CN117636065A - Concrete dam displacement deep learning prediction method and system considering crack influence - Google Patents
Concrete dam displacement deep learning prediction method and system considering crack influence Download PDFInfo
- Publication number
- CN117636065A CN117636065A CN202311777157.0A CN202311777157A CN117636065A CN 117636065 A CN117636065 A CN 117636065A CN 202311777157 A CN202311777157 A CN 202311777157A CN 117636065 A CN117636065 A CN 117636065A
- Authority
- CN
- China
- Prior art keywords
- displacement
- influence
- model
- factor
- deep learning
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000006073 displacement reaction Methods 0.000 title claims abstract description 123
- 238000000034 method Methods 0.000 title claims abstract description 79
- 238000013135 deep learning Methods 0.000 title claims abstract description 35
- 238000012544 monitoring process Methods 0.000 claims abstract description 49
- 238000011134 hematopoietic stem cell transplantation Methods 0.000 claims abstract description 31
- 238000012216 screening Methods 0.000 claims abstract description 21
- 238000010187 selection method Methods 0.000 claims abstract description 17
- 238000004458 analytical method Methods 0.000 claims abstract description 12
- 230000006870 function Effects 0.000 claims description 15
- 238000012360 testing method Methods 0.000 claims description 14
- 238000004364 calculation method Methods 0.000 claims description 13
- 238000013136 deep learning model Methods 0.000 claims description 11
- 230000015654 memory Effects 0.000 claims description 10
- 238000012549 training Methods 0.000 claims description 10
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims description 9
- 238000004590 computer program Methods 0.000 claims description 8
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 claims description 6
- 238000004422 calculation algorithm Methods 0.000 claims description 6
- 230000032683 aging Effects 0.000 claims description 5
- 238000003860 storage Methods 0.000 claims description 5
- 230000008901 benefit Effects 0.000 abstract description 4
- 230000007774 longterm Effects 0.000 abstract description 3
- 230000009467 reduction Effects 0.000 description 8
- 230000000694 effects Effects 0.000 description 5
- 238000005259 measurement Methods 0.000 description 5
- 238000003491 array Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 238000000692 Student's t-test Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 238000010835 comparative analysis Methods 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 238000001816 cooling Methods 0.000 description 1
- 238000005336 cracking Methods 0.000 description 1
- 238000007418 data mining Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 230000005484 gravity Effects 0.000 description 1
- 230000002706 hydrostatic effect Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000003973 irrigation Methods 0.000 description 1
- 230000002262 irrigation Effects 0.000 description 1
- 238000012886 linear function Methods 0.000 description 1
- 238000012417 linear regression Methods 0.000 description 1
- 230000007787 long-term memory Effects 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000013178 mathematical model Methods 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 239000013307 optical fiber Substances 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000035515 penetration Effects 0.000 description 1
- 238000010248 power generation Methods 0.000 description 1
- 230000006403 short-term memory Effects 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 238000013179 statistical model Methods 0.000 description 1
- 238000012353 t test Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
- 238000012800 visualization Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/18—Complex mathematical operations for evaluating statistical data, e.g. average values, frequency distributions, probability functions, regression analysis
-
- 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
- G06N3/0442—Recurrent networks, e.g. Hopfield networks characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
-
- 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
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- 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
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/774—Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Software Systems (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- Computing Systems (AREA)
- Artificial Intelligence (AREA)
- Health & Medical Sciences (AREA)
- Mathematical Physics (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Business, Economics & Management (AREA)
- Molecular Biology (AREA)
- Databases & Information Systems (AREA)
- Biomedical Technology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computational Mathematics (AREA)
- Operations Research (AREA)
- Mathematical Optimization (AREA)
- Strategic Management (AREA)
- Pure & Applied Mathematics (AREA)
- Human Resources & Organizations (AREA)
- Medical Informatics (AREA)
- Economics (AREA)
- Multimedia (AREA)
- Mathematical Analysis (AREA)
- Tourism & Hospitality (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- General Business, Economics & Management (AREA)
- Probability & Statistics with Applications (AREA)
- Quality & Reliability (AREA)
- Algebra (AREA)
- Marketing (AREA)
Abstract
The invention discloses a concrete dam displacement deep learning prediction method and a system considering crack influence, comprising the following steps: acquiring monitoring data of a concrete dam, obtaining an initial modeling factor set, taking the crack opening and closing degree as an influence factor of displacement, and establishing a displacement monitoring model HSCT taking the crack influence into consideration; matching the influence factor set considering the crack with the observed displacement value to obtain a primary factor set; screening influence factors of displacement by adopting variable selection methods mRMR and Lasso, and removing modeling factors which do not meet the threshold requirement; and carrying out predictive analysis on the data set after factor screening according to an optimal CNN-LSTM model, and selecting a better variable selection method according to the predictive accuracy ratio. Compared with the traditional HST model, the method has the advantages that the precision is improved greatly, unnecessary complexity is reduced, long-term dependence and short-term characteristics of displacement data can be captured, and the model is more stable and reliable.
Description
Technical Field
The invention relates to the technical field of concrete dam safety monitoring, in particular to a concrete dam displacement deep learning prediction method and system considering crack influence.
Background
Since the 20 th century, many arch dams were constructed, and arch dam engineering plays an important role in flood control, power generation, water supply, irrigation and the like. However, as the engineering service time is continuously increased, the structural state of the dam is continuously changed, and dam materials are damaged to different degrees, so that potential safety hazards appear. Among the numerous effects, displacement is the monitoring quantity that most intuitively reflects the overall performance of the concrete dam. The reasonable mathematical model can effectively explain the influence of environmental load on arch dam displacement, so that the establishment of a high-precision displacement monitoring model is important to ensure long-term safe operation of the arch dam.
Hydrostatic pressure-season-time (HST) models are widely used in practical engineering. However, the existing monitoring method mainly uses an HST model, and model input only covers main influencing factors, so that the influence of other complex factors is ignored. This greatly affects the framework of the model and thus limits its prediction accuracy. For concrete dams, crack generation and propagation until penetration is the whole process of dam instability and failure. The integral strength and rigidity of the dam body are obviously reduced due to crack cracking, so that the dam body is easier to deform in the working process, and the safety and stability of the dam body are threatened.
However, due to the excessive number of seam gauges, adding all of the measured fracture opening to the model may lead to over-fitting problems. This problem can be avoided by introducing feature selection class methods, selecting strongly explanatory variables. The traditional concrete dam monitoring model mainly regards the effect quantity as a linear function of a related modeling factor, and engineering practice shows that the effect quantity and the modeling factor are in a complex nonlinear relation. In this regard, deep learning models, such as long and short term memory neural networks (LSTM), have good nonlinear data mining capabilities and excellent performance in processing time series data, and thus are widely used in the field of dam safety monitoring.
Disclosure of Invention
The present invention has been made in view of the above-described problems.
Therefore, the technical problems solved by the invention are as follows: the existing prediction method has the problems of low prediction precision and optimization of how to reduce the complexity of feature selection.
In order to solve the technical problems, the invention provides the following technical scheme: the concrete dam displacement deep learning prediction method considering the influence of cracks comprises the following steps:
acquiring monitoring data of a concrete dam, obtaining an initial modeling factor set, taking the crack opening and closing degree as an influence factor of displacement, and establishing a displacement monitoring model HSCT considering the crack influence;
based on a HSCT regression model, obtaining an influence factor set considering cracks, and matching the factors with an observed displacement value to obtain a primary factor set;
screening influence factors of displacement by adopting variable selection methods mRMR and Lasso, removing modeling factors which do not meet the threshold requirement, and constructing the rest modeling factors into a screened factor set;
and constructing a deep learning model CNN-LSTM with a mean square error MSE as a loss function, optimizing the model by using an Adam algorithm, obtaining an optimal parameter combination when the MSE is minimum, constructing an optimal CNN-LSTM model, carrying out predictive analysis on the screened factor set, and selecting a better variable selection method according to the predictive accuracy ratio.
As a preferable scheme of the concrete dam displacement deep learning prediction method considering the crack influence, the invention comprises the following steps: the initial modeling factor set includes a water pressure component, a temperature component, an aging component, and a fracture component;
the displacement monitoring model HSCT considering the crack influence is established, and the mathematical expression is as follows:
wherein H represents the depth of water in front of the dam; a, a 0 Representing constant terms; a, a i ，b 1i ，b 2i ，c 1 ，c 2 ，d i Representing fitting coefficients; m represents a coefficient, taking 3 or 4; n represents a time period, 1 or 2; t represents the accumulated number of days from the displacement monitoring date to the initial displacement monitoring date; θ=t/100; p represents the number of the seam gauges, J i The degree of fracture opening and closing of each joint meter is shown.
As a preferable scheme of the concrete dam displacement deep learning prediction method considering the crack influence, the invention comprises the following steps: the primary selection factor set is obtained by matching the primary selection factor with an observed displacement value; the observed displacement value is a displacement specific value measured by monitoring points arranged on the dam body.
As a preferable scheme of the concrete dam displacement deep learning prediction method considering the crack influence, the invention comprises the following steps: the factor removing step of the variable selecting method mRMR comprises the following steps:
calculating mutual information between two variables:
wherein x is i Represents the influence factor of displacement, y represents the measured displacement, p (x i ) And p (y) represents x i And an edge probability density of y; p (x) i Y) equals the joint probability density;
representing D as x i And mutual information I (x) i Average value of y):
wherein F represents x i Is equal to the number of features in F;
r is represented as x i And redundancy between y:
the operator phi is defined to combine the operations D and R:
let feature set F n-1 Is from the full feature set T m The extracted n-1 feature components are used for searching the optimal feature set by using an incremental search method:
and sequentially searching the features with the maximum mRMR value from the features, calculating the value phi, eliminating the corresponding influence factors when the value phi is smaller than 2, forming a candidate feature set, and taking the candidate feature set as the input of a subsequent prediction model.
As a preferable scheme of the concrete dam displacement deep learning prediction method considering the crack influence, the invention comprises the following steps: the factor eliminating step of the variable selecting method Lasso comprises the following steps:
compressing the coefficients of the variables and making some regression coefficients become 0 by constructing a penalty function; the mathematical expression is as follows:
wherein X represents a displacement influence factor, Y represents a displacement monitoring value, lambda represents an adjustment parameter, and a penalty term ism is more than or equal to 0, and beta represents the regression coefficient of the variable;
adding L in a linear model 1 Penalty term, lasso estimation for linear model:
order theWhen t is less than t 0 A portion of the coefficients are compressed to 0, thereby reducing the dimension of X;
after Lasso calculation, the regression coefficient of part of factors is changed into 0, and the factors with the regression coefficient of 0 are removed to form an input data set of a subsequent prediction model.
As a preferable scheme of the concrete dam displacement deep learning prediction method considering the crack influence, the invention comprises the following steps: the loss function MSE is defined as:
wherein n represents the number of prediction data,represents the ith monitoring data,/-)>Representing the ith prediction data.
As a preferable scheme of the concrete dam displacement deep learning prediction method considering the crack influence, the invention comprises the following steps: the method for establishing the deep learning model CNN-LSTM comprises the following steps:
taking the current day value corresponding to the displacement influence factor as the input of the CNN-LSTM model, and taking the current day measured displacement as the output of the CNN-LSTM model; and (3) at the same time node, actually measured data of each influence factor and displacement are calculated according to 4:1 are divided and respectively used as a training set and a testing set, wherein the training set is used for training a corresponding CNN-LSTM model, and the testing set is used for evaluating the prediction precision of the corresponding CNN-LSTM model and selecting a better variable selection method according to the prediction precision ratio;
by calculating the Root Mean Square Error (RMSE), mean Absolute Error (MAE) and square correlation coefficient (R) of measured value and predicted value in the test set 2 Three indexes are used for verifying the accuracy of the optimal CNN-LSTM model, and the calculation formulas of the three indexes are as follows:
wherein n represents a time series length, y i The monitoring data is represented by a representation of the data,representing predicted values +.>Mean value representing the sequence of observed data, +.>Representing the average of the predicted data sequence.
The invention relates to a concrete dam displacement deep learning prediction system taking crack influence into consideration by adopting the method, which is characterized in that:
the factor set generation module is used for acquiring monitoring data of the concrete dam, obtaining an initial modeling factor set, taking the crack opening and closing degree as an influence factor of displacement, and establishing a displacement monitoring model HSCT taking the crack influence into consideration;
based on a HSCT regression model, obtaining an influence factor set considering cracks, and matching the factors with an observed displacement value to obtain a primary factor set;
the screening module is used for screening influence factors of displacement by adopting variable selection methods mRMR and Lasso, eliminating modeling factors which do not meet the threshold requirement, and constructing the rest modeling factors into a screened factor set;
and the prediction module is used for constructing a deep learning model CNN-LSTM taking a mean square error MSE as a loss function, optimizing the model by using an Adam algorithm, obtaining an optimal parameter combination when the MSE is minimum, constructing an optimal CNN-LSTM model, carrying out prediction analysis on a screened factor set, and selecting a better variable selection method according to the prediction precision ratio.
A computer device, comprising: a memory and a processor; the memory stores a computer program characterized in that: the processor, when executing the computer program, implements the steps of the method of any of the present invention.
A computer-readable storage medium having stored thereon a computer program, characterized by: the computer program, when executed by a processor, implements the steps of the method of any of the present invention.
The invention has the beneficial effects that: according to the concrete dam displacement deep learning prediction method considering the crack influence, the crack opening and closing degree is used as the influence factor of displacement, and the HSCT model for displacement monitoring is constructed, so that the method has higher precision compared with a conventional HST model. Second, the feature selection method eliminates extraneous or redundant factors, reducing unnecessary complexity. Finally, the deep learning model CNN-LSTM can be used for predicting by combining the influence factors and the displacement at the last moment, so that the time-varying characteristic of the displacement data can be reflected better, and the stability and the reliability of the model are improved.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for the description of the embodiments will be briefly described below, it being obvious that the drawings in the following description are only some embodiments of the present invention, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art. Wherein:
FIG. 1 is an overall flowchart of a concrete dam displacement deep learning prediction method taking the influence of cracks into consideration according to a first embodiment of the present invention;
FIG. 2 is a detailed flowchart of a concrete dam displacement deep learning prediction method considering crack influence according to a second embodiment of the present invention;
FIG. 3 is a schematic diagram of a concrete dam vertical line monitoring station arrangement for a concrete dam displacement deep learning prediction method taking the influence of cracks into consideration according to a second embodiment of the present invention;
FIG. 4 is a histogram of prediction performance indexes of two measuring points of a concrete dam displacement deep learning prediction method considering crack influence provided by a second embodiment of the present invention under an HST model;
FIG. 5 is a histogram of predicted performance indexes of two measuring points of a concrete dam displacement deep learning prediction method considering crack influence according to a second embodiment of the present invention under a HSCT model;
FIG. 6 is a histogram of multi-model predictive performance indexes of measuring points PL8-U of a concrete dam displacement deep learning prediction method considering crack influence according to a second embodiment of the invention under different screening methods;
fig. 7 is a histogram of multi-model prediction performance indexes of the measuring points PL18-U of the concrete dam displacement deep learning prediction method considering the influence of cracks according to the second embodiment of the present invention under different screening methods.
Detailed Description
So that the manner in which the above recited objects, features and advantages of the present invention can be understood in detail, a more particular description of the invention, briefly summarized above, may be had by reference to the embodiments, some of which are illustrated in the appended drawings. All other embodiments, which can be made by one of ordinary skill in the art based on the embodiments of the present invention without making any inventive effort, shall fall within the scope of the present invention.
Example 1
Referring to fig. 1, for one embodiment of the present invention, there is provided a concrete dam displacement deep learning prediction method considering the influence of cracks, including:
s1: and obtaining an initial modeling factor set based on the HST model, and simultaneously taking the fracture opening and closing degree as an influence factor of displacement to establish a displacement monitoring model HSCT considering the fracture influence.
It is noted that the initial set of modeling factors includes a hydraulic pressure component, a temperature component, an aging component, and a fracture component. Establishing a displacement monitoring model HSCT considering crack influence, wherein the mathematical expression is as follows:
wherein H represents the depth of water in front of the dam; a, a 0 Representing constant terms; a, a i ，b 1i ，b 2i ，c 1 ，c 2 ，d i Representing fitting coefficients; m represents a coefficient, taking 3 or 4; n represents a time period, 1 or 2; t represents the accumulated number of days from the displacement monitoring date to the initial displacement monitoring date; θ=t/100; p represents the number of the seam gauges, J i The degree of fracture opening and closing of each joint meter is shown.
S2: based on the HSCT regression model, an influence factor set considering the crack is obtained, and a primary factor set is obtained after the factors are matched with the observed displacement value.
The preliminary set of factors includes a plurality of modeling factors including a hydraulic pressure component, a temperature component, an aging component, and a fracture component. And matching the primary selection factors with the observed displacement values according to dates to obtain a primary selection factor set. The observed displacement value is a displacement specific value measured by monitoring points arranged on the dam body.
S3: considering that the displacement influence factors of the cracks intensively contain a large number of features, the information is redundant, and only part of the features are closely related to the displacement. In order to obtain an optimal factor set, variable selection methods mRMR and Lasso are adopted to screen influence factors of displacement, modeling factors which do not meet the threshold requirement are removed, and the rest modeling factors are built into a factor set after screening.
Further, the factor removing step of the variable selecting method mRMR includes:
calculating mutual information between two variables:
wherein x is i Represents the influence factor of displacement, y represents the measured displacement, p (x i ) And p (y) represents x i And an edge probability density of y; p (x) i Y) equals the joint probability density;
representing D as x i And mutual information I (x) i Average value of y):
wherein F represents x i Is equal to the number of features in F;
r is represented as x i And redundancy between y:
the operator phi is defined to combine the operations D and R:
let feature set F n-1 Is from the full feature set T m The extracted n-1 feature components are used for searching the optimal feature set by using an incremental search method:
and sequentially searching the features with the maximum mRMR value from the features, calculating the value phi, eliminating the corresponding influence factors when the value phi is smaller than 2, forming a candidate feature set, and taking the candidate feature set as the input of a subsequent prediction model.
Further, the factor removing step of the variable selecting method Lasso includes:
compressing the coefficients of the variables and making some regression coefficients become 0 by constructing a penalty function; the mathematical expression is as follows:
wherein X represents a displacement influence factor, Y represents a displacement monitoring value, lambda represents an adjustment parameter, and a penalty term ism is more than or equal to 0, and beta represents the regression coefficient of the variable;
adding L in a linear model 1 Penalty term, lasso estimation for linear model:
order theWhen t is less than t 0 A portion of the coefficients are compressed to 0, thereby reducing the dimension of X;
after Lasso calculation, the regression coefficient of part of factors is changed into 0, and the factors with the regression coefficient of 0 are removed to form an input data set of a subsequent prediction model.
S4: and constructing a deep learning model CNN-LSTM with a mean square error MSE as a loss function, optimizing the model by using an Adam algorithm, obtaining an optimal parameter combination when the MSE is minimum, constructing an optimal CNN-LSTM model, carrying out predictive analysis on the screened factor set on the basis, and selecting a better variable selection method according to the predictive accuracy ratio.
Wherein the loss function MSE is defined as:
wherein n represents the number of prediction data,represents the ith monitoring data,/-)>Representing the ith prediction data.
Further, the method for establishing the deep learning model CNN-LSTM comprises the following steps:
the deep learning prediction model CNN-LSTM. Wherein CNN is used for extracting local features, LSTM is used for modeling long-term dependency, and the model mainly comprises an input layer, a convolution layer, an LSTM layer and an output layer.
Taking the current day value corresponding to the displacement influence factor as the input of the CNN-LSTM model, and taking the current day measured displacement as the output of the CNN-LSTM model; and (3) at the same time node, actually measured data of each influence factor and displacement are calculated according to 4:1 are divided and respectively used as a training set and a test set, wherein the training set is used for training the corresponding CNN-LSTM model, and the test set is used for evaluating the prediction precision of the corresponding CNN-LSTM model and selecting a better variable selection method according to the prediction precision ratio.
By calculating the Root Mean Square Error (RMSE), mean Absolute Error (MAE) and the square correlation coefficient (R) of the measured and predicted values in the test set 2 ) The accuracy of the optimal CNN-LSTM model is verified by three indexes, and the calculation formulas of the three indexes are as follows:
wherein n represents a time series length, y i The monitoring data is represented by a representation of the data,representing predicted values +.>Mean value representing the sequence of observed data, +.>Representing the average of the predicted data sequence. These three metrics are used to verify model performance, i.e., RMSE and MAE are as small as possible, R 2 As large as possible, and R 2 The closer to 1, the better the model effect.
On the other hand, the embodiment also provides a concrete dam displacement deep learning prediction system considering the influence of cracks, which comprises:
the factor set generation module is used for acquiring monitoring data of the concrete dam, obtaining an initial modeling factor set, taking the crack opening and closing degree as an influence factor of displacement, and establishing a displacement monitoring model HSCT taking the crack influence into consideration.
Based on the HSCT regression model, an influence factor set considering the crack is obtained, and a primary factor set is obtained after the factors are matched with the observed displacement value.
And the screening module is used for screening influence factors of displacement by adopting variable selection methods mRMR and Lasso, eliminating modeling factors which do not meet the threshold requirement, and constructing the rest modeling factors into a screened factor set.
And the prediction module is used for constructing a deep learning model CNN-LSTM taking a mean square error MSE as a loss function, optimizing the model by using an Adam algorithm, obtaining an optimal parameter combination when the MSE is minimum, constructing an optimal CNN-LSTM model, carrying out prediction analysis on a screened factor set, and selecting a better variable selection method according to the prediction precision ratio.
The above functions, if implemented in the form of software functional units and sold or used as a stand-alone product, may be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention may be embodied essentially or in a part contributing to the prior art or in a part of the technical solution, in the form of a software product stored in a storage medium, comprising several instructions for causing a computer device (which may be a personal computer, a server, a network device, etc.) to perform all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-only memory (ROM), a random access memory (RAM, random Access Memory), a magnetic disk, or an optical disk, or other various media capable of storing program codes.
Logic and/or steps represented in the flowcharts or otherwise described herein, e.g., a ordered listing of executable instructions for implementing logical functions, can be embodied in any computer-readable medium for use by or in connection with an instruction execution system, apparatus, or device, such as a computer-based system, processor-containing system, or other system that can fetch the instructions from the instruction execution system, apparatus, or device and execute the instructions. For the purposes of this description, a "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transport the program for use by or in connection with the instruction execution system, apparatus, or device.
More specific examples (a non-exhaustive list) of the computer-readable medium would include the following: an electrical connection (electronic device) having one or more wires, a portable computer diskette (magnetic device), a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disc read-only memory (CDROM). In addition, the computer readable medium may even be paper or other suitable medium on which the program is printed, as the program may be electronically captured, via, for instance, optical scanning of the paper or other medium, then compiled, interpreted or otherwise processed in a suitable manner, if necessary, and then stored in a computer memory.
It is to be understood that portions of the present invention may be implemented in hardware, software, firmware, or a combination thereof. In the above-described embodiments, the various steps or methods may be implemented in software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, may be implemented using any one or combination of the following techniques, as is well known in the art: discrete logic circuits having logic gates for implementing logic functions on data signals, application specific integrated circuits having suitable combinational logic gates, programmable Gate Arrays (PGAs), field Programmable Gate Arrays (FPGAs), and the like.
Example 2
Referring to fig. 2 to 7, for one embodiment of the present invention, a concrete dam displacement deep learning prediction method considering the influence of cracks is provided, and in order to verify the beneficial effects of the present invention, scientific demonstration is performed through economic benefit calculation and simulation experiments.
As shown in fig. 2, a concrete dam displacement deep learning prediction method considering crack influence includes:
the embodiment specifically comprises the following steps: a concrete gravity arch dam has a maximum dam height of 76.30m, a dam top elevation of 126.30m and plumb line observation points arranged on three representative dam sections of No. 8, no. 18 and No. 26. The vertical line monitoring point layout of the arch dam is shown in fig. 3. Because of unreasonable cooling measures, cracks are generated at the 105m elevation and transversely penetrate through a plurality of dam segments.
In order to explore the influence of cracks on the displacement of the measuring points, in the embodiment, the vertical displacement of the measuring points PL8-U and PL18-U of the No. 8 dam segment and the No. 18 dam segment near 105m elevation is selected for analysis from 24 days in 1 month in 2005 to 31 days in 12 months in 2018, and the front and back 2 segments of data are respectively used for training and evaluating the prediction performance of the CNN-LSTM model by taking the 3 months in 2016 as time demarcation points.
The measuring point of the arch dam near the 105m elevation is obviously affected by the crack, so that crack opening displacement is introduced as a crack component on the basis of a water pressure component, a temperature component and an aging component, a displacement monitoring model HSCT considering the crack effect is established, and the mathematical expression is as follows:
wherein H represents the depth of water in front of the dam; a, a 0 Representing constant terms; a, a i ，b 1i ，b 2i ，c 1 ，c 2 ，d i Representing fitting coefficients; m represents a coefficient, taking 3 or 4; n represents a time period, 1 or 2; t represents the accumulated number of days from the displacement monitoring date to the initial displacement monitoring date; θ=t/100; p represents the number of the seam gauges, J i The degree of fracture opening and closing of each joint meter is shown.
And selecting parameters in the statistical model according to dam characteristics, wherein m is 4 in the water pressure component, n is 2 in the temperature component, and monitoring data of 19 seam meters are selected, namely, p is 19. Based on a HSCT regression model, a factor set influencing displacement is obtained, and a primary factor set is obtained after the factors are matched with an observed displacement value:
wherein: x is X m Is a displacement influencing factor; y is Y m For actually measuring displacement, J i For the crack opening and closing degree measured by the joint meter, i=1, 2, …,19. In order to improve the convergence rate of the model, normalization processing is carried out on the data.
For comparison purposes, a traditional HST model is established, and the prediction of the HSCT model and the prediction of the HST model are compared and analyzed to demonstrate the necessity of researching a displacement monitoring method of the slotted concrete arch dam by considering the influence of cracks. Multiple Linear Regression (MLR), relevance Vector Machine (RVM), extreme Learning Machine (ELM) and LSTM model displacement prediction are also implemented on the same dataset.
Fig. 4 and 5 show the predictive model test set statistics for two stations under HST model and HSCT model. Compared with the HST model, the accuracy of the HSCT model considering the crack influence is improved during prediction, and the error of the MLR model is greatly reduced; the RMSE of RVM model was reduced by 17.5% on average and the MAE was reduced by 12.4% on average; RMSE of ELM model was reduced by 31.8% on average and MAE was reduced by 29.4% on average; RMSE of LSTM model was reduced by 11.7% on average and MAE was reduced by 6.9% on average; the average RMSE reduction of the CNN-LSTM model was 8.5% and the average MAE reduction was 7.9%.
The verification of the above models shows that the model accuracy is improved and the model performance is further enhanced by using the factor set under the HSCT model for prediction compared with the traditional HST model, which proves the necessity of considering the influence of cracks.
When there are only a few factors, the model may not capture complex relationships in the data, resulting in lower prediction accuracy. By increasing the number of factors, the model can better fit the changes and trends in the data, providing more accurate predictions. However, increasing the number of factors may also introduce noise or uncorrelated variables, which may lead to over-fitting problems.
In order to obtain an optimal factor set, screening influence factors by adopting a feature selection class method: eliminating factors with phi less than 2 by adopting an mRMR method; compressing the coefficients of the variables by using a Lasso method, and eliminating factors with regression coefficients of 0. For comparative analysis, variable significance analysis based on stepwise regression by the conventional method is also realized. Tables 1 and 2 show the calculation results of the influence factors of two measuring points under the HSCT model by different screening methods.
TABLE 1 calculation results of influence factors of measurement points PL8-U by different screening methods
TABLE 2 calculation results of influence factors of measurement points PL18-U by different screening methods
For the measuring point PL8-U, the factor x is calculated by mRMR 1 、x 2 、x 4 、x 10 、x 17 、x 19 、x 23 、x 28 Removing; after Lasso calculation, factor x 2 、x 13 、x 22 、x 23 Rejection was performed. For the measuring point PL18-U, the factor x is calculated by mRMR 1 、x 2 、x 3 、x 10 、x 17 、x 23 Removing; after Lasso calculation, factor x 13 、x 15 、x 23 Rejection was performed. Table 3 shows the t-statistic and p-statistic for the final set of factors for points PL8-U at 95% confidence intervals. Table 4 shows the t-statistic and p-statistic for the final set of factors for points PL18-U at 95% confidence intervals.
TABLE 3 influence factor significance analysis of measurement points PL8-U under HSCT model
TABLE 4 influence factor significance analysis of measurement points PL18-U under HSCT model
Multiple times of detection is carried out on the HSCT model of the measuring point PL8-U by utilizing stepwise regression, and factors are removed: x is x 28 、x 22 、x 26 、x 19 The values of the F statistic are 1750.53, 1815.41, 1884.15, 1958.10, respectively. Counter point PLThe HSCT model of 18-U is subjected to multiple tests, and corresponding rejection factors are obtained: x is x 20 、x 26 、x 22 、x 28 The values of the F statistic are 1655.30, 1716.85, 1783.11, 1854.33, respectively. From the F statistic, the value of F test shows an increasing trend, which indicates that the significant relationship between the target variable and the independent variable is gradually enhanced. As can be seen from tables 3 and 4, as the factors are progressively removed from the regression model, the p-test values are less than 0.05 and the absolute values of the t-test are greater than 2, indicating that each independent variable is significant in the regression model. Table 5 shows the final set of factors after various methods of calculation.
TABLE 5 factor sets after screening by various methods
After the variables of the measuring points PL8-U and PL18-U under the HSCT model are selected by different screening methods, the final factor set is put into various models for predictive analysis. Tables 6 and 7 calculate statistical indices for each model prediction after the points PL8-U and PL18-U have undergone different variable selection methods.
Table 6 statistical index of model predictions after measuring points PL8-U are subjected to factor screening
Table 7 statistical index of model predictions after measuring points PL18-U are subjected to factor screening
In the prediction of the two measurement point test sets, compared with the HSCT model which is not subjected to variable selection, the RMSE average is reduced by 8.73% and the MAE average is reduced by 11.1% in the stepwise regression method, which proves the reliability of deleting factors one by one; the improvement of the mRMR method and the Lasso method is larger, the average reduction of the RMSE is 20.6 percent under the mRMR method, the average reduction of the MAE is 26.9 percent under the Lasso method, the average reduction of the RMSE is 16.1 percent under the Lasso method, and the average reduction of the MAE is 21.2 percent. Compared with stepwise regression, after variable selection is performed by utilizing mRMR and Lasso, the accuracy of the two methods is higher than that of the stepwise regression according to the statistical index, and the variable selection method provided by the invention is proved to be reasonable and feasible. The regression model processed by the mRMR method has higher precision, compared with the Lasso method, the average RMSE is reduced by 5.2 percent, the average MAE is reduced by 6.8 percent, and the optimal method is mRMR.
Fig. 6 and 7 are visualizations of performance metrics for points PL8-U and PL18-U, respectively, showing an average 24.2% reduction in RMSE and 28.5% reduction in MAE compared to RVM and ELM for the CNN-LSTM model, based on statistical metrics. Compared with LSTM, the RMSE average of the CNN-LSTM model is reduced by 9.2%, the MAE average is reduced by 16.5%, namely the optimal model is CNN-LSTM, the fitting goodness is respectively up to 0.973 and 0.975, and the predicted value is basically consistent with the actual monitoring value, so that the advantages of the CNN-LSTM in the treatment of high-dimensional nonlinear relation are reflected, and the strong adaptability of the model is proved.
It should be noted that the above embodiments are only for illustrating the technical solution of the present invention and not for limiting the same, and although the present invention has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical solution of the present invention may be modified or substituted without departing from the spirit and scope of the technical solution of the present invention, which is intended to be covered in the scope of the claims of the present invention.
Claims (10)
1. The concrete dam displacement deep learning prediction method considering the influence of cracks is characterized by comprising the following steps:
acquiring monitoring data of a concrete dam, obtaining an initial modeling factor set, taking the crack opening and closing degree as an influence factor of displacement, and establishing a displacement monitoring model HSCT considering the crack influence;
based on a HSCT regression model, obtaining an influence factor set considering cracks, and matching the factors with an observed displacement value to obtain a primary factor set;
screening influence factors of displacement by adopting variable selection methods mRMR and Lasso, removing modeling factors which do not meet the threshold requirement, and constructing the rest modeling factors into a screened factor set;
and constructing a deep learning model CNN-LSTM with a mean square error MSE as a loss function, optimizing the model by using an Adam algorithm, obtaining an optimal parameter combination when the MSE is minimum, constructing an optimal CNN-LSTM model, and carrying out predictive analysis on the screened factor set.
2. The concrete dam displacement deep learning prediction method considering crack influence as claimed in claim 1, wherein: the initial modeling factor set includes a water pressure component, a temperature component, an aging component, and a fracture component;
the displacement monitoring model HSCT considering the crack influence is established, and the mathematical expression is as follows:
wherein H represents the depth of water in front of the dam; a, a 0 Representing constant terms; a, a i ，b 1i ，b 2i ，c 1 ，c 2 ，d i Representing fitting coefficients; m represents a coefficient, taking 3 or 4; n represents a time period, 1 or 2; t represents the accumulated number of days from the displacement monitoring date to the initial displacement monitoring date; θ=t/100; p represents the number of the seam gauges, J i The degree of fracture opening and closing of each joint meter is shown.
3. The concrete dam displacement deep learning prediction method considering crack influence as claimed in claim 2, wherein: the primary selection factor set is obtained by matching the primary selection factor with an observed displacement value; the observed displacement value is a displacement specific value measured by monitoring points arranged on the dam body.
4. The concrete dam displacement deep learning prediction method considering crack influence as claimed in claim 3, wherein: the factor removing step of the variable selecting method mRMR comprises the following steps:
calculating mutual information between two variables:
wherein x is i Represents the influence factor of displacement, y represents the measured displacement, p (x i ) And p (y) represents x i And an edge probability density of y; p (x) i Y) equals the joint probability density;
representing D as x i And mutual information I (x) i Average value of y):
wherein F represents x i Is equal to the number of features in F;
r is represented as x i And redundancy between y:
the operator phi is defined to combine the operations D and R:
let feature set F n-1 Is from the full feature set T m The extracted n-1 feature components are used for searching the optimal feature set by using an incremental search method:
and sequentially searching the features with the maximum mRMR value from the features, calculating the value phi, eliminating the corresponding influence factors when the value phi is smaller than 2, forming a candidate feature set, and taking the candidate feature set as the input of a subsequent prediction model.
5. The concrete dam displacement deep learning prediction method considering crack influence as claimed in claim 4, wherein: the factor eliminating step of the variable selecting method Lasso comprises the following steps:
compressing the coefficients of the variables and making some regression coefficients become 0 by constructing a penalty function; the mathematical expression is as follows:
wherein X represents a displacement influence factor, Y represents a displacement monitoring value, lambda represents an adjustment parameter, and a penalty term ism is more than or equal to 0, and beta represents the regression coefficient of the variable;
adding L in a linear model 1 Penalty term, lasso estimation for linear model:
order theWhen t is less than t 0 A portion of the coefficients are compressed to 0, thereby reducing the dimension of X;
after Lasso calculation, the regression coefficient of part of factors is changed into 0, and the factors with the regression coefficient of 0 are removed to form an input data set of a subsequent prediction model.
6. The concrete dam displacement deep learning prediction method considering crack influence as claimed in claim 5, wherein: the loss function MSE is defined as:
wherein n represents the number of prediction data,represents the ith monitoring data,/-)>Representing the ith prediction data.
7. The concrete dam displacement deep learning prediction method considering crack influence as claimed in claim 6, wherein: the method for establishing the deep learning model CNN-LSTM comprises the following steps:
taking the current day value corresponding to the displacement influence factor as the input of the CNN-LSTM model, and taking the current day measured displacement as the output of the CNN-LSTM model; and (3) at the same time node, actually measured data of each influence factor and displacement are calculated according to 4:1 are divided to be respectively used as a training set and a testing set, wherein the training set is used for training the corresponding CNN-LSTM model, and the testing set is used for evaluating the prediction precision of the corresponding CNN-LSTM model;
by calculating the Root Mean Square Error (RMSE), mean Absolute Error (MAE) and square correlation coefficient (R) of measured value and predicted value in the test set 2 Three indexes are used for verifying the accuracy of the optimal CNN-LSTM model, and the calculation formulas of the three indexes are as follows:
wherein n represents a time series length, y i The monitoring data is represented by a representation of the data,representing predicted values +.>Mean value representing the sequence of observed data, +.>Representing the average of the predicted data sequence.
8. A concrete dam displacement deep learning prediction system taking into account crack influence by using the method as set forth in any one of claims 1 to 7, characterized in that:
the factor set generation module is used for acquiring monitoring data of the concrete dam, obtaining an initial modeling factor set, taking the crack opening and closing degree as an influence factor of displacement, and establishing a displacement monitoring model HSCT taking the crack influence into consideration;
based on a HSCT regression model, obtaining an influence factor set considering cracks, and matching the factors with an observed displacement value to obtain a primary factor set;
the screening module is used for screening influence factors of displacement by adopting variable selection methods mRMR and Lasso, eliminating modeling factors which do not meet the threshold requirement, and constructing the rest modeling factors into a screened factor set;
and the prediction module is used for constructing a deep learning model CNN-LSTM with a mean square error MSE as a loss function, optimizing the model by using an Adam algorithm, obtaining an optimal parameter combination when the MSE is minimum, constructing an optimal CNN-LSTM model, and carrying out prediction analysis on the screened factor set.
9. A computer device, comprising: a memory and a processor; the memory stores a computer program characterized in that: and the processor realizes the concrete dam displacement deep learning prediction method considering the influence of cracks when executing the computer program.
10. A computer-readable storage medium having stored thereon a computer program, characterized by: the computer program, when executed by a processor, implements the steps of a concrete dam displacement deep learning prediction method taking the influence of cracks into account.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311777157.0A CN117636065A (en) | 2023-12-22 | 2023-12-22 | Concrete dam displacement deep learning prediction method and system considering crack influence |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311777157.0A CN117636065A (en) | 2023-12-22 | 2023-12-22 | Concrete dam displacement deep learning prediction method and system considering crack influence |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117636065A true CN117636065A (en) | 2024-03-01 |
Family
ID=90016370
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311777157.0A Pending CN117636065A (en) | 2023-12-22 | 2023-12-22 | Concrete dam displacement deep learning prediction method and system considering crack influence |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117636065A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118606657A (en) * | 2024-08-07 | 2024-09-06 | 长江水利委员会长江科学院 | Dam deformation prediction method, system, equipment and storage medium |
| CN118606657B (en) * | 2024-08-07 | 2024-10-25 | 长江水利委员会长江科学院 | Dam deformation prediction method, system, equipment and storage medium |
-
2023
- 2023-12-22 CN CN202311777157.0A patent/CN117636065A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118606657A (en) * | 2024-08-07 | 2024-09-06 | 长江水利委员会长江科学院 | Dam deformation prediction method, system, equipment and storage medium |
| CN118606657B (en) * | 2024-08-07 | 2024-10-25 | 长江水利委员会长江科学院 | Dam deformation prediction method, system, equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113298288A (en) | Power supply station operation and maintenance cost prediction method integrating time sequence and neural network | |
| CN109271319B (en) | Software fault prediction method based on panel data analysis | |
| CN113822499B (en) | Train spare part loss prediction method based on model fusion | |
| Zhu et al. | Res-HSA: Residual hybrid network with self-attention mechanism for RUL prediction of rotating machinery | |
| CN117149581A (en) | Abnormal information analysis and early warning method and system based on association rule algorithm | |
| CN118211882B (en) | Product quality management system and method based on big data | |
| CN117029968A (en) | Traffic data diagnosis method, system, storage medium and electronic equipment | |
| CN116597939A (en) | Medicine quality control management analysis system and method based on big data | |
| CN115759409A (en) | Water gate deformation prediction method for optimizing LSTM (least Square TM) model by multi-time mode attention mechanism | |
| CN114638696A (en) | Credit risk prediction model training method and system | |
| CN115718880A (en) | Prediction method for degradation stage of complex equipment | |
| CN115906337A (en) | Engineering structure sectional type main aftershock toughness analysis method and system | |
| CN117633475A (en) | Analog method-based oilfield single well yield prediction method and device | |
| Kumar et al. | Cryptocurrency price forecasting in a volatile landscape: Sarimax modeling and short-term strategies | |
| CN117076856A (en) | Defect root cause positioning method, system, equipment and storage medium | |
| CN116881640A (en) | Method and system for predicting core extraction degree and computer-readable storage medium | |
| CN116722653A (en) | Dynamic state detection method and system for electric power system | |
| CN117636065A (en) | Concrete dam displacement deep learning prediction method and system considering crack influence | |
| CN114925476B (en) | Rolling bearing degradation track augmentation quaternion prediction method and storage medium | |
| CN115809725A (en) | Multi-factor short-term electric quantity prediction method and device | |
| CN113391987A (en) | Quality prediction method and device for online software system | |
| CN117909717B (en) | Engineering quantity auxiliary acceptance settlement method based on deep learning and data mining | |
| CN118171047B (en) | Filling method and device of missing data, electronic equipment and storage medium | |
| Hou et al. | Unsupervised graph anomaly detection with discriminative embedding similarity for viscoelastic sandwich cylindrical structures | |
| WO2024040801A1 (en) | Transverse wave time difference prediction method and apparatus |
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 |