CN115310348A - Stacking-based grouting amount integrated agent prediction model and prediction method - Google Patents
Stacking-based grouting amount integrated agent prediction model and prediction method Download PDFInfo
- Publication number
- CN115310348A CN115310348A CN202210798033.XA CN202210798033A CN115310348A CN 115310348 A CN115310348 A CN 115310348A CN 202210798033 A CN202210798033 A CN 202210798033A CN 115310348 A CN115310348 A CN 115310348A
- Authority
- CN
- China
- Prior art keywords
- parameters
- grouting
- model
- prediction
- value
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 66
- 238000012549 training Methods 0.000 claims abstract description 60
- 238000013528 artificial neural network Methods 0.000 claims abstract description 30
- 238000012795 verification Methods 0.000 claims abstract description 9
- 241000287127 Passeridae Species 0.000 claims description 64
- 238000010276 construction Methods 0.000 claims description 51
- 238000004088 simulation Methods 0.000 claims description 40
- 239000002002 slurry Substances 0.000 claims description 30
- 239000003795 chemical substances by application Substances 0.000 claims description 26
- 230000000739 chaotic effect Effects 0.000 claims description 22
- 238000010845 search algorithm Methods 0.000 claims description 21
- 238000013507 mapping Methods 0.000 claims description 19
- 238000005457 optimization Methods 0.000 claims description 17
- 238000012360 testing method Methods 0.000 claims description 16
- 238000005291 chaos (dynamical) Methods 0.000 claims description 11
- 239000011440 grout Substances 0.000 claims description 9
- 239000004568 cement Substances 0.000 claims description 7
- 238000002790 cross-validation Methods 0.000 claims description 6
- 238000005070 sampling Methods 0.000 claims description 5
- 238000003066 decision tree Methods 0.000 claims description 4
- 239000011159 matrix material Substances 0.000 claims description 4
- 238000012163 sequencing technique Methods 0.000 claims description 4
- 238000012856 packing Methods 0.000 claims description 2
- 238000013277 forecasting method Methods 0.000 claims 3
- 238000004422 calculation algorithm Methods 0.000 description 19
- 230000004044 response Effects 0.000 description 14
- 230000008569 process Effects 0.000 description 12
- 238000010801 machine learning Methods 0.000 description 10
- 230000008901 benefit Effects 0.000 description 6
- 238000004364 calculation method Methods 0.000 description 5
- 238000003062 neural network model Methods 0.000 description 5
- 238000007637 random forest analysis Methods 0.000 description 3
- 230000003044 adaptive effect Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 239000011435 rock Substances 0.000 description 2
- 238000012706 support-vector machine Methods 0.000 description 2
- 241000282461 Canis lupus Species 0.000 description 1
- 241001206881 Myrmeleon inconspicuus Species 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000007636 ensemble learning method Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 239000012530 fluid Substances 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 230000005484 gravity Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000006698 induction Effects 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 239000002245 particle Substances 0.000 description 1
- 230000008092 positive effect Effects 0.000 description 1
- 230000002265 prevention Effects 0.000 description 1
- 238000004886 process control Methods 0.000 description 1
- 230000008439 repair process Effects 0.000 description 1
- 230000001360 synchronised effect Effects 0.000 description 1
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
- G06N20/20—Ensemble learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N7/00—Computing arrangements based on specific mathematical models
- G06N7/08—Computing arrangements based on specific mathematical models using chaos models or non-linear system models
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2111/00—Details relating to CAD techniques
- G06F2111/10—Numerical modelling
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Software Systems (AREA)
- Evolutionary Computation (AREA)
- Business, Economics & Management (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Computing Systems (AREA)
- Data Mining & Analysis (AREA)
- Strategic Management (AREA)
- Human Resources & Organizations (AREA)
- Mathematical Physics (AREA)
- Economics (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Health & Medical Sciences (AREA)
- Marketing (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Quality & Reliability (AREA)
- Operations Research (AREA)
- Entrepreneurship & Innovation (AREA)
- Primary Health Care (AREA)
- Game Theory and Decision Science (AREA)
- Development Economics (AREA)
- Nonlinear Science (AREA)
- Algebra (AREA)
- Computational Mathematics (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Pure & Applied Mathematics (AREA)
- Computer Hardware Design (AREA)
Abstract
The invention discloses a Stacking-based grouting quantity integrated proxy prediction model, which comprises an integrated proxy model, wherein the integrated proxy model is provided with two layers, the first layer comprises three base learners which are trained and verified by adopting a five-fold cross verification method, and the second layer comprises a meta-learner; the three base learners are respectively an SVR neural network, a BPNN neural network and an RF model; the meta learner is an ANFIS neural network; the training data in the training set of the meta learner includes predicted outcome data for the three base learners. The invention also discloses a method for predicting the grouting amount integrated proxy prediction model based on Stacking. The method can increase the diversity of the model, reduce overfitting and prediction uncertainty and generate a more accurate and more stable prediction result.
Description
Technical Field
The invention relates to the field of dam bedrock grouting construction in water conservancy and hydropower engineering, in particular to a grouting amount integrated proxy prediction model and a grouting amount integrated proxy prediction method based on Stacking.
Background
At present, grouting construction is used as a main method for dam foundation seepage prevention, foundation improvement and repair, and is very important for ensuring the stable operation of hydraulic buildings. The grouting amount is an important parameter for representing the grouting construction quality, and is closely related to the cost benefit of grouting construction. By predicting the grouting amount, the grouting design scheme can be perfected, the grouting material and equipment investment can be saved, and the follow-up grouting construction control and engineering amount optimization can be guided.
However, due to the concealment and complexity of grouting engineering, the grouting amount estimated according to traditional methods such as experience of constructors and simplified physical model tests has a large error from the actual situation. With the rapid development of computational fluid mechanics methods and discrete fracture network modeling methods, the numerical simulation technology becomes a powerful tool for calculating the grouting amount of the dam foundation. However, the numerical simulation has the problems of complex modeling process, large calculated amount, long time consumption and the like, so that the grouting amount calculation efficiency is low, and the requirement of guiding grouting engineering construction cannot be met. The agent model based on the machine learning algorithm can replace a complex time-consuming numerical simulation process, and meanwhile, both the calculation efficiency and the prediction precision are considered. Therefore, the agent model technology based on the machine learning algorithm becomes an effective way for solving the problem of grouting amount prediction of complex engineering.
However, the single machine learning agent model has poor prediction stability, may have poor generalization performance due to randomness, and using only one machine learning model tends to underestimate the uncertainty of prediction to poor accuracy. Therefore, some studies improve the overall prediction accuracy by adopting a combined proxy model. However, the combined proxy model has subjectivity and uncertainty in determining the weighting value of each algorithm, and a more excellent model cannot be trained. The ensemble learning (ensemble learning) method fuses information of a plurality of single models together through an ensemble strategy to construct an ensemble learning model, so that the diversity of the model can be increased, overfitting and prediction uncertainty can be reduced, and a more accurate and more stable prediction result can be generated. Ensemble learning generally comprises three strategies, namely Bagging, boosting and Stacking, wherein a Stacking algorithm can flexibly combine a plurality of different types of base learners according to actual problems to construct a proper ensemble learning model, so that better prediction performance and model generalization capability are obtained, and the ensemble learning method is widely applied to various fields for processing prediction problems.
In addition, the selection of the model parameters is considered to have a significant influence on the final prediction effect of the model, and how to determine the optimal parameters of the model is also the core problem of model establishment. The group intelligent algorithm is simple to operate, has strong problem solving capability and is outstanding in the field of parameter optimization, and is widely applied to parameter optimization of machine learning algorithms. The Sparrow Search Algorithm (SSA) is a novel swarm intelligence optimization algorithm proposed by Xue and the like in 2020, and experiments prove that the algorithm is superior to the current mainstream algorithms such as a Grey wolf optimization algorithm (GWOO), a particle swarm optimization algorithm (PSO), a Gravity Search Algorithm (GSA), an ant lion optimization Algorithm (ALO) and the like in the aspects of search accuracy, convergence speed, stability and the like. However, the SSA has a problem that the population diversity is gradually reduced in the later stage of the search and the algorithm cannot jump out of the local extremum, so that an optimization algorithm capable of considering the global search capability, the local search capability and the calculation efficiency is urgently needed.
In summary, in the research of the existing dam foundation grouting amount prediction model, the numerical simulation has the problems of complex modeling process, large calculated amount, long time consumption and the like, and only one machine learning model is used, so that the uncertainty of prediction is easily underestimated, and the accuracy is poor.
Disclosure of Invention
The invention provides a grouting quantity integrated agent prediction model and a prediction method based on Stacking for solving the technical problems in the prior art.
The technical scheme adopted by the invention for solving the technical problems in the prior art is as follows: a grouting amount integrated proxy prediction model based on Stacking comprises an integrated proxy model, wherein the integrated proxy model is provided with two layers, the first layer comprises three base learners which are trained and verified by adopting a five-fold cross verification method, and the second layer comprises a meta-learner; the three base learners are respectively an SVR neural network, a BPNN neural network and an RF model; the meta learner is an ANFIS neural network; the training data in the training set of the meta-learner includes predicted outcome data for the three base learners.
Further, the grouting numerical simulation model based on three-dimensional fine fracture modeling is further included, geological parameters, construction parameters and slurry characteristic parameters are input into the grouting numerical simulation model, and a grouting quantity simulation value is output; the training data in the training set of the meta-learner also includes a grout amount analog value.
The invention also provides a prediction method of the grouting amount integrated agent prediction model based on the Stacking, which comprises the steps of constructing the grouting amount integrated agent prediction model based on the Stacking; the integrated agent prediction model is provided with two layers, wherein the first layer is provided with three base learners, the second layer is provided with a meta-learner, and the three base learners are respectively an SVR neural network, a BPNN neural network and an RF model; the element learning device is an ANFIS neural network; collecting historical data including geological parameters, construction parameters and slurry characteristic parameters as training samples, and constructing a training set and a verification set; training three base learners by using a five-fold cross validation method; constructing a training set including prediction result data obtained by the three base learners to train the meta-learner; inputting input parameters including geological parameters, construction parameters and slurry characteristic parameters into the three base learners at the same time, inputting prediction results obtained by the three base learners into the meta-learner, and outputting grouting amount prediction values by the meta-learner.
Further, a grouting numerical simulation model based on three-dimensional fine crack modeling is also set, so that the grouting numerical simulation model inputs geological parameters, construction parameters and slurry characteristic parameters and outputs a grouting quantity simulation value; and combining the prediction results obtained by the three base learners with the corresponding grouting quantity analog values to construct a new training set to train the meta-learner.
Further, the method for constructing the training set and the verification set comprises the following steps: extracting multiple groups of geological parameters by utilizing a Latin hypercube sampling method, combining the geological parameters with different construction parameters and slurry characteristic parameters, and constructing a parameter sample set representing various fracture geological conditions and construction working conditions; inputting the data in the parameter sample set into a grouting numerical simulation model to obtain a grouting quantity simulation value; and taking the geological parameters, the construction parameters, the slurry characteristic parameters and the corresponding generated grouting quantity simulation values as sample data, constructing a sample set, and dividing the sample set into a training set and a testing set according to a proportion.
Further, the method for collecting historical data comprising geological parameters, construction parameters and slurry characteristic parameters as training samples comprises the following steps: obtaining the following geological parameters according to the three-dimensional fine fracture network model: the number of cracks, the average tendency of the cracks, the average inclination angle and the average crack width; obtaining the following construction condition parameters according to the actual construction scheme of the project and the cement grouting construction technical specification of the hydraulic structure: sequencing, hole order, hole depth and grouting pressure of grouting holes; the following slurry parameters are obtained according to the actual construction scheme of the project: slurry water-to-cement ratio.
Further, a sparrow search algorithm is improved based on the chaos theory and the Levy flight strategy, and the improved sparrow search algorithm is adopted to synchronously optimize model parameters of the base learner and the meta learner.
Further, the method for improving the sparrow search algorithm based on the chaos theory and the Levy flight strategy comprises the following steps:
initializing sparrow population based on chaos theory, and initializing sparrow position xp by using Tent chaotic mapping to generate chaotic sequence i,j Wherein i =1,2,3 \8230n, n represents the number of sparrow populations, j =1,2,3 \8230d, d represents the number of sparrows to be treatedOptimizing the dimensionality of the variables;
random generation of [0,1]Initial value x in between pi,0 When j =0;
generating a chaotic sequence using Tent chaotic mapping:
mapping the chaotic sequence to a search space of a solution to obtain a chaotic initialization population:
the position updating formulas of the explorer, the follower and the scout are improved by adopting a Levy flight strategy, the searching range is expanded, the global searching capability is improved, and the improved explorer position updating formulas are as follows:
the improved seeker location update formula is as follows:
the improved seeker location update formula is as follows:
in the above formulas:
Y={y i i =1,2,3 \ 8230n } represents responsive grout values;
F={f i i =1,2,3 \8230, k represents the final grouting amount predicted value;
xp represents a population vector set of the parameter to be optimized;
i represents an individual in a sparrow population;
j represents a variable to be optimized;
xp min,j is the minimum value of the vector value of the j-dimension population;
xp max,j is the maximum value of the jth dimension population vector value;
representing the jth dimension population vector value of the ith sparrow individual at the tth iteration;
expressed as the jth dimension population vector value of the ith sparrow individual at the t +1 iteration;
t represents the current iteration number;
iter max representing the maximum number of iterations;
α ∈ (0, 1) represents a random coefficient;
R 2 ∈[0,1]representing an early warning value;
ST ∈ [0.5,1] represents a safety value;
≧ represents the dot product;
l vy (lambda) represents a L vy random search path;
n is the total number of individuals in the sparrow population,
a represents a 1 × d matrix in which each element takes on the value 1 or-1, and A + =A T (AA T ) -1 ;
f b Representing a current global optimal fitness value;
f w representing a current global worst fitness value;
f i representing the fitness value of the sparrow individual to be updated;
ε is a small constant that prevents the denominator from being zero.
Further, the method for synchronously optimizing the model parameters of the base learner and the meta learner by adopting the improved sparrow search algorithm comprises the following steps of:
step 1, determining the population number, the maximum iteration number, the adaptability value range, the solution interval range, the seeker proportion and the reconnaissance proportion;
step 2, initializing a population by Tent chaotic mapping;
step 3, adding 1 to the iteration times; calculating the fitness value of the sparrow population;
step 4, selecting a part from sparrows with better fitness values as an explorer, and updating the positions;
step 5, updating the positions of the remaining sparrows serving as followers;
step 6, selecting a part from the sparrow population as a reconnaissance person to update the position;
step 7, calculating the fitness value of the sparrow population;
step 8, judging whether the fitness value meets the condition, and if not, executing step 9; if yes, executing step 10;
step 9, judging whether the iteration times are less than the maximum iteration times, if so, executing the step 3, otherwise, executing the step 10;
and step 10, finishing the optimization to obtain the optimal model parameters.
Further, the model parameters optimized by the sparrow search algorithm include: penalty factors and kernel parameters of the SVR neural network, initial threshold and weight of the BPNN neural network, decision tree number n and maximum depth h of the RF model, and front-part parameters of the ANFIS neural network.
The invention has the advantages and positive effects that: according to the invention, the grouting quantity integrated agent prediction model is built by adopting the Stacking-based integrated agent prediction model, so that the model diversity can be increased, the overfitting and prediction uncertainty can be reduced, and a more accurate and more stable prediction result can be generated. The improved sparrow search algorithm is adopted to optimize the model parameters, so that the uniformity and diversity of population distribution of the initialized model parameters can be ensured, the problem that the population diversity is reduced and the local optimization is avoided in the search process can be solved. The invention solves the following technical problems of the existing machine learning grouting quantity prediction model: geological parameters are not considered, grouting numerical simulation calculation is complex and time-consuming, the precision of a single agent model is low, and the weighting subjectivity of a combined agent model is high. Compared with a single agent model, the method has the advantages that the prediction performance is improved, and the grouting amount can be rapidly and accurately predicted, so that a more accurate and reliable prediction result is obtained, guidance is provided for decision making, and the safety and the quality of grouting are ensured; the method can provide reliable method support for the grouting amount estimation of the area to be grouted in the actual grouting engineering, and has important engineering application value; meanwhile, the model provided by the invention also provides a new idea for predicting other engineering parameters, and has a good engineering application prospect.
Drawings
FIG. 1 is a schematic structural diagram of a grouting quantity integrated agent prediction model based on Stacking.
FIG. 2 is a schematic diagram of a grouting quantity integrated agent prediction model training process based on Stacking.
FIG. 3 is a flow chart of a prediction method of a grouting quantity integrated proxy prediction model based on Stacking according to the invention.
In the figure:
SVR: vector regression is supported.
BPNN: BP neural network model.
RF: and (4) random forest model.
ANFIS: an adaptive neuro-fuzzy inference system.
X: and (5) deciding a variable set.
Y: and (5) response quantity collection.
X tr : the decision variables in the sample are trained.
Y tr : the amount of response in the training samples.
X te : the decision variables in the sample are tested.
Y te : the amount of response in the sample is tested.
P tr-SVR : and obtaining a response quantity predicted value based on the training sample by using a support vector machine model.
P tr-BPNN : and obtaining a response quantity predicted value based on the training sample by using a BP neural network model.
P tr-RF : and obtaining a response quantity predicted value based on the training sample by utilizing a random forest model.
P tr-1~5 : and training the response quantity predicted values obtained by the base learner five times based on the training samples.
P tr : and (4) assembling new training samples formed by the outputs of the three base learners to the training samples.
P te-SVR : and obtaining a response quantity predicted value based on the test sample by using a support vector machine model.
P te-BPNN : and obtaining a response quantity predicted value based on the test sample by using the BP neural network model.
P te-RF Using stochastic SensorsThe forest model is based on the response prediction value obtained by the test sample.
P te : and (4) integrating the outputs of the three base learners on the test samples to form a new test sample book.
F te : and finally predicting the result.
Detailed Description
For further understanding of the contents, features and effects of the present invention, the following embodiments are enumerated in conjunction with the accompanying drawings, and the following detailed description is given:
the Chinese definitions of the following foreign words and abbreviations in this application are as follows:
and (4) Stacking: one type of ensemble learning strategy is to combine different types of basis learners to construct an ensemble model by training a meta-learner to learn the relationship between the model output and the actual output of each basis learner.
Bagging: one type of ensemble learning strategy is to generate multiple base learners of the same type in parallel through sample training and to linearly combine their results.
Boosting: one type of ensemble learning strategy is to generate multiple base learners of the same type in a serial fashion through sample training and linearly combine their results.
SSA: a sparrow search algorithm, a novel group intelligent optimization algorithm.
ISSA-Stacking: an integrated learning algorithm based on an improved sparrow search algorithm.
SVR: the support vector regression is based on a prediction model of a hyperplane, and the support vector regression has unique advantages for solving the problem of nonlinear regression prediction of small samples and high latitude.
BPNN: the BP neural network model and the classical neural network model have better numerical processing and approximation capability.
RF: the random forest model and the representation of the Bagging integration algorithm have the advantages of low generalization error, good stability, difficulty in overfitting and the like.
ANFIS: an adaptive neuro-fuzzy inference system.
LHS: latin hypercube sampling method.
L vy: the levy flight strategy is a type of non-gaussian random process.
Tent: tent chaotic mapping, which refers to a piecewise linear mapping.
Referring to fig. 1 to 3, an integrated proxy prediction model for grouting amount based on Stacking includes an integrated proxy model, which has two layers, the first layer includes three base learners trained and verified by a five-fold cross-validation method, and the second layer includes a meta-learner; the three base learners are respectively an SVR neural network, a BPNN neural network and an RF model; the element learning device is an ANFIS neural network; the training data in the training set of the meta learner includes predicted outcome data for the three base learners.
Preferably, the grouting device also comprises a grouting numerical simulation model based on three-dimensional fine fracture modeling, and the grouting numerical simulation model can input geological parameters, construction parameters and slurry characteristic parameters and output a grouting quantity simulation value; the training data in the training set of the meta-learner may also include a grout amount simulation value.
The invention also provides a prediction method of the grouting amount integrated proxy prediction model based on the Stacking, which is used for constructing the grouting amount integrated proxy prediction model based on the Stacking; the integrated agent prediction model is provided with two layers, wherein the first layer is provided with three base learners, the second layer is provided with a meta-learner, and the three base learners are respectively an SVR neural network, a BPNN neural network and an RF model; the element learning device is an ANFIS neural network; collecting historical data including geological parameters, construction parameters and slurry characteristic parameters as training samples, and constructing a training set and a verification set; training three base learners by using a five-fold cross validation method; constructing a training set including prediction result data obtained by the three base learners to train the meta-learner; inputting input parameters including geological parameters, construction parameters and slurry characteristic parameters into the three base learners at the same time, inputting prediction results obtained by the three base learners into the meta-learner, and outputting grouting amount prediction values by the meta-learner.
Preferably, a grouting numerical simulation model based on three-dimensional fine crack modeling can be further arranged, so that geological parameters, construction parameters and slurry characteristic parameters can be input into the grouting numerical simulation model, and a grouting quantity simulation value can be output; the prediction results obtained by the three base learners can be combined with the corresponding grouting quantity analog values to construct a new training set to train the meta-learner.
Preferably, the method for constructing the training set and the verification set may include: a plurality of groups of geological parameters can be extracted by utilizing a Latin hypercube sampling method, and are combined with different construction parameters and slurry characteristic parameters to construct a parameter sample set representing various fracture geological conditions and construction working conditions; the data in the parameter sample set can be input into a grouting numerical simulation model to obtain a grouting quantity simulation value; the geological parameters, the construction parameters, the slurry characteristic parameters and the corresponding generated grouting amount simulation values can be used as sample data to construct a sample set, and the sample set can be divided into a training set and a testing set according to the proportion.
Preferably, the method of collecting historical data including geological parameters, construction parameters and grout characteristic parameters as training samples may include: the following geological parameters can be obtained according to the three-dimensional fine fracture network model: the number of cracks, the average tendency of the cracks, the average inclination angle and the average crack width; the following construction condition parameters can be obtained according to the actual construction scheme of the project and the cement grouting construction technical specification of the hydraulic structure: sequencing, hole order, hole depth and grouting pressure of grouting holes; the following slurry parameters can be obtained according to the actual construction scheme of the project: water-to-ash ratio of the slurry.
Preferably, the sparrow search algorithm can be improved based on the chaos theory and the Levy flight strategy, and the improved sparrow search algorithm can be adopted to synchronously optimize the model parameters of the base learner and the meta learner.
Preferably, the method for improving the sparrow search algorithm based on the chaos theory and the Levy flight strategy may include:
the sparrow population initialization can be carried out based on the chaos theory, and the Tent chaos mapping can be used for generating a chaos sequence to initialize the sparrow position xp i,j Wherein i =1,2,3 \8230n, n denotes the number of sparrow populations, j =1,2,3 \8230d, d denotes the dimension of the variable to be optimized.
Can be randomly generatedTo [0, 1]]Initial value x in between pi,0 When j =0.
The chaotic sequence may be generated using Tent chaotic mapping:
the chaotic sequence can be mapped to a search space of a solution to obtain a chaotic initialization population:
the position updating formulas of the seeker, the follower and the reconnaissance can be improved by adopting a Levy flight strategy, the searching range is expanded, the global searching capability is improved, and the improved seeker position updating formula can be as follows:
the improved seeker location update formula may be as follows:
the improved seeker location update formula may be as follows:
in the above formulas:
Y={y i i =1,2,3 \ 8230n represents responsive grout values.
F={f i I =1,2,3 \8230;, k } denotes the final grouting amount pre-groutingAnd (6) measuring.
xp represents a population vector set of the parameter to be optimized.
i represents an individual in a sparrow population.
j denotes the variable to be optimized.
xp min,j Is the minimum value of the vector value of the j-dimension population.
xp max,j Is the maximum value of the vector value of the j-th dimension population.
Expressed as the jth dimension population vector value of the ith sparrow individual at the tth iteration.
Expressed as the jth dimension population vector value of the ith sparrow individual at the t +1 iteration.
t represents the current number of iterations.
iter max The maximum number of iterations is indicated.
α ∈ (0, 1) represents a random coefficient.
R 2 ∈[0,1]Indicating an early warning value.
ST ∈ [0.5,1] indicates a security value.
≧ represents the dot product.
L vy (lambda) represents a L vy random search path.
n is the total number of individuals in the sparrow population,
a represents a 1 × d matrix in which each element takes on the value 1 or-1, and A + =A T (AA T ) -1 。
f b Representing the current global best fitness value.
f w Representing the current global worst fitness value.
f i Representing the fitness value of the sparrow individual to be updated.
ε is a small constant that prevents the denominator from being zero.
Preferably, the method for synchronously optimizing the model parameters of the base learner and the meta learner by adopting the improved sparrow search algorithm comprises the following steps:
step 1, the population quantity, the maximum iteration times, the adaptability value range, the solution interval range, the seeker ratio and the reconnaissance ratio can be determined.
And 2, initializing the population by Tent chaotic mapping.
And step 3, adding 1 to the iteration times. And calculating the fitness value of the sparrow population.
And 4, selecting a part from sparrows with better fitness values as an explorer to update the positions.
And 5, the remaining sparrows can be used as followers to update the positions.
And 6, selecting a part from the sparrow population as a reconnaissance person to update the position.
And 7, calculating the adaptability value of the sparrow population.
Step 8, judging whether the fitness value meets the condition, and if not, executing step 9; if so, step 10 is performed.
And 9, judging whether the iteration times are less than the maximum iteration times, if so, executing the step 3, and otherwise, executing the step 10.
And step 10, finishing the optimization to obtain the optimal model parameters.
Preferably, the model parameters optimized by the sparrow search algorithm may include: penalty factors and kernel parameters of the SVR neural network, initial threshold and weight of the BPNN neural network, the number n and maximum depth h of decision trees of the RF model, and the front-part parameters of the ANFIS neural network.
The working process and working principle of the present invention are further explained by a preferred embodiment of the present invention as follows:
the invention adopts the Stacking integrated learning strategy which can increase the diversity of the model, reduce overfitting and prediction uncertainty and generate more accurate and more stable prediction results, can ensure the uniformity and diversity of the initialized population distribution, and can overcome the problem that the population diversity is reduced and avoid falling into the local optimum in the searching process. The invention solves the problems that the existing machine learning grouting amount prediction model does not consider geological parameters, grouting numerical simulation calculation is complex and time-consuming, the precision of a single agent model is low, and the empowerment subjectivity of a combined agent model is large. Compared with a single agent model, the prediction performance of the method is greatly improved, and the rapid and accurate grouting amount prediction can be realized, so that a more accurate and reliable prediction result is obtained, guidance is provided for decision making, and the safety and quality of grouting are ensured; the method can provide reliable method support for the grouting amount estimation of the area to be grouted in the actual grouting engineering, and has important engineering application value; meanwhile, the model provided by the text also provides a new idea for the prediction of other engineering parameters, and has a good engineering application prospect.
In actual engineering, accurate and reliable grouting amount prediction has important significance for grouting construction process control. Aiming at the defects that a numerical simulation method is complex and time-consuming, a single machine learning method-based proxy model is low in precision and the like in the conventional grouting amount prediction research, an ISSA-Stacking-based dam foundation grouting amount prediction integrated proxy model is adopted, and accurate prediction of grouting amounts under various geological conditions and grouting working conditions is rapidly realized, so that effective and reliable theoretical guidance is provided for subsequent grouting construction control and engineering amount optimization. The dam foundation grouting quantity prediction integrated agent model based on ISSA-Stacking specifically comprises the following steps:
A. and acquiring input parameters including three factors of geological properties, construction conditions and slurry properties.
B. A data set is generated.
C. And constructing an integrated agent model based on Stacking.
D. And optimizing the Stacking-based integrated proxy model parameters by using the improved SSA, and establishing the Stacking-based integrated proxy model.
In order to make the objects, technical solutions and advantages of the present invention more apparent, embodiments of the present invention are described in detail with reference to the accompanying drawings.
1. Obtaining input parameters including three factors of geological property, construction condition and slurry property
Acquiring fractured rock parameters such as the number of fractures, the average tendency of the fractures, the average inclination angle, the average gap width and the like according to the three-dimensional fine fracture network model; determining the sequencing, hole sequence, hole depth and grouting pressure of grouting holes according to actual engineering construction measures and the technical Specification for hydraulic structure cement grouting construction (DL/T5148-2012); and determining the water-cement ratio of the slurry according to actual construction measures of the project.
2. Generating a data set
Based on a fractured rock mass parameter sample space, a plurality of groups of geological parameters are extracted by utilizing a Latin Hypercube (LHS) sampling method, and are combined with different construction parameters and slurry characteristic parameters to construct parameter sample points representing various fractured geological conditions and construction working conditions. And substituting the parameter sample points into a grouting numerical simulation model based on three-dimensional fine fracture modeling to calculate and obtain grouting quantity simulation values, generating a data set consisting of the parameter sample points and the corresponding grouting quantity simulation values, and further dividing the data set into a training set and a test set according to the proportion.
3. Building Stacking-based integrated proxy model
And training the three basis learners of SVR, BPNN and RF by using the obtained data training set through five-fold cross validation, and improving the overall generalization and diversity of the model. And (3) forming a new training set by the obtained prediction result and the simulation response value to train the ANFIS meta-learner, and realizing induction and fusion of the base learner result under the condition of considering uncertainty of the prediction process, wherein the flow of the method is shown in FIG. 1. The method comprises the following specific steps:
3-1, dividing a sample data set (X, Y) consisting of input data and grouting amount analog values into training sets (X) in proportion tr ,Y tr ) And test set (X) te ,Y te );
3-2, training the SVR, the BPNN and the RF base learner by adopting five-fold cross validation, and obtaining a prediction result p tr-j Repeating the process five times to obtain the prediction result P of each base learner on the whole original training set tr-SVR ={p tr-j ,j=1,2,3,4,5}、P tr-BPNN ={p tr-j J =1,2,3,4,5} and P tr-RF ={p tr-j J =1,2,3,4,5}. Constructing a new training set (P) with the response values in the original training set tr ,Y tr ) For training the Meta-learner ANFIS, where P tr ={P tr-SVR ,P tr-BPNN ,P tr-RF };
3-3, in the testing process, each trained base learning device respectively obtains the corresponding prediction result of the original testing set and constructs a new testing set (P) with the response value in the original testing set te ,Y te ) For testing the trained Meta-learner ANFIS, where P te ={P te -SVR,P te -BPANN,P te -RF }. Thereby obtaining the final predicted result F te ={f i ,i=1,2,3…,k}。
And 3-4, optimizing the Stacking-based integrated proxy model parameters by using the improved SSA, and establishing a Stacking-based grouting amount integrated proxy prediction model.
And performing synchronous optimization on model parameters of the base learner and the meta learner by adopting a sparrow search algorithm improved by a chaos theory and a Levy flight strategy, and further establishing a packing-based grouting amount integrated proxy prediction model to realize high-precision prediction of grouting amount. The method comprises the following specific steps:
3-4-1, initializing sparrow population by adopting a chaos theory, and initializing a sparrow position x by using a Tent chaos mapping to generate a chaos sequence pi,j Wherein i =1,2,3 \8230n, n represents the number of sparrow populations, and j =1,2,3 \8230d, d represents the dimension of the variable to be optimized.
a. Random generation of [0,1]Initial value x in between pi,0 When j =0.
b. Generating a chaotic sequence using Tent chaotic mapping:
c. mapping the chaotic sequence to a search space of a solution to obtain a chaotic initialization population:
in the formula, xp min,j ,xp max,j Respectively, the minimum and maximum values of the j-th dimension.
3-4-2, improving the position updating formulas of the seeker, the follower and the reconnaissance by adopting a Levy flight strategy, enlarging the searching range and improving the global searching capability, wherein the improved seeker position updating formula is as follows:
where t denotes the current iteration number, iter max Represents the maximum number of iterations, α ∈ (0, 1) is a random number, R 2 ∈[0,1]Represents an early warning value, ST ∈ [0.5,1]Representing a security value.Is the current global optimum position, # is the point product, and L vy (λ) is the L vy random search path.
The improved seeker location update formula is as follows:
in the formulaIndicating the optimal position to be occupied by the seeker at the time,is the current global worst position, A represents a 1 × d matrix, where each element takes the value of 1 or-1, and A + =A T (AA T ) -1 。
The improved seeker location update formula is as follows:
in the formula (f) b And f w Respectively, the current global best fitness value and the worst fitness value. f. of i Representing the fitness value of the sparrow individual to be updated; ε is a small constant that acts to prevent the denominator from being zero.
3-4-2, performing parameter synchronization optimization on each machine learning algorithm in the Stacking-based integrated agent model by using the ISSA with more excellent global search capability and convergence performance.
The predictive performance of the Stacking-based integrated proxy model is affected by the following parameters: the method mainly comprises a penalty factor C and a kernel parameter g of SVR, an initial threshold b and a weight w of BPNN, the number n and the maximum depth h of a decision tree of RF and the front-part parameters ai, bi and ci of ANFIS. The ISSA-based parameter search process can be translated into the following optimization problem:
wherein Y = { yi, i =1,2,3 \8230;, n } and F = { fi, i =1,2,3 \8230; n } respectively represent response grout magnitude and final grout magnitude predicted values, xp represents a population vector set of parameters to be optimized, and lb and xp ub Respectively the lower limit and the upper limit of the value of the parameter to be optimized. Therefore, an integrated agent model based on ISSA-Stacking is constructed.
The above-mentioned embodiments are only for illustrating the technical ideas and features of the present invention, and the purpose thereof is to enable those skilled in the art to understand the contents of the present invention and to carry out the same, and the present invention shall not be limited to the embodiments, i.e. the equivalent changes or modifications made within the spirit of the present invention shall fall within the scope of the present invention.
Claims (10)
1. A grouting amount integrated proxy prediction model based on Stacking is characterized by comprising an integrated proxy model, wherein the integrated proxy model is provided with two layers, the first layer comprises three base learners which are trained and verified by adopting a five-fold cross verification method, and the second layer comprises a meta-learner; the three base learners are respectively an SVR neural network, a BPNN neural network and an RF model; the meta learner is an ANFIS neural network; the training data in the training set of the meta-learner includes predicted outcome data for the three base learners.
2. The Stacking-based grouting quantity integrated proxy prediction model according to claim 1, further comprising a grouting numerical simulation model based on three-dimensional fine fracture modeling, wherein the grouting numerical simulation model inputs geological parameters, construction parameters and slurry characteristic parameters and outputs a grouting quantity simulation value; the training data in the training set of the meta learner also includes a grout amount simulation value.
3. A prediction method of a grouting amount integrated proxy prediction model based on Stacking is characterized in that the grouting amount integrated proxy prediction model based on Stacking is constructed; the integrated agent prediction model is provided with two layers, wherein the first layer is provided with three base learners, the second layer is provided with a meta-learner, and the three base learners are respectively an SVR neural network, a BPNN neural network and an RF model; the meta learner is an ANFIS neural network; collecting historical data including geological parameters, construction parameters and slurry characteristic parameters as training samples, and constructing a training set and a verification set; training three base learners by using a five-fold cross validation method; constructing a training set including prediction result data obtained by the three base learners to train the meta-learner; and simultaneously inputting input parameters including geological parameters, construction parameters and slurry characteristic parameters into the three base learners, inputting prediction results obtained by the three base learners into the meta learner, and outputting a grouting amount prediction value by the meta learner.
4. The prediction method of the Stacking-based grouting quantity integrated proxy prediction model according to claim 3, characterized in that a grouting numerical simulation model based on three-dimensional fine fracture modeling is further provided, so that the grouting numerical simulation model inputs geological parameters, construction parameters and slurry characteristic parameters and outputs a grouting quantity simulation value; and combining the prediction results obtained by the three base learners with the corresponding grouting quantity analog values to construct a new training set to train the meta-learner.
5. The forecasting method of the grouting quantity integrated agent forecasting model based on Stacking according to claim 4, wherein the method for constructing the training set and the verification set comprises the following steps: extracting multiple groups of geological parameters by utilizing a Latin hypercube sampling method, combining the geological parameters with different construction parameters and slurry characteristic parameters, and constructing a parameter sample set representing various fracture geological conditions and construction working conditions; inputting the data in the parameter sample set into a grouting numerical simulation model to obtain a grouting quantity simulation value; and taking the geological parameters, the construction parameters, the slurry characteristic parameters and the corresponding generated grouting amount simulation values as sample data, constructing a sample set, and dividing the sample set into a training set and a testing set according to a proportion.
6. The prediction method of the Stacking-based grouting quantity integrated proxy prediction model according to claim 3, wherein the method for collecting historical data including geological parameters, construction parameters and slurry characteristic parameters as training samples comprises the following steps: obtaining the following geological parameters according to the three-dimensional fine fracture network model: the number of cracks, the average tendency of the cracks, the average inclination angle and the average crack width; obtaining the following construction condition parameters according to the actual construction scheme of the project and the cement grouting construction technical specification of the hydraulic structure: sequencing grouting holes, hole sequence, hole depth and grouting pressure; the following slurry parameters are obtained according to the actual construction scheme of the project: slurry water-to-cement ratio.
7. The prediction method of the Stacking-based grouting quantity integrated proxy prediction model according to claim 3, characterized in that a sparrow search algorithm is improved based on a chaos theory and a Levy flight strategy, and model parameters of a base learner and a meta learner are synchronously optimized by adopting the improved sparrow search algorithm.
8. The prediction method of the Stacking-based grouting quantity integrated proxy prediction model according to claim 7, wherein the method for improving the sparrow search algorithm based on the chaos theory and the Levy flight strategy comprises the following steps:
initializing sparrow population based on chaos theory, and initializing sparrow position xp by using Tent chaos mapping to generate chaos sequence i,j Wherein i =1,2,3 \8230n, n represents the number of sparrow populations, j =1,2,3 \8230d, d represents the dimension of the variable to be optimized;
random generation of [0,1]Initial value x in between pi,0 When j =0;
generating a chaotic sequence using Tent chaotic mapping:
mapping the chaotic sequence to a search space of a solution to obtain a chaotic initialization population:
the position updating formulas of the explorer, the follower and the scout are improved by adopting a Levy flight strategy, the searching range is expanded, the global searching capability is improved, and the improved explorer position updating formulas are as follows:
the improved seeker location update formula is as follows:
the improved seeker location update formula is as follows:
in the above formulas:
Y={y i i =1,2,3 \ 8230n represents responsive grout magnitude;
F={f i i =1,2,3 \8230, k represents the final grouting amount predicted value;
xp represents a population vector set of the parameter to be optimized;
i represents an individual in a sparrow population;
j represents a variable to be optimized;
xp min,j is the minimum value of the jth dimension population vector value;
xp max,j is the maximum value of the vector value of the j-dimension population;
expressed as the jth dimension population vector value of the ith sparrow individual at the tth iteration;
expressed as the jth dimension population vector value of the ith sparrow individual at the t +1 iteration;
t represents the current iteration number;
iter max representing the maximum number of iterations;
α ∈ (0, 1) denotes a random coefficient;
R 2 ∈[0,1]representing an early warning value;
ST ∈ [0.5,1] represents a safety value;
≧ represents the dot product;
levy (lambda) represents a Levy random search path;
n is the total number of individuals in the sparrow population,
a represents a 1 × d matrix in which each element takes on the value 1 or-1, and A + =A T (AA T ) -1 ;
f b Representing a current global optimal fitness value;
f w representing a current global worst fitness value;
f i representing the fitness value of the sparrow individual to be updated;
ε is a small constant that prevents the denominator from being zero.
9. The forecasting method of the padding-based grouting quantity integrated agent forecasting model according to claim 8, characterized in that the method for synchronously optimizing the model parameters of the base learner and the meta learner by adopting the improved sparrow search algorithm comprises the following steps:
step 1, determining the population number, the maximum iteration number, the adaptability value range, the solution interval range, the seeker proportion and the reconnaissance proportion;
step 2, initializing a population by Tent chaotic mapping;
step 3, adding 1 to the iteration times; calculating a sparrow population fitness value;
step 4, selecting a part from sparrows with better fitness values as an explorer, and updating the positions;
step 5, the position of the remaining sparrows is updated as followers;
step 6, selecting a part from the sparrow population as a scout, and updating the position;
step 7, calculating the fitness value of the sparrow population;
step 8, judging whether the fitness value meets the condition, and if not, executing step 9; if yes, executing step 10;
step 9, judging whether the iteration times are less than the maximum iteration times, if so, executing the step 3, otherwise, executing the step 10;
and step 10, finishing the optimization to obtain the optimal model parameters.
10. The forecasting method of the grouting quantity integrated proxy forecasting model based on packing according to claim 7, characterized in that the model parameters optimized by adopting a sparrow search algorithm comprise: penalty factors and kernel parameters of the SVR neural network, initial threshold and weight of the BPNN neural network, the number n and maximum depth h of decision trees of the RF model, and the front-part parameters of the ANFIS neural network.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210798033.XA CN115310348A (en) | 2022-07-08 | 2022-07-08 | Stacking-based grouting amount integrated agent prediction model and prediction method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210798033.XA CN115310348A (en) | 2022-07-08 | 2022-07-08 | Stacking-based grouting amount integrated agent prediction model and prediction method |
Publications (1)
Publication Number | Publication Date |
---|---|
CN115310348A true CN115310348A (en) | 2022-11-08 |
Family
ID=83856439
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202210798033.XA Pending CN115310348A (en) | 2022-07-08 | 2022-07-08 | Stacking-based grouting amount integrated agent prediction model and prediction method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN115310348A (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117215728A (en) * | 2023-11-06 | 2023-12-12 | 之江实验室 | Agent model-based simulation method and device and electronic equipment |
CN117455890A (en) * | 2023-11-20 | 2024-01-26 | 浙江大学 | Child intussusception air enema result prediction device based on improved integrated deep learning |
CN117455890B (en) * | 2023-11-20 | 2024-05-31 | 浙江大学 | Child intussusception air enema result prediction device based on improved integrated deep learning |
-
2022
- 2022-07-08 CN CN202210798033.XA patent/CN115310348A/en active Pending
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117215728A (en) * | 2023-11-06 | 2023-12-12 | 之江实验室 | Agent model-based simulation method and device and electronic equipment |
CN117215728B (en) * | 2023-11-06 | 2024-03-15 | 之江实验室 | Agent model-based simulation method and device and electronic equipment |
CN117455890A (en) * | 2023-11-20 | 2024-01-26 | 浙江大学 | Child intussusception air enema result prediction device based on improved integrated deep learning |
CN117455890B (en) * | 2023-11-20 | 2024-05-31 | 浙江大学 | Child intussusception air enema result prediction device based on improved integrated deep learning |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109492822B (en) | Air pollutant concentration time-space domain correlation prediction method | |
Kim et al. | Neural network model incorporating a genetic algorithm in estimating construction costs | |
Chen et al. | Deformation prediction of landslide based on improved back-propagation neural network | |
CN109493599A (en) | A kind of Short-time Traffic Flow Forecasting Methods based on production confrontation network | |
CN104965969B (en) | A kind of Large-scale Underground Tunnels And Chambers rock reaction force inversion method | |
CN112733997B (en) | Hydrological time series prediction optimization method based on WOA-LSTM-MC | |
CN106529818A (en) | Water quality evaluation prediction method based on fuzzy wavelet neural network | |
CN109635245A (en) | A kind of robust width learning system | |
CN105335619A (en) | Collaborative optimization method applicable to parameter back analysis of high calculation cost numerical calculation model | |
Ning et al. | GA-BP air quality evaluation method based on fuzzy theory. | |
CN112307536B (en) | Dam seepage parameter inversion method | |
CN108805346A (en) | A kind of hot continuous rolling force forecasting method based on more hidden layer extreme learning machines | |
CN112765902A (en) | RBF neural network soft measurement modeling method based on TentFWA-GD and application thereof | |
Zhang et al. | Support vector machine weather prediction technology based on the improved quantum optimization algorithm | |
CN115310348A (en) | Stacking-based grouting amount integrated agent prediction model and prediction method | |
CN115948964A (en) | Road flatness prediction method based on GA-BP neural network | |
CN109408896B (en) | Multi-element intelligent real-time monitoring method for anaerobic sewage treatment gas production | |
Zhu et al. | Structural safety monitoring of high arch dam using improved ABC‐BP model | |
CN109493921A (en) | A kind of atmospheric distillation process modeling approach based on multi-agent system model | |
CN112183721B (en) | Construction method of combined hydrological prediction model based on self-adaptive differential evolution | |
CN108256688B (en) | Steam flooding development effect prediction method for optimizing GRNN by improving firefly algorithm | |
CN117371616A (en) | Soil frost heaving rate prediction method for optimizing generalized regression neural network | |
CN108876038A (en) | Big data, artificial intelligence, the Optimization of Material Property method of supercomputer collaboration | |
CN109116300A (en) | A kind of limit learning position method based on non-abundant finger print information | |
Velasco et al. | Performance analysis of artificial neural networks training algorithms and transfer functions for medium-term water consumption forecasting |
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 |