium based on DBN-LSTM-BWOA |
| CN117236137A (en) * | 2023-11-01 | 2023-12-15 | 龙建路桥股份有限公司 | Winter continuous construction control system for deep tunnel in high and cold area |
| CN117236137B (en) * | 2023-11-01 | 2024-02-02 | 龙建路桥股份有限公司 | Winter continuous construction control system for deep tunnel in high and cold area |
| CN118133635A (en) * | 2024-05-07 | 2024-06-04 | 深地科学与工程云龙湖实验室 | Tunnel contact channel deformation control method, system, equipment and storage medium |
| CN118332521A (en) * | 2024-06-17 | 2024-07-12 | 山东建筑大学 | Crust deformation time sequence simulation method based on particle swarm optimization random forest |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN116739176A (en) | Tunnel mechanized construction risk prediction method based on deep belief network | |
| Jong et al. | State-of-the-art review of geotechnical-driven artificial intelligence techniques in underground soil-structure interaction | |
| Wang et al. | Application of artificial neural network in tunnel engineering: a systematic review | |
| CN112529341B (en) | Drilling well leakage probability prediction method based on naive Bayesian algorithm | |
| Zhang et al. | A failure probability evaluation method for collapse of drill-and-blast tunnels based on multistate fuzzy Bayesian network | |
| KR101967978B1 (en) | Apparatus for predicting net penetration rate of shield tunnel boring machine and method thereof | |
| Zhang et al. | Collapse risk analysis of deep foundation pits in metro stations using a fuzzy Bayesian network and a fuzzy AHP | |
| CN108280289B (en) | Rock burst danger level prediction method based on local weighted C4.5 algorithm | |
| CN112926267B (en) | TBM tunnel rock burst grade prediction method and system based on tunneling parameter inversion | |
| Cheng et al. | Groutability prediction of microfine cement based soil improvement using evolutionary LS-SVM inference model | |
| CN109948198A (en) | A kind of fender graded method for evaluating reliability based on nonlinear function | |
| CN116084892B (en) | Automatic perforation system based on fuzzy neural network | |
| CN115017791A (en) | Tunnel surrounding rock grade identification method and device | |
| Gong et al. | Safety risk evaluations of deep foundation construction schemes based on imbalanced data sets | |
| Wu et al. | A dynamic decision approach for risk analysis in complex projects | |
| CN117252236B (en) | Tunnel blasting vibration peak prediction method, system, equipment and medium based on DBN-LSTM-BWOA | |
| CN115654381A (en) | Water supply pipeline leakage detection method based on graph neural network | |
| CN115146714A (en) | Big data management method related to underground space collapse | |
| CN116050603A (en) | Method and equipment for predicting and optimizing deformation of undercut tunnel based on hybrid intelligent method | |
| CN115481565A (en) | Earth pressure balance shield tunneling parameter prediction method based on LSTM and ant colony algorithm | |
| CN117910336A (en) | Prediction method and system for ground subsidence probability of shield construction in initial stage | |
| Wang et al. | Novel intelligent adjustment height method of Shearer drum based on adaptive fuzzy reasoning Petri net | |
| Lee et al. | Quick prediction of tunnel displacements using Artificial Neural Network and field measurement results | |
| CN113361649A (en) | Autonomous ship navigation scene clustering method for improving fuzzy C-means algorithm | |
| CN111946258A (en) | GRU-based sliding orientation intelligent control method |
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 |