CN111460708A - Dam mechanical parameter prediction method based on optimized neural network - Google Patents

Dam mechanical parameter prediction method based on optimized neural network Download PDF

Info

Publication number
CN111460708A
CN111460708A CN202010226886.7A CN202010226886A CN111460708A CN 111460708 A CN111460708 A CN 111460708A CN 202010226886 A CN202010226886 A CN 202010226886A CN 111460708 A CN111460708 A CN 111460708A
Authority
CN
China
Prior art keywords
neural network
dam
output
mechanical parameters
sample
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.)
Withdrawn
Application number
CN202010226886.7A
Other languages
Chinese (zh)
Inventor
黄丹
关志豪
吕小龙
姜冬菊
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Hohai University HHU
Original Assignee
Hohai University HHU
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Hohai University HHU filed Critical Hohai University HHU
Priority to CN202010226886.7A priority Critical patent/CN111460708A/en
Publication of CN111460708A publication Critical patent/CN111460708A/en
Withdrawn legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/044—Recurrent networks, e.g. Hopfield networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/086—Learning methods using evolutionary algorithms, e.g. genetic algorithms or genetic programming
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/12—Computing arrangements based on biological models using genetic models
    • G06N3/126—Evolutionary algorithms, e.g. genetic algorithms or genetic programming

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Biophysics (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Biomedical Technology (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Computational Linguistics (AREA)
  • Software Systems (AREA)
  • Artificial Intelligence (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Evolutionary Biology (AREA)
  • Physiology (AREA)
  • Genetics & Genomics (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention discloses a dam mechanical parameter prediction method based on an optimized neural network, which comprises the steps of obtaining multiple groups of mechanical parameters of multiple dams and corresponding measuring point displacements of the multiple groups of mechanical parameters, determining the multiple groups of mechanical parameters as sample labels, determining the corresponding measuring point displacements of the multiple groups of mechanical parameters as sample characteristics to obtain training samples, inputting the sample characteristics into the neural network, training by taking the sample labels as the output of the neural network, so that a mapping relation between the measuring point displacements and the mechanical parameters is formed in the neural network, inputting the actual measurement displacements of the dam to be measured into the trained neural network, and enabling the neural network to output the mechanical parameters of the dam to be measured, so that the mechanical parameters of the dam to be measured can be rapidly and accurately obtained, and the accuracy of the determined mechanical parameters of the dam to be measured can be improved.

Description

Dam mechanical parameter prediction method based on optimized neural network
Technical Field
The invention relates to the technical field of dam mechanical parameter inverse analysis, in particular to a dam mechanical parameter prediction method based on an optimized neural network.
Background
The construction and operation conditions of dams such as high arch dams and the like are complex, the dams are subjected to load effects such as seepage, temperature, flood discharge and the like for a long time, and the safety state of the dams is in a dynamic process for a long time. There is a great uncertainty about the physical-mechanical parameters of the dam, and the parameters of different areas change with the passage of time. The method for analyzing the constantly changing mechanical state of the arch dam by adopting the fixed and unchangeable mechanical parameters often cannot obtain an accurate result. The inversion of the mechanical parameters of the dam by using a reverse analysis method has become one of the important means for analyzing the actual operation condition of the dam. The intelligent algorithm has certain defects such as poor global search capability of a neural network, low calculation speed, easy premature convergence of a genetic algorithm, dependence of results on selection of initial values, and low accuracy of dam mechanical parameters determined by the traditional scheme.
Disclosure of Invention
Aiming at the problems, the invention provides a dam mechanics parameter prediction method based on an optimized neural network.
In order to realize the aim of the invention, the dam mechanics parameter prediction method based on the optimized neural network comprises the following steps:
s10, acquiring multiple groups of mechanical parameters of multiple dams and corresponding measuring point displacements of the multiple groups of mechanical parameters, determining the multiple groups of mechanical parameters as sample labels, and determining the corresponding measuring point displacements of the multiple groups of mechanical parameters as sample characteristics to obtain training samples;
s20, inputting the sample characteristics into a neural network, and training by taking the sample labels as the output of the neural network so as to form a mapping relation between measuring point displacement and mechanical parameters in the neural network;
and S30, inputting the actual measurement displacement of the dam to be measured into the trained neural network, and enabling the neural network to output the mechanical parameters of the dam to be measured.
In an embodiment, the dam mechanics parameter prediction method based on an optimized neural network further includes:
s40, establishing a finite element model according to the mechanical parameters of the dam to be detected output by the neural network, carrying out finite element forward analysis calculation to obtain the calculated displacement of the dam to be detected, comparing the calculated displacement with the actually measured displacement value, and determining the inversion precision of the neural network according to the comparison result.
In one embodiment, acquiring multiple sets of mechanical parameters of multiple dams, and corresponding displacement of measuring points of the sets of mechanical parameters comprises:
acquiring multiple groups of mechanical parameters of multiple dams, establishing a dam finite element model, and respectively carrying out forward analysis calculation on each group of mechanical parameters by adopting the dam finite element model through a finite element method to obtain corresponding measuring point displacement of each group of mechanical parameters.
In one embodiment, before inputting the sample features into a neural network and training with the sample labels as the output of the neural network, the method further includes:
taking the weight and threshold value of the BP neural network random initialization as an initial feasible solution group of the genetic algorithm to carry out population initialization, wherein the individual of the genetic algorithm comprises the connection weight value of an input layer and a hidden layer of the BP neural network, the threshold value of the hidden layer, the connection weight value of the hidden layer and an output layer and the threshold value of the output layer;
and setting a mapping relation among network layers after population initialization so as to determine the neural network.
As an embodiment, the mapping relationship between network layers includes:
the output of the hidden layer neurons is:
Figure BDA0002427981550000021
wherein x isiIn order to input the samples, the method,
Figure BDA0002427981550000022
the connection weights between the input layer and the hidden layer,
Figure BDA0002427981550000023
for hidden layer neuron thresholds, PjF () is an activation function of the hidden layer neuron, and is usually taken as a Sigmoid function;
the output of the output layer neurons is:
Figure BDA0002427981550000024
wherein the content of the first and second substances,
Figure BDA0002427981550000025
the connection weights between the hidden layer and the output layer,
Figure BDA0002427981550000026
as output layer neuron threshold, QkIs the output value, P, of an output layer neuroniFor the output values of hidden layer neurons, g () is an output layer neuron activation function, typically taking a linear function.
As an embodiment, inputting the sample features into a neural network, and training with the sample labels as the output of the neural network comprises:
inputting the sample characteristics into the neural network, calculating to obtain output values corresponding to the sample characteristics according to the mapping relation between network layers in the neural network, and taking the sum of absolute values of differences between the output values and expected values as an individual fitness value in a genetic algorithm;
the method comprises the steps that initial weights and threshold values randomly determined by a neural network are used as initial feasible solution groups of a genetic algorithm, the fitness values of individuals are calculated, the groups are gradually evolved to the optimal solution through the genetic algorithm operation, and the optimal weights and threshold values are continuously searched;
and determining the optimized neural network according to the determined optimal weight and the optimal threshold.
According to the dam mechanical parameter prediction method based on the optimized neural network, multiple groups of mechanical parameters of multiple dams and corresponding measuring point displacements of the multiple groups of mechanical parameters are obtained, the multiple groups of mechanical parameters are determined as sample labels, the corresponding measuring point displacements of the multiple groups of mechanical parameters are determined as sample characteristics, training samples are obtained, the sample characteristics are input into the neural network, the sample labels are used as output of the neural network for training, so that a mapping relation between the measuring point displacements and the mechanical parameters is formed in the neural network, the actual measurement displacements of the dam to be measured are input into the trained neural network, the mechanical parameters of the dam to be measured are output by the neural network, the mechanical parameters of the dam to be measured are rapidly and accurately obtained, and the accuracy of the determined mechanical parameters of the dam to be measured can be improved.
Drawings
FIG. 1 is a flow chart of a dam mechanics parameter prediction method based on an optimized neural network according to an embodiment;
FIG. 2 is a flow chart of a dam mechanics parameter prediction method based on an optimized neural network according to another embodiment;
FIG. 3 is a schematic diagram of a BP neural network topology according to an embodiment;
FIG. 4 is a schematic representation of a concrete calculation region model of example 1 in one embodiment;
FIG. 5 is a comparison graph of mean error of elastic modulus inversion values of example 1 in one embodiment;
FIG. 6 is a comparison graph of 20 inversion results for partition No. 1 in example 1 in one embodiment;
FIG. 7 is a dam calculation model of example 2 in one embodiment;
FIG. 8 is a comparison graph of the number 1 bullet-mode partitioning inversion results for the model of example 2 in one embodiment;
FIG. 9 is a comparison graph of the results of the number 2 bullet-mode partition inversion of the example 2 model in one embodiment;
FIG. 10 is a comparison graph of the inversion results of the bullet-mode partitioning for model No. 3 of example 2 in one embodiment.
Detailed Description
In order to make the objects, technical solutions and advantages of the present application more apparent, the present application is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present application and are not intended to limit the present application.
Reference herein to "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment can be included in at least one embodiment of the application. The appearances of the phrase in various places in the specification are not necessarily all referring to the same embodiment, nor are separate or alternative embodiments mutually exclusive of other embodiments. It is explicitly and implicitly understood by one skilled in the art that the embodiments described herein can be combined with other embodiments.
Referring to fig. 1, fig. 1 is a flowchart of a dam mechanical parameter prediction method based on an optimized neural network according to an embodiment, in this embodiment, a relationship between a mechanical parameter value and a displacement value of a dam may be mapped by using a BP neural network, when the mapping relationship is determined, a group of displacement values is input, and a result obtained through network prediction is a parameter to be inverted, which may specifically include the following steps:
s10, obtaining multiple groups of mechanical parameters of the dams and corresponding measuring point displacements of the multiple groups of mechanical parameters, determining the multiple groups of mechanical parameters as sample labels, and determining the corresponding measuring point displacements of the multiple groups of mechanical parameters as sample characteristics to obtain training samples.
The multiple groups of mechanical parameters are mechanical parameters of multiple dams, and a large number of mechanical parameters can be obtained for the multiple dams. The set of mechanical parameters may include elastic modulus, poisson's ratio, and other mechanical parameters.
In one embodiment, acquiring multiple sets of mechanical parameters of multiple dams, and corresponding displacement of measuring points of the sets of mechanical parameters comprises:
acquiring multiple groups of mechanical parameters of multiple dams, establishing a dam finite element model, and respectively carrying out forward analysis calculation on each group of mechanical parameters by adopting the dam finite element model through a finite element method to obtain corresponding measuring point displacement of each group of mechanical parameters.
In this embodiment, multiple sets of mechanical parameters including elastic modulus, poisson ratio and other mechanical parameters may be used as labels of a sample, a large number of samples are given, a corresponding dam finite element model is established according to the given mechanical parameters, forward analysis and calculation are performed by a finite element method, and displacement of a measuring point corresponding to each set of mechanical parameters is obtained as characteristics of the sample, so as to obtain a training sample including a large number of sample labels and sample characteristics corresponding to each sample label.
And S20, inputting the sample characteristics into a neural network, and training by taking the sample labels as the output of the neural network so as to form a mapping relation between the measuring point displacement and the mechanical parameters in the neural network.
The neural network can be a BP neural network, and the BP neural network can be optimized by adopting a genetic algorithm in a specific training process. The initialized connection weight and the threshold of the simple BP neural network are randomly determined, the defects of easy falling into local optimal solution, low calculation speed and the like exist in the process of prediction, and the inherent defect of the BP neural network caused by the random initialized weight threshold can be avoided by using a genetic algorithm for optimization.
And S30, inputting the actual measurement displacement of the dam to be measured into the trained neural network, and enabling the neural network to output the mechanical parameters of the dam to be measured.
Inputting the actual measurement displacement of the dam to be tested into the trained neural network to obtain the mechanical parameter value to be inverted so as to obtain the mechanical parameters of the dam to be tested, such as elastic modulus, Poisson ratio and the like.
According to the dam mechanical parameter prediction method based on the optimized neural network, multiple groups of mechanical parameters of multiple dams and corresponding measuring point displacements of the multiple groups of mechanical parameters are obtained, the multiple groups of mechanical parameters are determined as sample labels, the corresponding measuring point displacements of the multiple groups of mechanical parameters are determined as sample characteristics, training samples are obtained, the sample characteristics are input into the neural network, the sample labels are used as output of the neural network for training, so that a mapping relation between the measuring point displacements and the mechanical parameters is formed in the neural network, the actual measurement displacements of the dam to be measured are input into the trained neural network, the mechanical parameters of the dam to be measured are output by the neural network, the mechanical parameters of the dam to be measured are rapidly and accurately obtained, and the accuracy of the determined mechanical parameters of the dam to be measured can be improved.
In an embodiment, the dam mechanics parameter prediction method based on an optimized neural network further includes:
s40, establishing a finite element model according to the mechanical parameters of the dam to be detected output by the neural network, carrying out finite element forward analysis calculation to obtain the calculated displacement of the dam to be detected, comparing the calculated displacement with the actually measured displacement value, and determining the inversion precision of the neural network according to the comparison result.
Specifically, the mechanical parameters of the dam to be tested output by the neural network are obtained by inversion, a finite element model can be established according to the obtained mechanical parameters by inversion, finite element forward analysis calculation is performed, the obtained calculated displacement value is compared with the actually measured displacement value, and the accuracy of neural network inversion can be further verified.
In one embodiment, before inputting the sample features into a neural network and training with the sample labels as the output of the neural network, the method further includes:
taking the weight and threshold value of the BP neural network random initialization as an initial feasible solution group of the genetic algorithm to carry out population initialization, wherein the individual of the genetic algorithm comprises the connection weight value of an input layer and a hidden layer of the BP neural network, the threshold value of the hidden layer, the connection weight value of the hidden layer and an output layer and the threshold value of the output layer;
and setting a mapping relation among network layers after population initialization so as to determine the neural network.
As an embodiment, the mapping relationship between the network layers includes:
the output of the hidden layer neurons is:
Figure BDA0002427981550000051
wherein x isiIn order to input the samples, the method,
Figure BDA0002427981550000052
the connection weights between the input layer and the hidden layer,
Figure BDA0002427981550000053
for hidden layer neuron thresholds, PjF () is an activation function of the hidden layer neuron, and is usually taken as a Sigmoid function;
the output of the output layer neurons is:
Figure BDA0002427981550000054
wherein the content of the first and second substances,
Figure BDA0002427981550000055
the connection weights between the hidden layer and the output layer,
Figure BDA0002427981550000056
as output layer neuron threshold, QkIs the output value, P, of an output layer neuroniFor the output values of hidden layer neurons, g () is an output layer neuron activation function, typically taking a linear function.
As an embodiment, inputting the sample features into a neural network, and training with the sample labels as the output of the neural network comprises:
inputting the sample characteristics into the neural network, calculating to obtain output values corresponding to the sample characteristics according to the mapping relation between network layers in the neural network, and taking the sum of absolute values of differences between the output values and expected values as an individual fitness value in a genetic algorithm;
the method comprises the steps that initial weights and threshold values randomly determined by a neural network are used as initial feasible solution groups of a genetic algorithm, the fitness values of individuals are calculated, the groups are gradually evolved to the optimal solution through the genetic algorithm operation, and the optimal weights and threshold values are continuously searched;
and determining the optimized neural network according to the determined optimal weight and the optimal threshold.
The embodiment may determine a neural network suitable for dam mechanics parameter prediction, and specifically, the determining process of the neural network may also include:
taking the weight value and the threshold value of the random initialization of the BP neural network as an initial feasible solution group of the genetic algorithm to perform population initialization, wherein the individual of the genetic algorithm comprises 4 parts of the connection weight value of an input layer and a hidden layer of the BP neural network, the threshold value of the hidden layer, the connection weight value of the hidden layer and an output layer and the threshold value of the output layer.
Step two, the mapping relation between the network layers comprises the following two parts:
(1) the output of the hidden layer neurons is:
Figure BDA0002427981550000061
wherein x isiIn order to input the samples, the method,
Figure BDA0002427981550000062
the connection weights between the input layer and the hidden layer,
Figure BDA0002427981550000063
is a hidden layer neuron threshold.
(2) The output of the output layer neurons is:
Figure BDA0002427981550000064
wherein the content of the first and second substances,
Figure BDA0002427981550000065
the connection weights between the hidden layer and the output layer,
Figure BDA0002427981550000066
is the output layer neuron threshold.
Further, based on the neural network determined in the above manner, the corresponding training process includes:
and step three, inputting sample characteristics, and calculating to obtain an output value, namely a predicted mechanical parameter, according to the interlayer mapping relation in the step two. Taking the sum of the absolute values of the differences between the predicted parameter value and the expected value as the fitness value of the individual in the genetic algorithm.
Step four, according to the initial weight and the threshold randomly determined by the neural network in the step one and the step two, the initial feasible solution group of the genetic algorithm is used, the fitness value of the individual is calculated through the step three, the group is gradually evolved to the optimal solution through the operation of the genetic algorithm, and the optimal weight and the optimal threshold are continuously searched.
And step five, obtaining the optimal solution individuals according to the step four, namely the optimal weight and the threshold of the optimized BP neural network, and then predicting by using the BP neural network to improve the precision of the prediction result.
In the embodiment, the interlayer connection weight and the threshold of the neural network are optimized through genetic operations such as selection, intersection, variation and the like in the genetic algorithm, so that the optimal solution individuals obtained after population evolution is finished are used as the optimal weight and the threshold of the BP neural network. The optimized novel algorithm improves the condition that the error of the simple BP neural network is larger due to the randomly determined initial weight and threshold, avoids falling into the local optimal solution, and can improve the prediction precision and stability of the BP network. The basic content of the determined optimized neural network prediction algorithm for dam mechanics parameter inversion is that the weight and the threshold of the BP neural network are evolved towards the optimal solution through genetic operations such as selection, intersection, variation and the like of the genetic algorithm, and the prediction precision of the BP neural network is improved after the BP neural network is optimized.
In an embodiment, the dam mechanics parameter prediction method based on the optimized neural network may also be as shown in fig. 2, and the BP neural network topology may be as shown in fig. 3. Assume that there are two sets of training samples, input samples:
Figure BDA0002427981550000071
outputting a sample:
Figure BDA0002427981550000072
wherein A isi＝[a1,a2,…,ax]，Bi＝[b1,b2,…,by]N is the sample size, x is the input sample dimension, and y is the output sample dimension. After two groups of training samples are trained by using a BP neural network, the input value A of the (n + 1) th group of test samplesn+1The test sample is brought into a trained network, and the corresponding test sample output value B can be predictedn+1. Applications ofIn the inversion of the mechanical parameters of the dam, the displacement value of the monitoring point of the dam can be used as the sample characteristic and input into the neural network, and the mechanical parameter value of the dam is used as the sample label. Under the given sample label, the corresponding sample characteristics are obtained through calculation by a finite element method, and the BP neural network is trained by using a large number of samples obtained through finite elements to obtain the trained neural network. And finally, inputting the displacement value of the actual monitoring measuring point into a trained network, wherein the output prediction result is the corresponding parameter value to be inverted.
Example 1: the method comprises the steps of taking a concrete block with a simple geometric shape as an object, dividing the block into 8 different elastic modulus areas, carrying out positive analysis calculation under the load of concentrated force to obtain displacement data of a plurality of measuring points, and inverting the elastic modulus of the concrete in the calculation area by using a BP (back propagation) neural network and a GA-BP (genetic algorithm-back propagation) neural network respectively. Fig. 4 is a calculation model, fig. 5 is a comparison of average errors of results obtained by inverting 100 with two algorithms, and fig. 6 is an error comparison graph of results obtained by inverting 20 times with two algorithms in the number 1 bullet-mode region.
Example 2: the method comprises the steps of taking the section of a certain arch dam as an object, dividing the dam into areas to be inverted with different elastic moduli along the elevation, and inverting the elastic modulus of the dam through a plurality of monitoring points on the dam body. The height of the dam is 100m, and the width of the dam bottom is 60 m. The upstream of the dam is subjected to hydrostatic pressure, the dam body is subjected to gravity, the unit thickness of the dam is taken, and the displacement value of each measuring point is obtained by performing positive analysis calculation according to the plane stress problem. FIG. 7 is a calculation model, and FIGS. 8 to 10 are graphs comparing results obtained by inverting the BP neural network and the GA-BP neural network 20 times respectively.
The technical effects obtained according to the above examples 1 and 2 include: the result shows that the minimum error of the inversion result of the basic BP neural network is 8.0 percent, and the maximum error is 23.0 percent; the minimum error of the neural network inversion result after the genetic algorithm is optimized is 3.1%, and the maximum error is 11.2%. Errors of inversion results of the two algorithms are within an allowable range, and prediction accuracy of a novel algorithm formed by optimizing the simple BP neural network by using a genetic algorithm is obviously improved.
The technical features of the above embodiments can be arbitrarily combined, and for the sake of brevity, all possible combinations of the technical features in the above embodiments are not described, but should be considered as the scope of the present specification as long as there is no contradiction between the combinations of the technical features.
It should be noted that the terms "first \ second \ third" referred to in the embodiments of the present application merely distinguish similar objects, and do not represent a specific ordering for the objects, and it should be understood that "first \ second \ third" may exchange a specific order or sequence when allowed. It should be understood that "first \ second \ third" distinct objects may be interchanged under appropriate circumstances such that the embodiments of the application described herein may be implemented in an order other than those illustrated or described herein.
The terms "comprising" and "having" and any variations thereof in the embodiments of the present application are intended to cover non-exclusive inclusions. For example, a process, method, apparatus, product, or device that comprises a list of steps or modules is not limited to the listed steps or modules but may alternatively include other steps or modules not listed or inherent to such process, method, product, or device.
The above-mentioned embodiments only express several embodiments of the present application, and the description thereof is more specific and detailed, but not construed as limiting the scope of the invention. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the concept of the present application, which falls within the scope of protection of the present application. Therefore, the protection scope of the present patent shall be subject to the appended claims.

Claims (6)

1. A dam mechanics parameter prediction method based on an optimized neural network is characterized by comprising the following steps:
s10, acquiring multiple groups of mechanical parameters of multiple dams and corresponding measuring point displacements of the multiple groups of mechanical parameters, determining the multiple groups of mechanical parameters as sample labels, and determining the corresponding measuring point displacements of the multiple groups of mechanical parameters as sample characteristics to obtain training samples;
s20, inputting the sample characteristics into a neural network, and training by taking the sample labels as the output of the neural network so as to form a mapping relation between measuring point displacement and mechanical parameters in the neural network;
and S30, inputting the actual measurement displacement of the dam to be measured into the trained neural network, and enabling the neural network to output the mechanical parameters of the dam to be measured.
2. The optimized neural network-based dam mechanics parameter prediction method of claim 1, further comprising:
s40, establishing a finite element model according to the mechanical parameters of the dam to be detected output by the neural network, carrying out finite element forward analysis calculation to obtain the calculated displacement of the dam to be detected, comparing the calculated displacement with the actually measured displacement value, and determining the inversion precision of the neural network according to the comparison result.
3. The optimized neural network-based dam mechanical parameter prediction method according to claim 1, wherein the obtaining of multiple sets of mechanical parameters of multiple dams and corresponding measuring point displacement of each set of mechanical parameters comprises:
acquiring multiple groups of mechanical parameters of multiple dams, establishing a dam finite element model, and respectively carrying out forward analysis calculation on each group of mechanical parameters by adopting the dam finite element model through a finite element method to obtain corresponding measuring point displacement of each group of mechanical parameters.
4. The optimized neural network-based dam mechanics parameter prediction method according to any one of claims 1 to 3, wherein before inputting the sample features into a neural network and training with the sample labels as the output of the neural network, further comprising:
taking the weight and threshold value of the BP neural network random initialization as an initial feasible solution group of the genetic algorithm to carry out population initialization, wherein the individual of the genetic algorithm comprises the connection weight value of an input layer and a hidden layer of the BP neural network, the threshold value of the hidden layer, the connection weight value of the hidden layer and an output layer and the threshold value of the output layer;
and setting a mapping relation among network layers after population initialization so as to determine the neural network.
5. The optimized neural network-based dam mechanics parameter prediction method according to claim 4, wherein the mapping relationship between network layers comprises:
the output of the hidden layer neurons is:
Figure FDA0002427981540000011
wherein x isiIn order to input the samples, the method,
Figure FDA0002427981540000012
the connection weights between the input layer and the hidden layer,
Figure FDA0002427981540000013
for hidden layer neuron thresholds, PjF () is the activation function of the hidden layer neuron, and is the output value of the hidden layer neuron;
the output of the output layer neurons is:
Figure FDA0002427981540000021
wherein the content of the first and second substances,
Figure FDA0002427981540000022
the connection weights between the hidden layer and the output layer,
Figure FDA0002427981540000023
as output layer neuron threshold, QkIs the output value, P, of an output layer neuroniFor the output values of hidden layer neurons, g () is the output layer neuron activation function.
6. The optimized neural network-based dam mechanics parameter prediction method of claim 5, wherein inputting the sample features into a neural network, and training with the sample labels as the output of the neural network comprises:
inputting the sample characteristics into the neural network, calculating to obtain output values corresponding to the sample characteristics according to the mapping relation between network layers in the neural network, and taking the sum of absolute values of differences between the output values and expected values as an individual fitness value in a genetic algorithm;
the method comprises the steps that initial weights and threshold values randomly determined by a neural network are used as initial feasible solution groups of a genetic algorithm, the fitness values of individuals are calculated, the groups are gradually evolved to the optimal solution through the genetic algorithm operation, and the optimal weights and threshold values are continuously searched;
and determining the optimized neural network according to the determined optimal weight and the optimal threshold.
CN202010226886.7A 2020-03-27 2020-03-27 Dam mechanical parameter prediction method based on optimized neural network Withdrawn CN111460708A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010226886.7A CN111460708A (en) 2020-03-27 2020-03-27 Dam mechanical parameter prediction method based on optimized neural network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010226886.7A CN111460708A (en) 2020-03-27 2020-03-27 Dam mechanical parameter prediction method based on optimized neural network

Publications (1)

Publication Number Publication Date
CN111460708A true CN111460708A (en) 2020-07-28

Family

ID=71685727

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010226886.7A Withdrawn CN111460708A (en) 2020-03-27 2020-03-27 Dam mechanical parameter prediction method based on optimized neural network

Country Status (1)

Country Link
CN (1) CN111460708A (en)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113191055A (en) * 2021-05-06 2021-07-30 河海大学 Dam material performance parameter inversion method based on deep reinforcement network
CN113468824A (en) * 2021-07-29 2021-10-01 北京全四维动力科技有限公司 Model training method and calculation method for calculating loss coefficient of mechanical blade of impeller
CN114034770A (en) * 2021-11-15 2022-02-11 金陵科技学院 Data detection method and system based on construction dam mechanics big data
CN114594706A (en) * 2022-02-07 2022-06-07 安徽聚源水利科技液压坝制造有限公司 Network control hinge dam and intelligent control system

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113191055A (en) * 2021-05-06 2021-07-30 河海大学 Dam material performance parameter inversion method based on deep reinforcement network
CN113191055B (en) * 2021-05-06 2022-05-10 河海大学 Dam material performance parameter inversion method based on deep reinforcement network
CN113468824A (en) * 2021-07-29 2021-10-01 北京全四维动力科技有限公司 Model training method and calculation method for calculating loss coefficient of mechanical blade of impeller
CN114034770A (en) * 2021-11-15 2022-02-11 金陵科技学院 Data detection method and system based on construction dam mechanics big data
CN114594706A (en) * 2022-02-07 2022-06-07 安徽聚源水利科技液压坝制造有限公司 Network control hinge dam and intelligent control system

Similar Documents

Publication Publication Date Title
CN111460708A (en) Dam mechanical parameter prediction method based on optimized neural network
CN109145464B (en) Structural damage identification method integrating multi-target ant lion optimization and trace sparse regularization
CN107480341A (en) A kind of dam safety comprehensive method based on deep learning
CN108304679A (en) A kind of adaptive reliability analysis method
CN108876021B (en) Medium-and-long-term runoff forecasting method and system
KR101919076B1 (en) Time-series data predicting system
CN107885928A (en) Consider the stepstress acceleration Degradation Reliability analysis method of measurement error
CN112307536B (en) Dam seepage parameter inversion method
CN115659729B (en) Dam safety monitoring analysis method and system based on structural simulation calculation
CN111428419A (en) Suspended sediment concentration prediction method and device, computer equipment and storage medium
CN106568647A (en) Nerve network-based concrete strength predication method
CN112765880B (en) Method for monitoring stratum saturated brine invasion amount based on Bi-LSTM
CN105933388A (en) WSN data layered fusion method for plant growth monitoring
CN111161799B (en) Method and system for acquiring polygenic risk scores based on multigroup study data
CN112598114A (en) Power consumption model construction method, power consumption measurement method and device and electronic equipment
CN117146954A (en) Weighing compensation method and device based on improved WOA-BP neural network
CN115099493B (en) Forest fire spreading rate prediction method in any direction based on CNN
CN109101759A (en) A kind of parameter identification method based on forward and reverse response phase method
CN113468795A (en) Universal building method for different parameter material partition pouring/filling dam deformation mixed model
CN111859744A (en) Node rigid domain identification method, device and equipment based on monitoring data
CN113657021A (en) Marine measurement period evaluation method based on BP neural network
CN111811827A (en) Product performance consistency inspection method based on Rayleigh distribution
CN111210877A (en) Method and device for deducing physical property parameters
CN117669394B (en) Mountain canyon bridge long-term performance comprehensive evaluation method and system
CN113152541B (en) Rapid detection method for horizontal bearing capacity of single pile

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
WW01 Invention patent application withdrawn after publication
WW01 Invention patent application withdrawn after publication

Application publication date: 20200728