CN113128108A - Method for determining diameter of jet grouting pile based on differential evolution artificial intelligence - Google Patents
Method for determining diameter of jet grouting pile based on differential evolution artificial intelligence Download PDFInfo
- Publication number
- CN113128108A CN113128108A CN202110373814.XA CN202110373814A CN113128108A CN 113128108 A CN113128108 A CN 113128108A CN 202110373814 A CN202110373814 A CN 202110373814A CN 113128108 A CN113128108 A CN 113128108A
- Authority
- CN
- China
- Prior art keywords
- jet grouting
- grouting pile
- neural network
- differential evolution
- artificial neural
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 56
- 238000013473 artificial intelligence Methods 0.000 title claims abstract description 21
- 238000013528 artificial neural network Methods 0.000 claims abstract description 66
- 238000010276 construction Methods 0.000 claims abstract description 41
- 238000012549 training Methods 0.000 claims abstract description 31
- 238000012360 testing method Methods 0.000 claims abstract description 18
- 230000035772 mutation Effects 0.000 claims abstract description 14
- 238000010606 normalization Methods 0.000 claims abstract description 11
- 238000005457 optimization Methods 0.000 claims abstract description 11
- 238000005215 recombination Methods 0.000 claims abstract description 10
- 230000006798 recombination Effects 0.000 claims abstract description 10
- 238000012545 processing Methods 0.000 claims abstract description 6
- 239000013598 vector Substances 0.000 claims description 45
- 239000002689 soil Substances 0.000 claims description 17
- 230000006870 function Effects 0.000 claims description 15
- 230000008569 process Effects 0.000 claims description 13
- 239000004568 cement Substances 0.000 claims description 9
- 230000002159 abnormal effect Effects 0.000 claims description 6
- 230000015572 biosynthetic process Effects 0.000 claims description 6
- 238000012937 correction Methods 0.000 claims description 6
- 239000011159 matrix material Substances 0.000 claims description 6
- 238000007781 pre-processing Methods 0.000 claims description 6
- 238000005070 sampling Methods 0.000 claims description 6
- 238000004080 punching Methods 0.000 claims description 3
- 230000001537 neural effect Effects 0.000 claims 1
- 238000004364 calculation method Methods 0.000 abstract description 5
- 238000012706 support-vector machine Methods 0.000 description 3
- 230000007547 defect Effects 0.000 description 2
- 238000004836 empirical method Methods 0.000 description 2
- 230000007246 mechanism Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 239000000243 solution Substances 0.000 description 2
- 230000003044 adaptive effect Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000006835 compression Effects 0.000 description 1
- 238000007906 compression Methods 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000002347 injection Methods 0.000 description 1
- 239000007924 injection Substances 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 230000002787 reinforcement Effects 0.000 description 1
- 230000003014 reinforcing effect Effects 0.000 description 1
- 239000002002 slurry Substances 0.000 description 1
- 238000007711 solidification Methods 0.000 description 1
- 230000008023 solidification Effects 0.000 description 1
- 238000012731 temporal analysis Methods 0.000 description 1
- 230000001550 time effect Effects 0.000 description 1
- 238000000700 time series analysis Methods 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
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- 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)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Software Systems (AREA)
- General Engineering & Computer Science (AREA)
- Computational Linguistics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Data Mining & Analysis (AREA)
- Biophysics (AREA)
- Mathematical Physics (AREA)
- Biomedical Technology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Business, Economics & Management (AREA)
- Marketing (AREA)
- Human Resources & Organizations (AREA)
- Economics (AREA)
- Primary Health Care (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Medical Informatics (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- Consolidation Of Soil By Introduction Of Solidifying Substances Into Soil (AREA)
Abstract
The invention discloses a method for determining the diameter of a jet grouting pile based on differential evolution artificial intelligence, which comprises the following steps: collecting and pretreating construction parameters and stratum parameters of the jet grouting pile and the corresponding jet grouting pile diameter; dividing the preprocessed data set into a training set and a testing set, and carrying out normalization processing on the data set; initializing a differential evolution artificial neural network; inputting a training set into the differential evolution artificial neural network, and performing iterative training by using an N-Adam optimization algorithm; carrying out mutation, recombination and selection of a differential evolution artificial neural network to obtain next generation population individuals; repeating the training step until the evolution algebra reaches a threshold value or the accuracy of the training set reaches a requirement, finishing the training, and storing the differential evolution artificial neural network at the moment; and inputting the data set of the jet grouting pile into a differential evolution artificial neural network to obtain the diameter of the jet grouting pile. The method improves the calculation accuracy by using the differential evolution artificial neural network, and realizes the efficient and accurate prediction of the diameter of the jet grouting pile.
Description
Technical Field
The invention relates to the field of constructional engineering, in particular to a method for determining the diameter of a jet grouting pile based on differential evolution artificial intelligence.
Background
The jet grouting pile is one of the most common foundation treatment technologies at present, and is widely applied to underground engineering construction such as subway construction. During construction, cement slurry is injected into a soil layer through a high-pressure rotating nozzle and is mixed with soil, and a cement reinforcing body is formed after solidification. The construction operation of the high-pressure jet grouting pile is relatively simple, so that the high-pressure jet grouting pile is widely applied to projects such as tunnels, foundation pits and the like. However, the pile diameter obtained by construction is different due to the guniting operation in soil layers with different properties. Therefore, how to determine the diameters of the rotary jet piles in different soil layers in the construction process is a key problem in the design and construction of the high-pressure rotary jet piles. At present, methods for determining the diameter of a jet grouting pile can be roughly divided into two types: (1) semi-theoretical and semi-empirical methods; (2) an artificial intelligence method. Due to the complexity of a jet flow damage mechanism and the limitation of time effect of construction of the jet grouting pile, the analytical solution of the diameter of the jet grouting pile cannot be accurately determined by a semi-theoretical semi-empirical method, so that the analysis result is inaccurate. Although the existing artificial intelligence method can directly obtain the nonlinear relation between the diameter of the jet grouting pile and the soil body and construction parameters from jet grouting construction data without considering the specific mechanism and theoretical hypothesis of soil body damage, the method is more flexible, but the problems of low calculation efficiency, no method for predicting time series analysis and large result error exist.
After searching the existing documents, the patent with the patent publication number CN201410452452.3 discloses a method for determining the diameter of a high-pressure jet grouting pile by considering all construction parameters and soil characteristics, which adopts a turbulent jet flow theoretical model and determines the diameter of the high-pressure jet grouting pile by combining construction data and soil characteristics. But the determination of the related construction parameters and the soil body parameters is difficult, so that the calculation result is incorrect. In addition, the currently employed artificial intelligence methods are mainly Artificial Neural Networks (ANN) and Support Vector Machines (SVM). Ochma ń ski is equal to the Prediction of the diameter of jet grouting piles with the actual neural networks published in "Soils and Foundations" 2015, and the Prediction results are good using ANN. However, the method needs to determine the optimal network structure and parameters of the ANN through a plurality of early-stage training trials, and the time cost is high. In addition, the model is prone to overfitting problems resulting in increased errors. Tinoco is equal to Jet grouping diameter prediction section a data-drive adaptive published in European Journal of Environmental and Civil Engineering in 2016, and SVM is adopted to predict the diameter of the Jet grouting pile, so that the accuracy of a prediction result is better, but the selection process is complex and the time cost is also high. Therefore, it is necessary to provide a more effective artificial intelligence method for determining the diameter of the jet grouting pile to improve the efficiency and accuracy of determining the diameter of the pile.
Disclosure of Invention
The invention provides a method for determining the diameter of a jet grouting pile based on differential evolution artificial intelligence, overcomes the defects of difficult determination of construction parameters, low algorithm prediction efficiency, large deviation and the like in the conventional method, and realizes efficient and accurate prediction of the diameter of the jet grouting pile.
In order to solve the technical problems, the technical scheme of the invention is as follows:
a method for determining the diameter of a jet grouting pile based on differential evolution artificial intelligence comprises the following steps:
s1: collecting construction parameters, stratum parameters and corresponding diameters of the jet grouting piles;
s2: preprocessing the collected construction parameters and stratum parameters of the jet grouting pile and the corresponding diameter of the jet grouting pile;
s3: dividing a data set formed by the construction parameters, the formation parameters and the corresponding jet grouting pile diameters of the pretreated jet grouting piles into a training set and a testing set, and carrying out normalization processing on the data set;
s4: initializing a differential evolution artificial neural network;
s5: inputting a training set into the differential evolution artificial neural network, and performing iterative training by using an N-Adam optimization algorithm;
s6: carrying out mutation, recombination and selection of a differential evolution artificial neural network to obtain next generation population individuals;
s7: repeating the steps S5 to S6 until the evolution algebra reaches a threshold value or the accuracy of the training set reaches a requirement, finishing the training, and storing the differential evolution artificial neural network at the moment;
s8: and (4) inputting the jet grouting pile data set into the differential evolution artificial neural network obtained in the step S7 to obtain the diameter of the jet grouting pile.
Preferably, the construction parameters of the jet grouting pile in the step S1 include jet flow Q, jet pressure F, number of nozzles M, and lifting rate vsWater-cement ratio W, rotational speed N and nozzle diameter doThe stratum parameters are geological parameters obtained by performing physical and mechanical tests on a quasi-jet grouting area or a jet grouting pile passing through a soil layer, and comprise drainage/non-drainage shear strength s and dry weight rd。
Preferably, the preprocessing in step S2 includes removing outliers, and calculating a minimum value, a maximum value, a standard deviation value, and an average value of the samples, specifically:
the abnormal value removal means that abnormal data which are larger than or smaller than the corresponding average value and exceed 3 times of standard deviation in each operating parameter of each sampling point are removed;
the average stratum parameter value is the average soil layer sample parameter value of each jet grouting pile punching sampling;
the construction parameter average value of the jet grouting pile refers to the average value of the construction parameters of each grouted jet grouting pile.
Preferably, the data set in step S3 is used as input data and an output result of training the artificial neural network, where the input data is construction parameters and formation parameters of the jet grouting pile, and the output result is a corresponding diameter of the jet grouting pile.
Preferably, the normalization processing in step S3 is to perform non-dimensionalization processing on the input data, and map the data to the (0, 1) range, and the normalization formula (1) is as follows:
in the formula, Xn,minAnd Xn,maxRespectively as the minimum value and the maximum value of X, wherein X is one of construction parameters and stratum parameters of the jet grouting pile, and X isnAnd the values are normalized by X.
Preferably, the differential evolution artificial neural network in step S4 is specifically:
obtaining N population individuals by using a differential evolution algorithm, wherein the population individuals represent a possible parameter combination of the artificial neural network and comprise a hidden layer number gamma of the artificial neural network, a node unit number phi of each layer, a regularization coefficient psi and iteration times omega, and the population individuals comprise:
the artificial neural network is a data network consisting of a plurality of connected layered node units and a weight w, and the weight w is a weighted value connected with the node units;
the hidden layer number gamma is the layer number of other layers of the artificial neural network except the input layer and the output layer;
the regularization coefficient Ψ is a coefficient of a regularization term in a cost function of the artificial neural network;
the iteration frequency omega refers to the frequency of carrying out reverse adjustment on the weight and the deviation so as to optimize the cost function.
Preferably, the initializing the differential evolution artificial neural network in step S4 specifically includes:
weight w for artificial neural networkGAnd population individuals x of a differential evolution algorithmi,GGiving an initial value, and enabling G to represent the nth generation generated by the ANN-DE method in an iteration mode; when G is 0, the model is in an initialization state, and initial weight w is obtained0And the first generation population individuals xi,0The initial value is shown as follows:
xi,0＝[x1,i,0,x2,i,0,...,xD,i,0]i＝1,2,...,N；G＝0 (2)
in the formula, D is a parameter which needs to be optimized by differential evolution, and specifically means four parameters (gamma, phi, psi and omega) contained in a population individual; n is the number of individuals for which the iteration contains a population.
Preferably, the N-adam optimization algorithm in step S5 is iteratively trained:
the N-Adam optimization algorithm comprises the following steps:
(a) calculating weight parameter gradient
(b) Updating gradients
(c) Calculating the first order momentum mt←μ·mt-1+(1-μ)gt
(d) Calculating a first order momentum correction term
(e) Calculating second order momentum
(f) Calculating a second order momentum correction term
(g) Calculating updated weight parameter
Wherein v and mu are momentum exponential decay parameters, and the defaults are 0.9 and 0.999; coefficient ε is 1e-8(ii) a Eta is the step length, and the default is 0.002; w is at-1Weight matrix and vector, w, referring to the last iteration of the differential evolution modeltRefers to the updated weight matrix and vector.
The iterative training means iteratively updating the weight w of the artificial neural network so as to minimize a cost function C of the artificial neural network;
the cost function C satisfies formula (3):
in the formula, n is the number of samples,
for the model output value of the i-th sample, piFor the target value of the ith sample, Ψ is a regularization coefficient, w is an assigned weight, and the weight and bias update process satisfies equation (4):
where δ represents the learning rate, n is the number of samples, and w is the assigned weight.
Preferably, mutation, recombination and selection of the differential evolution artificial neural network are performed in step S6 to obtain next generation population individuals, specifically:
the mutation refers to the vector x of individuals for a given populationi,G＝[x1,i,G,x2,i,G,x3,i,G]Adding the weighted difference of the two vectors to the third vector to obtain a new vector, as shown in equation (5):
vi,G＝x1,i,G-1+F(x2,i,G-1-x3,i,G-1) (5)
wherein v isi,GIs a donor vector, F is a mutation factor, and F is epsilon [0,2 ∈]；
The recombination refers to a test vector ui,GFrom the target vector xi,G-1And donor vector vi,GAssuming that the elements of the donor vector enter the test vector with a probability CR, ui,GThe jth component u ofj,i,GIs expressed as in equation (6):
wherein, randj,i～U[0,1]And I andrand[1,2,...,D]to ensure vi,G≠xi,G-1A range of random integers;
the selection refers to the target vector xi,G-1And the test vector vi,GThe comparison is performed and then the target vector corresponding to the smallest function value is selected for the next generation, as shown in equation (7):
preferably, the accuracy in step S7 is the standard deviation RMSE of the prediction error, as shown in equation (8):
where N is the number of data points, YobsIs the actual jet grouting pile diameter, YpredThe method is a predicted value of the diameter of the jet grouting pile of the differential evolution artificial neural network.
Compared with the prior art, the technical scheme of the invention has the beneficial effects that:
aiming at the defects of difficult construction parameter determination, low algorithm prediction efficiency, large deviation and the like in the existing method, the invention improves the calculation efficiency by utilizing an N-Adam method optimization algorithm, improves the calculation accuracy by utilizing a differential evolution artificial neural network, does not need to assume a model theory, and realizes the efficient and accurate prediction of the diameter of the jet grouting pile.
Drawings
FIG. 1 is a schematic flow chart of the method of the present invention.
FIG. 2 is a comparison graph of predicted results and measured results of the present invention in an example.
Detailed Description
The drawings are for illustrative purposes only and are not to be construed as limiting the patent;
for the purpose of better illustrating the embodiments, certain features of the drawings may be omitted, enlarged or reduced, and do not represent the size of an actual product;
it will be understood by those skilled in the art that certain well-known structures in the drawings and descriptions thereof may be omitted.
The technical solution of the present invention is further described below with reference to the accompanying drawings and examples.
Example 1
The embodiment provides a method for determining the diameter of a jet grouting pile based on differential evolution artificial intelligence, as shown in fig. 1, the method comprises the following steps:
s1: collecting construction parameters, stratum parameters and corresponding diameters of the jet grouting piles;
s2: preprocessing the collected construction parameters and stratum parameters of the jet grouting pile and the corresponding diameter of the jet grouting pile;
s3: dividing a data set formed by the construction parameters, the formation parameters and the corresponding jet grouting pile diameters of the pretreated jet grouting piles into a training set and a testing set, and carrying out normalization processing on the data set;
s4: initializing a differential evolution artificial neural network;
s5: inputting a training set into the differential evolution artificial neural network, and performing iterative training by using an N-Adam optimization algorithm;
s6: carrying out mutation, recombination and selection of a differential evolution artificial neural network to obtain next generation population individuals;
s7: repeating the steps S5 to S6 until the evolution algebra reaches a threshold value or the accuracy of the training set reaches a requirement, finishing the training, and storing the differential evolution artificial neural network at the moment;
s8: and (4) inputting the jet grouting pile data set into the differential evolution artificial neural network obtained in the step S7 to obtain the diameter of the jet grouting pile.
The construction parameters of the jet grouting pile in the step S1 comprise jet flow Q (X)1,m3S), injection pressure F (X)2kN), number of nozzles M (X)3) A lifting rate vs(X4M/s), water-cement ratio W (X)5) Rotational speed N (X)6Rpm) and nozzle diameter do(X7M), the stratum parameter is a geological parameter obtained by performing a physical mechanical test on the jet grouting area or the jet grouting pile passing through the soil layer, and the geological parameter is obtained byShear strength with/without drainage s (X)8kPa) and dry weight rd(X9,kN/m3)。
Geological exploration refers to taking soil from a region needing reinforcement on a construction site, exploration holes are arranged according to 12-35 m according to geotechnical engineering exploration specification GB 50021, exploration holes 1/3-1/2 are control holes, the control holes penetrate through the thickness of a compression layer below a pile end plane, and general exploration holes are required to penetrate 3-5 times of the designed diameter of a pile body below the pile end plane and are not smaller than 3 m.
The preprocessing in step S2 includes removing outliers, and calculating a minimum value, a maximum value, a standard deviation value, and an average value of the samples, and specifically includes:
the abnormal value removal means that abnormal data which are larger than or smaller than the corresponding average value and exceed 3 times of standard deviation in each operating parameter of each sampling point are removed;
the average stratum parameter value is the average soil layer sample parameter value of each jet grouting pile punching sampling;
the construction parameter average value of the jet grouting pile refers to the average value of the construction parameters of each grouted jet grouting pile.
In this embodiment, the number of nozzles M (X)3) The maximum value is 2 and the minimum value is 1; rate of lift vs(X4M/s) maximum value of 0.0103m/s and minimum value of 0.0020 m/s; water-cement ratio W (X)5) Maximum value of 0.0103 and minimum value of 0.8; rotational speed N (X)6Rpm) maximum of 1.5rpm and minimum of 0.8; diameter d of nozzleo(X7M) a maximum of 0.008m and a minimum of 0.0018 m; drainage/non-drainage shear strength s (X)8kPa) maximum value of 212.00kPa and minimum value of 42.70 kPa; dryness and severity degree rd(X9,kN/m3) Maximum value of 16.11kN/m3Minimum value of 13.70kN/m3；
In this embodiment, the number of nozzles M (X)3) Standard deviation of 0.47, lifting rate vs(X4M/s) standard deviation of 0.0014m/s, water-cement ratio W (X)5) Standard deviation of 1.181, rotational speed N (X)6Rpm) standard deviation of 0.18rpm, nozzle diameter do(X7M) standard deviation of 0.0009m,drainage/non-drainage shear strength s (X)8kPa) standard deviation of 33.1kPa, dry weight rd(X9,kN/m3) The standard deviation is 0.63kN/m3；
In the present embodiment, the shear strength s (X) of drainage/non-drainage8kPa) average value of 119.88kPa, dry weight rd(X9,kN/m3) The average value is 15.236kN/m3；
In this embodiment, the number of nozzles M (X)3) Average value of 1.31, lifting rate vs(X4M/s) average value of 0.0053m/s, water-cement ratio W (X)5) Average value of 1.006, speed N (X)6Rpm) mean value of 1.01rpm, nozzle diameter do(X7M) has an average value of 0.0031 m.
And the data set in the step S3 is used as input data and an output result of training the artificial neural network, wherein the input data are construction parameters and formation parameters of the jet grouting pile, and the output result is the corresponding diameter of the jet grouting pile.
The normalization process in step S3 is a non-dimensionalization process performed on the input data to map the data to a (0, 1) range, and the normalization formula (1) is as follows:
in the formula, Xn,minAnd Xn,maxRespectively as the minimum value and the maximum value of X, wherein X is one of construction parameters and stratum parameters of the jet grouting pile, and X isnIs the value after X normalization;
in this embodiment, the number of nozzles X3,maxIs 2, X3,minIs 1; lifting rate X4,maxIs 0.0103, X4,minIs 0.0020; water-cement ratio X5,maxIs 0.0103, X5,minIs 0.8; rotational speed X6,maxAt 1.5rpm, X6,minIs 0.8; nozzle diameter X7,maxIs 0.008m, X7,minIs 0.0018 m; shear strength X of drainage/non-drainage8,maxAt 212.00kPa, X8,min42.70 kPa; dryness and severity X9,maxIs 16.11kN/m3，X9,minIs 13.70kN/m3。
The differential evolution artificial neural network in the step S4 specifically includes:
obtaining N population individuals by using a differential evolution algorithm, wherein the population individuals represent a possible parameter combination of the artificial neural network and comprise a hidden layer number gamma of the artificial neural network, a node unit number phi of each layer, a regularization coefficient psi and iteration times omega, and the population individuals comprise:
the artificial neural network is a data network consisting of a plurality of connected layered node units and a weight w, and the weight w is a weighted value connected with the node units;
the hidden layer number gamma is the layer number of other layers of the artificial neural network except the input layer and the output layer;
the regularization coefficient Ψ is a coefficient of a regularization term in a cost function of the artificial neural network;
the iteration frequency omega refers to the frequency of carrying out reverse adjustment on the weight and the deviation so as to optimize the cost function.
In step S4, initializing the differential evolution artificial neural network specifically includes:
weight w for artificial neural networkGAnd population individuals x of a differential evolution algorithmi,GGiving an initial value, and enabling G to represent the nth generation generated by the ANN-DE method in an iteration mode; when G is 0, the model is in the initialized state, and in the present embodiment, the number of hidden layers Γ is [1,2 ]]The number of node units per layer is [1,8 ]]The regularization coefficient Ψ is [0.0,0.4 ]]The number of iterations omega is [1000,3000]. Get the initial weight w0And the first generation population individuals xi,0The initial value is shown as follows:
xi,0＝[x1,i,0,x2,i,0,...,xD,i,0]i＝1,2,...,N；G＝0 (2)
in the formula, D is a parameter which needs to be optimized by differential evolution, and specifically means four parameters (gamma, phi, psi and omega) contained in a population individual; n is the number of population-containing individuals in the iteration, and N is 50 in this embodiment.
The N-Adam optimization algorithm in step S5 is iteratively trained:
the N-Adam optimization algorithm comprises the following steps:
(a) calculating weight parameter gradient
(b) Updating gradients
(c) Calculating the first order momentum mt←μ·mt-1+(1-μ)gt
(d) Calculating a first order momentum correction term
(e) Calculating second order momentum
(f) Calculating a second order momentum correction term
(g) Calculating updated weight parameter
Wherein v and mu are momentum exponential decay parameters, and the defaults are 0.9 and 0.999; coefficient ε is 1e-8(ii) a Eta is the step length, and the default is 0.002; w is at-1Weight matrix and vector, w, referring to the last iteration of the differential evolution modeltRefers to the updated weight matrix and vector.
The iterative training means iteratively updating the weight w of the artificial neural network so as to minimize a cost function C of the artificial neural network;
the cost function C satisfies formula (3):
in the formula, n is the number of samples,
for the model output value of the i-th sample, piFor the target value of the ith sample, Ψ is a regularization coefficient, w is an assigned weight, and the weight and bias update process satisfies equation (4):
where δ represents the learning rate, n is the number of samples, and w is the assigned weight.
In step S6, mutation, recombination, and selection of the artificial neural network of the differential evolution are performed to obtain next generation population individuals, specifically:
the mutation refers to the vector x of individuals for a given populationi,G＝[x1,i,G,x2,i,G,x3,i,G]Adding the weighted difference of the two vectors to the third vector to obtain a new vector, as shown in equation (5):
vi,G＝x1,i,G-1+F(x2,i,G-1-x3,i,G-1) (5)
wherein v isi,GCalled donor vector, F is a mutation factor, and F is equal to [0,2 ]]；
The recombination refers to a test vector ui,GFrom the target vector xi,G-1And donor vector vi,GAssuming that the elements of the donor vector enter the test vector with a probability CR, ui,GThe jth component u ofj,i,GIs expressed as in equation (6):
wherein, randj,i～U[0,1]And I andrand[1,2,...,D]to ensure vi,G≠xi,G-1A range of random integers;
the selectionSelecting refers to converting the target vector xi,G-1And the test vector vi,GThe comparison is performed and then the target vector corresponding to the smallest function value is selected for the next generation, as shown in equation (7):
the accuracy in step S7 is the standard deviation RMSE of the prediction error, as shown in equation (8):
where N is the number of data points, YobsIs the actual jet grouting pile diameter, YpredThe method is a predicted value of the diameter of the jet grouting pile of the differential evolution artificial neural network.
Predesignated maximum number of evolutionary iterations Gmax＝100。
The test model results show that compared with the traditional optimized neural network and the theoretical model, the proposed ANN-DE can achieve higher accuracy while ensuring the reliability and efficiency of the training process. In this case, when G is 27, the optimal hyper-parameter can be extracted; at this time, the number of hidden layers Γ is 2, the number of node elements Φ in each layer is 8, the regularization coefficient Ψ is 0.24469, and the number of iterations Ω is 2995. The results show that the standard deviation RMSE obtained from training is 0.05339 and the standard deviation RMSE obtained from testing is 0.04238.
In the present embodiment, the average value of predicted values of the jet grouting pile diameters is 0.65071 m.
The prediction results are shown in FIG. 2, where the block is the result of the ANN-DE method and the cross point is the result of the theoretical method. The distribution of points on the graph can obtain better fitting degree of the ANN-DE method, and the correlation R2The ANN-DE process has a higher 0.97 than the theoretical process 0.93. Meanwhile, the model performance of the RMSE and coefficient-of-determination mapping shows that the deviation of the ANN-DE method from the theoretical method RMSE is 0.02818, and the deviation of the coefficient-of-determination is 0.04, and therefore,the ANN-DE method can achieve higher accuracy while ensuring the reliability and efficiency of the training process.
The same or similar reference numerals correspond to the same or similar parts;
the terms describing positional relationships in the drawings are for illustrative purposes only and are not to be construed as limiting the patent;
it should be understood that the above-described embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the embodiments of the present invention. Other variations and modifications will be apparent to persons skilled in the art in light of the above description. And are neither required nor exhaustive of all embodiments. Any modification, equivalent replacement, and improvement made within the spirit and principle of the present invention should be included in the protection scope of the claims of the present invention.
Claims (10)
1. A method for determining the diameter of a jet grouting pile based on differential evolution artificial intelligence is characterized by comprising the following steps:
s1: collecting construction parameters, stratum parameters and corresponding diameters of the jet grouting piles;
s2: preprocessing the collected construction parameters and stratum parameters of the jet grouting pile and the corresponding diameter of the jet grouting pile;
s3: dividing a data set formed by the construction parameters, the formation parameters and the corresponding jet grouting pile diameters of the pretreated jet grouting piles into a training set and a testing set, and carrying out normalization processing on the data set;
s4: initializing a differential evolution artificial neural network;
s5: inputting a training set into the differential evolution artificial neural network, and performing iterative training by using an N-Adam optimization algorithm;
s6: carrying out mutation, recombination and selection of a differential evolution artificial neural network to obtain next generation population individuals;
s7: repeating the steps S5 to S6 until the evolution algebra reaches a threshold value or the accuracy of the training set reaches a requirement, finishing the training, and storing the differential evolution artificial neural network at the moment;
s8: and (4) inputting the jet grouting pile data set into the differential evolution artificial neural network obtained in the step S7 to obtain the diameter of the jet grouting pile.
2. The method for determining the diameter of the jet grouting pile based on the differential evolution artificial intelligence of claim 1, wherein the construction parameters of the jet grouting pile in the step S1 comprise jet flow Q, jet pressure F, number M of nozzles, lifting rate vsWater-cement ratio W, rotational speed N and nozzle diameter doThe stratum parameters are geological parameters obtained by performing physical and mechanical tests on a quasi-jet grouting area or a jet grouting pile passing through a soil layer, and comprise drainage/non-drainage shear strength s and dry weight rd。
3. The method for determining the diameter of the jet grouting pile based on the evolutionary diversity artificial intelligence of claim 1, wherein the preprocessing in step S2 includes removing outliers, and calculating a minimum value, a maximum value, a standard deviation value and an average value of the samples, specifically:
the abnormal value removal means that abnormal data which are larger than or smaller than the corresponding average value and exceed 3 times of standard deviation in each operating parameter of each sampling point are removed;
the average stratum parameter value is the average soil layer sample parameter value of each jet grouting pile punching sampling;
the construction parameter average value of the jet grouting pile refers to the average value of the construction parameters of each grouted jet grouting pile.
4. The method for determining the diameter of the jet grouting pile based on the differential evolution artificial intelligence is characterized in that the data set in the step S3 is used as input data and an output result for training an artificial neural network, wherein the input data are construction parameters and formation parameters of the jet grouting pile, and the output result is the corresponding diameter of the jet grouting pile.
5. The method for determining a diameter of a jet grouting pile based on a differential evolution artificial intelligence according to claim 4, wherein the normalization process in step S3 is a non-dimensionalization process of the input data, and the data is mapped to a range of (0, 1), and the normalization formula (1) is as follows:
in the formula, Xn,minAnd Xn,maxRespectively as the minimum value and the maximum value of X, wherein X is one of construction parameters and stratum parameters of the jet grouting pile, and X isnAnd the values are normalized by X.
6. The method for determining the diameter of the jet grouting pile based on the evolutionary diversity artificial intelligence of claim 5, wherein the evolutionary diversity artificial neural network in the step S4 is specifically:
obtaining N population individuals by using a differential evolution algorithm, wherein the population individuals represent a possible parameter combination of the artificial neural network and comprise a hidden layer number gamma of the artificial neural network, a node unit number phi of each layer, a regularization coefficient psi and iteration times omega, and the population individuals comprise:
the artificial neural network is a data network consisting of a plurality of connected layered node units and a weight w, and the weight w is a weighted value connected with the node units;
the hidden layer number gamma is the layer number of other layers of the artificial neural network except the input layer and the output layer;
the regularization coefficient Ψ is a coefficient of a regularization term in a cost function of the artificial neural network;
the iteration frequency omega refers to the frequency of carrying out reverse adjustment on the weight and the deviation so as to optimize the cost function.
7. The method for determining the diameter of the jet grouting pile based on the evolutionary diversity artificial intelligence of claim 6, wherein the initializing the evolutionary diversity artificial neural network in step S4 specifically comprises:
weights for artificial neural networkswGAnd population individuals x of a differential evolution algorithmi,GGiving an initial value, and enabling G to represent the nth generation generated by the ANN-DE method in an iteration mode; when G is 0, the model is in an initialization state, and initial weight w is obtained0And the first generation population individuals xi,0The initial value is shown as follows:
xi,0＝[x1,i,0,x2,i,0,...,xD,i,0]i＝1,2,...,N；G＝0 (2)
in the formula, D is a parameter which needs to be optimized by differential evolution, and specifically means four parameters (gamma, phi, psi and omega) contained in a population individual; n is the number of individuals for which the iteration contains a population.
8. The method for determining the diameter of the jet grouting pile based on the differential evolution artificial intelligence is characterized in that the N-Adam optimization algorithm in the step S5 is iteratively trained:
the N-Adam optimization algorithm comprises the following steps:
(a) calculating weight parameter gradient
(b) Updating gradients
(c) Calculating the first order momentum mt←μ·mt-1+(1-μ)gt
(d) Calculating a first order momentum correction term
(e) Calculating second order momentum
(f) Calculating a second order momentum correction term
(g) Calculating updated weight parameter
Wherein v and mu are momentum exponential decay parameters, and the defaults are 0.9 and 0.999; coefficient ε is 1e-8(ii) a Eta is the step length, and the default is 0.002; w is at-1Weight matrix and vector, w, referring to the last iteration of the differential evolution modeltRefers to the updated weight matrix and vector.
The iterative training means iteratively updating the weight w of the artificial neural network so as to minimize a cost function C of the artificial neural network;
the cost function C satisfies formula (3):
in the formula, n is the number of samples,
for the model output value of the i-th sample, piFor the target value of the ith sample, Ψ is a regularization coefficient, w is an assigned weight, and the weight and bias update process satisfies equation (4):
where δ represents the learning rate, n is the number of samples, and w is the assigned weight.
9. The method for determining the diameter of the jet grouting pile based on the differential evolution artificial intelligence of claim 8, wherein mutation, recombination and selection of the differential evolution artificial neural network are performed in step S6 to obtain a next generation population, specifically:
the mutation refers to a mutation for a given speciesGroup individual vector xi,G＝[x1,i,G,x2,i,G,x3,i,G]Adding the weighted difference of the two vectors to the third vector to obtain a new vector, as shown in equation (5):
wherein v isi,GIs a donor vector, F is a mutation factor, and F is epsilon [0,2 ∈]；
vi,G＝x1,i,G-1+F(x2,i,G-1-x3,i,G-1) (5)
The recombination refers to a test vector ui,GFrom the target vector xi,G-1And donor vector vi,GAssuming that the elements of the donor vector enter the test vector with a probability CR, ui,GThe jth component u ofj,i,GIs expressed as in equation (6):
wherein, randj,i～U[0,1]And I andrand[1,2,...,D]to ensure vi,G≠xi,G-1A range of random integers;
the selection refers to the target vector xi,G-1And the test vector vi,GThe comparison is performed and then the target vector corresponding to the smallest function value is selected for the next generation, as shown in equation (7):
10. the method for determining the diameter of the jet grouting pile based on the differential evolution artificial intelligence of claim 9, wherein the accuracy in the step S7 is the standard deviation RMSE of the prediction error, as shown in formula (8):
where N is the number of data points, YobsIs the actual jet grouting pile diameter, YpredThe method is a predicted value of the diameter of the jet grouting pile of the differential evolution artificial neural network.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110373814.XA CN113128108B (en) | 2021-04-07 | 2021-04-07 | Method for determining diameter of jet grouting pile based on differential evolution artificial intelligence |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110373814.XA CN113128108B (en) | 2021-04-07 | 2021-04-07 | Method for determining diameter of jet grouting pile based on differential evolution artificial intelligence |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113128108A true CN113128108A (en) | 2021-07-16 |
| CN113128108B CN113128108B (en) | 2023-05-26 |
Family
ID=76775338
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110373814.XA Active CN113128108B (en) | 2021-04-07 | 2021-04-07 | Method for determining diameter of jet grouting pile based on differential evolution artificial intelligence |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113128108B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113777647A (en) * | 2021-09-29 | 2021-12-10 | 北京中岩大地科技股份有限公司 | Gamma ray data processing method for monitoring jet grouting pile diameter |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6269351B1 (en) * | 1999-03-31 | 2001-07-31 | Dryken Technologies, Inc. | Method and system for training an artificial neural network |
| CN107194524A (en) * | 2017-07-28 | 2017-09-22 | 合肥工业大学 | A kind of Coal and Gas Outbursts Prediction method based on RBF neural |
| CN107563518A (en) * | 2017-09-12 | 2018-01-09 | 太原理工大学 | A kind of learning method of the extreme learning machine based on social force model colony optimization algorithm |
| CN107812343A (en) * | 2017-08-29 | 2018-03-20 | 浙江理工大学 | A kind of vault sports training method |
| CN107947738A (en) * | 2017-12-11 | 2018-04-20 | 南京航空航天大学 | A kind of Forecasting Methodology of solar energy unmanned plane cell plate voltage |
| CN110059348A (en) * | 2019-03-12 | 2019-07-26 | 南京工程学院 | A kind of axial phase magnetically levitated flywheel motor suspending power numerical modeling method |
| US20200196870A1 (en) * | 2018-04-16 | 2020-06-25 | Beijing University Of Technology | A near-infrared spectroscopy tomography reconstruction method based on neural network |
-
2021
- 2021-04-07 CN CN202110373814.XA patent/CN113128108B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6269351B1 (en) * | 1999-03-31 | 2001-07-31 | Dryken Technologies, Inc. | Method and system for training an artificial neural network |
| CN107194524A (en) * | 2017-07-28 | 2017-09-22 | 合肥工业大学 | A kind of Coal and Gas Outbursts Prediction method based on RBF neural |
| CN107812343A (en) * | 2017-08-29 | 2018-03-20 | 浙江理工大学 | A kind of vault sports training method |
| CN107563518A (en) * | 2017-09-12 | 2018-01-09 | 太原理工大学 | A kind of learning method of the extreme learning machine based on social force model colony optimization algorithm |
| CN107947738A (en) * | 2017-12-11 | 2018-04-20 | 南京航空航天大学 | A kind of Forecasting Methodology of solar energy unmanned plane cell plate voltage |
| US20200196870A1 (en) * | 2018-04-16 | 2020-06-25 | Beijing University Of Technology | A near-infrared spectroscopy tomography reconstruction method based on neural network |
| CN110059348A (en) * | 2019-03-12 | 2019-07-26 | 南京工程学院 | A kind of axial phase magnetically levitated flywheel motor suspending power numerical modeling method |
Non-Patent Citations (4)
| Title |
|---|
| BISONG E: "《Building Machine Learning》", 31 December 2019 * |
| KUN ZHANG ET AL: "Evolutionary hybrid neural network approach to predict shield tunneling induced ground settlements", 《ELSEVIER:TUNNELLING AND UNDERGROUND SPACE TECHNOLOGY》 * |
| M. OCHMAŃSKI ET AL: "Prediction of the diameter of jet grouting columns with artificial neural networks", 《ELSEVIER:SOILS AND FOUNDATIONS》 * |
| TIMOTHY DOZAT: "Incorporating Nesterov Momentum Into ADAM", 《ICLR 2016 WORKSHOP》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113777647A (en) * | 2021-09-29 | 2021-12-10 | 北京中岩大地科技股份有限公司 | Gamma ray data processing method for monitoring jet grouting pile diameter |
| CN113777647B (en) * | 2021-09-29 | 2024-04-19 | 北京中岩大地科技股份有限公司 | Gamma ray data processing method for jet grouting pile diameter monitoring |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113128108B (en) | 2023-05-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Jong et al. | State-of-the-art review of geotechnical-driven artificial intelligence techniques in underground soil-structure interaction | |
| Zhang et al. | Application of LSTM approach for modelling stress–strain behaviour of soil | |
| Zhang et al. | Estimation of diaphragm wall deflections for deep braced excavation in anisotropic clays using ensemble learning | |
| Gandomi et al. | A new multi-gene genetic programming approach to non-linear system modeling. Part II: geotechnical and earthquake engineering problems | |
| Kohestani et al. | Prediction of maximum surface settlement caused by earth pressure balance shield tunneling using random forest | |
| Chou et al. | Peak shear strength of discrete fiber-reinforced soils computed by machine learning and metaensemble methods | |
| Njock et al. | Artificial neural network optimized by differential evolution for predicting diameters of jet grouted columns | |
| Jan et al. | Neural network forecast model in deep excavation | |
| Ghanizadeh et al. | Developing bearing capacity model for geogrid-reinforced stone columns improved soft clay utilizing MARS-EBS hybrid method | |
| CN107153843A (en) | Surface subsidence forecasting system and method based on SVMs | |
| Liu et al. | Rock Strength Estimation Using Several Tree-Based ML Techniques. | |
| Harandizadeh et al. | TBM performance prediction developing a hybrid ANFIS-PNN predictive model optimized by imperialism competitive algorithm | |
| Jahed Armaghani et al. | Load carrying capacity assessment of thin-walled foundations: an ANFIS–PNN model optimized by genetic algorithm | |
| Mbarak et al. | SPT based determination of undrained shear strength: Regression models and machine learning | |
| Liu et al. | A hybrid data-driven model for geotechnical reliability analysis | |
| CN113323676B (en) | Method for determining cutter torque of shield machine by using principal component analysis-long and short memory model | |
| CN113128108A (en) | Method for determining diameter of jet grouting pile based on differential evolution artificial intelligence | |
| Kong et al. | Novel hybrid method to predict the ground-displacement field caused by shallow tunnel excavation | |
| CN115659758A (en) | Shield tunnel rock-soil parameter inversion and tunneling parameter optimization method based on approximate model | |
| Fattahi et al. | Application of adaptive neuro-fuzzy inference system for the assessment of damaged zone around underground spaces | |
| Dashtgoli et al. | Performance comparison of machine learning algorithms for maximum displacement prediction in soldier pile wall excavation | |
| Cividini et al. | Deterministic and probabilistic back analysis in rock mechanics | |
| Qiao et al. | Artificial neural network to predict the surface maximum settlement by shield tunneling | |
| CN112329287B (en) | P-y curve Bayes learning method based on test pile monitoring data | |
| JIANG et al. | Inversion iterative correction method for estimating shear strength of rock and soil mass in slope engineering |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |