CN111209620B - Method for predicting residual bearing capacity and crack propagation path of crack-containing structure - Google Patents
Method for predicting residual bearing capacity and crack propagation path of crack-containing structure Download PDFInfo
- Publication number
- CN111209620B CN111209620B CN201911400482.9A CN201911400482A CN111209620B CN 111209620 B CN111209620 B CN 111209620B CN 201911400482 A CN201911400482 A CN 201911400482A CN 111209620 B CN111209620 B CN 111209620B
- Authority
- CN
- China
- Prior art keywords
- crack
- crack propagation
- propagation path
- network
- strength
- 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.)
- Active
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
- G06N3/084—Backpropagation, e.g. using gradient descent
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a prediction method of residual bearing capacity and crack propagation path of a crack-containing structure based on LSTM-cGAN, which comprises the steps of obtaining the strength of the crack-containing structure with different degrees and the crack propagation path under the loading condition through finite element calculation or field actual measurement in a training stage, producing an antagonistic network model and a long-time and short-time memory method based on conditions, and simultaneously training four deep neural networks including a generation network G, a judgment network D, an LSTM network for processing a time sequence and a convolution neural network CNN for judging the strength of the crack structure. After the training is finished, the network G and the LSTM network are generated by inputting the structural crack propagation history measured on site, and the corresponding structural strength and the crack propagation path are predicted. The method can effectively predict the strength and the crack propagation path of the crack-containing structure, and can effectively realize the in-situ nondestructive monitoring of the crack-containing structure.
Description
Technical Field
The invention belongs to the crossing field of civil structure engineering and computer vision, and particularly relates to a prediction method of residual bearing capacity and crack propagation path of a crack-containing structure based on LSTM-cGAN.
Background
The occurrence and propagation of cracks are important risks leading to the reduction of the durability and the failure of the structure, and the catastrophic failure of the structure can be caused. In actual engineering, most reinforced concrete structures are in a crack working state in the operation period, so that monitoring of crack-containing structures and prediction of crack propagation paths and structural strength of the crack-containing structures are important contents for monitoring of the operation period of civil engineering structures.
In a traditional crack propagation path and strength prediction method, a physical model is often established according to the mechanical state at the current moment, and the crack propagation path and strength at the next moment are obtained by iteratively solving a control equation of a structure, but after the structural performance is degraded or when the internal structure of a composite material is unknown, the assumed physical model is often incorrect, so that the simulation effect is not satisfactory.
With the progress of data analysis means and the accumulation of a large amount of past engineering data, the data-driven method for solving engineering problems has been widely used. The invention applies the latest algorithm in deep learning, and can efficiently predict the crack propagation path and the corresponding structural strength according to the monitoring history.
Disclosure of Invention
In order to solve the technical problem of predicting crack propagation path and strength of crack structure, the invention provides a method for predicting residual bearing capacity and crack propagation path of crack structure based on LSTM-cGAN, which comprises the following steps:
a prediction method of residual bearing capacity and crack propagation path of a crack-containing structure based on LSTM-cGAN comprises the following steps:
(1) firstly, the intensities of crack structures with different degrees and crack propagation paths of the crack structures with different degrees under the loading condition are obtained through finite element calculation or field actual measurement, the crack propagation paths are drawn into a series of images according to color gradient strips, and classification, sorting and sequencing are carried out to establish a sample library.
(2) Generating a confrontation network model based on a long-and-short-term memory method and conditions, and training four deep neural networks through the sample library in the step (1): generating a network G and a judging network D, processing an LSTM network of a time sequence and judging a convolutional neural network CNN of crack structure strength, wherein the LSTM network is built in the generating network G;
(3) and calibrating the crack propagation path of the next time step output by the generated network G through the pixel value, and judging the structural strength of the crack propagation path according to the convolutional neural network CNN.
(4) And (3) in the testing stage, drawing the history of the structural crack propagation path measured on site into an image, inputting the image into the generation network G trained in the step (2), obtaining a predicted crack propagation path and obtaining the corresponding strength at each moment.
Further, the predicted structural strength is specifically tensile strength, compressive strength, bending strength, shear strength and torsional strength of the structure, and the input structural response information is specifically displacement field, strain field, stress field and phase field.
Further, in the step (1), the method for obtaining the crack propagation path and the structural strength at each time point includes:
when the internal condition of the structure is unknown, measuring crack propagation paths of the structure under different load conditions and corresponding structural strength during crack propagation by arranging strain gauges and combining video monitoring;
when the internal condition of the structure is known, the structure response under different load conditions is obtained through finite element phase field theory calculation, and the corresponding crack propagation path and the structure strength at each moment are obtained.
Further, in the step (1), the method for drawing the image according to the color gradient bar specifically includes: and converting the structural response information of each position into a corresponding pixel value according to the color gradient bar jet, and drawing into a picture.
Further, in the step (2), the generating network G is used to generate a crack propagation path at the next time according to the structural crack propagation path history, and the determining network D is used to determine whether the generated structural crack propagation path is correct, and the conditional generation countermeasure network takes the structural crack propagation path history as input, and has an objective function of:
wherein x represents an image of a structural crack propagation path from within the database; y represents the structural crack propagation path history; v (D, G) represents a cost function in the game problem, namely an objective function needing optimization; x to pdata(x) Representing x-obedient structural crack propagation path distribution p in databasedata，
Is shown in pdataDistributing and calculating expectations; x to pz(z) denotes that x obeys a prior distribution pz，pzIs [ -1,1 [ ]]Inner uniform distribution, i.e. z is [ -1,1 [)]A vector of intra-random samples is generated,
is shown in pzDistributing and calculating expectations; d (x | y) represents the output of input x through the judgment network D under the condition of the control parameter y; g (z | y) represents the output image of the input vector z passing through the generation network G under the condition of the control parameter y; d (G (z | y)) represents the output of G (z | y) through the judgment network D.
Further, in the step (2), the LSTM network built in the generation network G is used for processing the convolution part in the generation network to extract the characteristics of the crack propagation path history and predict the crack propagation path at the next time. The LSTM network structure is as follows:
inputting a layer: taking the extracted crack propagation path historical characteristics as input, converting the extracted crack propagation path historical characteristics through a weight and a bias, entering an activation function tanh, converting data into a three-dimensional array as input of entering a hidden layer, wherein the x coordinate is the number of batches of batch training, batch _ size, the y coordinate is the number of data of one batch, namely the number of data of one batch of batch training, the number of the data of one batch of long-time and short-time memory model lstm training is entered, and the z coordinate is cell _ size, namely the number of hidden layer units;
hiding the layer: inputting the processed data and the hidden layer state obtained by the last lstm unit into the lstm unit, if the training is carried out for the first time, initializing the hidden layer to be 0, if the training is not carried out for the first time, continuing to use the parameters of the previous hidden layer, and then obtaining the output data of the hidden layer and the hidden layer state at the moment through an input gate, a forgetting gate and an output gate;
output layer: and performing re-deformation on the output obtained by the hidden layer, and converting the weight and the bias of the output layer to obtain the characteristic of the stress expansion path at the next moment.
Fourthly, other: in the echo propagation in the training period, the adjustment of the weight, the bias and the hidden layer state is determined by a loss function, and each parameter is corrected by back propagation through constructing the loss function and carrying out gradient descent processing.
Further, in the step (3), the method for calibrating the structural response information by the pixel value specifically includes: and converting the pixel value of each position into corresponding structural response information according to the color gradient bar jet.
Further, in the step (3), the method for calibrating the strength of the crack-containing structure produced by the G model specifically comprises the following steps: and (5) obtaining the strength of the composite structure containing the cracks according to CNN training.
The invention has the beneficial effects that:
the method combines the latest algorithm conditions in the field of deep learning to produce the anti-network cGAN and the long-term memory method LSTM, can efficiently predict the crack propagation path and the corresponding structure residual bearing capacity according to the in-situ monitoring history, and solves the problem that the traditional method can not predict the crack propagation path and the corresponding structure strength after the structure performance is degraded according to in-situ monitoring data. The cGAN is the latest algorithm in the field of deep learning, is high in calculation speed and accuracy, and is used for extracting features from a crack propagation history picture containing a crack structure and generating a crack propagation path at the next moment according to the features. The LSTM is the most elegant algorithm for processing the time sequence problem at present, and can solve the shaving disappearance or gradient explosion phenomenon which occurs when the recurrent neural network RNN predicts. And finally, predicting the residual bearing capacity of the structure in the crack propagation process of the structure by adopting a convolutional neural network. The invention applies the latest algorithm in the field of deep learning, can efficiently predict the crack propagation path and the corresponding structure residual bearing capacity according to the in-situ monitoring history in a data driving mode, and provides a new thought for monitoring the structure state of the crack-containing structure.
Drawings
FIG. 1 is a schematic diagram of the overall structure of a conditional generation countermeasure network;
FIG. 2 is a network structure diagram of the generation network G;
FIG. 3 is a diagram illustrating a network structure of the judgment network D;
FIG. 4 is a diagram of a long-short term memory method network architecture;
FIG. 5 is a structural diagram of CNN for predicting structural strength of crack-containing structures;
fig. 6 is a picture of a sample containing crack propagation paths, wherein fig. 6(a) is a model of a crack-containing structure, fig. 6(b) is a structural crack propagation process, and fig. 6(c) is a change in structural counter force and structural strength during loading.
Detailed Description
The present invention will be described in detail below with reference to the accompanying drawings and preferred embodiments, and the objects and effects of the invention will become more apparent. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
The invention provides a prediction method of residual bearing capacity and crack propagation path of a crack-containing structure based on LSTM-cGAN, which applies the latest algorithm condition generation countermeasure network in the field of deep learning, and the condition generation countermeasure network is proved to be capable of efficiently processing images and generating images. The long-time and short-time memory method adopted in the invention is proved to be the most suitable algorithm for processing the time sequence problem in the existing machine learning algorithm. The invention applies the latest algorithm in the field of deep learning and can efficiently predict the crack propagation containing the crack structure and the strength thereof.
The method comprises the following specific steps:
(1) firstly, the strength of crack structures with different degrees and crack propagation paths of the crack structures with different degrees under the loading condition are obtained through finite element calculation, the crack propagation paths are drawn into a series of images according to color gradient strips, and classification, sorting and sequencing are carried out to establish a sample library.
Specifically, the problem of plane strain of mass concrete is solved by Abaqus finite element software in combination with a phase field theory, the strength of 220 groups of concrete structures containing cracks of different degrees and the crack propagation path of 220 groups of concrete structures containing cracks of different degrees under the displacement loading rate of 0.1m/s are obtained through calculation, the concrete structures are drawn into images of sequences according to color gradient strips jet, classification, sorting and sorting are carried out according to a calculation example, 200 groups are randomly taken out to establish a training sample library, and the rest 20 groups are established to establish a verification sample library.
2) Generating a confrontation network model based on a long-time and short-time memory method and conditions, and training four deep neural networks by adopting data of a training sample library: a generation network G (shown in fig. 2) and a judgment network D (shown in fig. 3), an LSTM network (shown in fig. 4) processing the time series and a convolutional neural network CNN (shown in fig. 5) judging the strength of the crack structure, wherein the LSTM network is built in the generation network G.
In this embodiment, four kinds of structural response information, namely, a displacement field, a strain field, a stress field, and a phase field, are used as inputs to train the four deep neural networks.
The generation network G is used for generating a crack propagation path at the next moment according to the structural crack propagation path history, and the judgment network D is used for distinguishing whether the input crack propagation path is from the generation network G or a real crack propagation path, wherein the condition generates a structural crack propagation path history of the countermeasure network, and the objective function is as follows:
wherein x represents an image of a structural crack propagation path from within the database; y represents the structural crack propagation path history; v (D, G) represents a cost function in the game problem, namely an objective function needing optimization; x to pdata(x) Representing x-obedient structural crack propagation path distribution p in databasedata，
Is shown in pdataDistributing and calculating expectations; x to pz(z) denotes that x obeys a prior distribution pz，pzIs [ -1,1 [ ]]In a uniform distribution, i.e.zIs [ -1,1 [ ]]A vector of intra-random samples is generated,
is shown in pzDistributing and calculating expectations; d (x | y) represents the output of input x through the judgment network D under the condition of the control parameter y;G(z | y) denotes an input vectorzAn output image through the generation network G under the condition of the control parameter y; d (G (z | y)) represents the output of G (z | y) through the judgment network D;
the adopted generation network G is a U-shaped network and comprises an Encode part, a long-time memory method network LSTM part and a Decode part, wherein the Encode part is a convolutional neural network, characteristics are extracted from a picture containing crack propagation historical information through eight convolutional layers, and each convolutional layer adopts a 4 x 4 convolutional core; the LSTM part predicts the crack propagation path characteristics of the next moment according to the crack propagation historical characteristics of the previous moments; the Decode part is a deconvolution neural network, which generates crack propagation paths at the next time according to the crack propagation path characteristics at the next time output by the LSTM through eight deconvolution layers, wherein each convolution layer adopts 4 × 4 convolution kernels, as shown in fig. 2.
In the adopted judgment network D, firstly, the picture (condition) including the crack propagation path history and the picture of the predicted crack propagation path at the next time are spliced together, and then, the output of the judgment network D is obtained through the network of five convolution layers and one full connection layer, as shown in fig. 3.
In the convolutional neural network CNN for judging the structural strength, the predicted picture of the crack propagation path at the next time passes through the network of five convolutional layers and one full-link layer, and finally the structural strength at this time is regressed, as shown in fig. 4.
And the long-time memory method network built in the generation network G is used for processing the convolution part in the generation network, extracting the characteristics of the crack propagation path history and predicting the crack propagation path at the next moment. The long-time memory method network structure is as follows:
inputting a layer: taking the extracted crack propagation path historical characteristics as input, converting the extracted crack propagation path historical characteristics through a weight and a bias, entering an activation function tanh, converting data into a three-dimensional array as input of entering a hidden layer, wherein the x coordinate is the number of batches of batch training, batch _ size, the y coordinate is the number of data of one batch, namely the number of data of one batch of batch training, the number of the data of one batch of long-time and short-time memory model lstm training is entered, and the z coordinate is cell _ size, namely the number of hidden layer units;
hiding the layer: inputting the processed data and the hidden layer state obtained by the last lstm unit into the lstm unit, if the training is carried out for the first time, initializing the hidden layer to be 0, and if the training is not carried out for the first time, continuing to use the parameters of the previous hidden layer, and obtaining the output data of the hidden layer and the hidden layer state at the moment through an input gate, a forgetting gate and an output gate;
output layer: and performing re-deformation on the output obtained by the hidden layer, and converting the weight and the bias of the output layer to obtain the characteristic of the stress expansion path at the next moment.
Fourthly, other: in the echo propagation in the training period, the adjustment of the weight, the bias and the hidden layer state is determined by a loss function, and each parameter is corrected by back propagation through constructing the loss function and carrying out gradient descent processing.
In this embodiment, the parameter in the network is set to time _ steps ═ 1; batch _ size ═ 5; input _ size ═ 3; cell _ size ═ 10; the learning rate was 0.006.
(3) And calibrating the structural response information in the generated crack propagation path image at the next time step output by the generated model through the combination of the pixel values and the jet of the color gradient bar, and judging the structural strength of the crack propagation path image according to the convolutional neural network CNN, as shown in FIG. 5. The predicted structural strength may specifically be: tensile strength, compressive strength, bending strength, shear strength, torsional strength of the structure.
(4) And in the testing stage, drawing the history of the structural crack propagation paths with concentrated verification into an image, inputting the image into the generation network G trained in the step 2 to obtain a predicted crack propagation path and obtain the corresponding strength at each moment, as shown in FIG. 6. Wherein fig. 6(a) is a crack-containing structure model, fig. 6(b) is a structure crack propagation process, and fig. 6(c) is a change in structural reaction force and structural strength during loading.
It will be understood by those skilled in the art that the foregoing is only a preferred embodiment of the present invention, and is not intended to limit the invention, and although the invention has been described in detail with reference to the foregoing examples, it will be apparent to those skilled in the art that various changes in the form and details of the embodiments may be made and equivalents may be substituted for elements thereof. All modifications, equivalents and the like which come within the spirit and principle of the invention are intended to be included within the scope of the invention.
Claims (8)
1. The method for predicting the residual bearing capacity and the crack propagation path of the crack-containing structure based on LSTM-cGAN is characterized by comprising the following steps of:
(1) firstly, obtaining the strength of crack structures with different degrees and crack propagation paths of the crack structures with different degrees under the loading condition through finite element calculation or field actual measurement, drawing the crack propagation paths into a series of images according to color gradient strips, classifying, sorting and sequencing the images, and establishing a sample library;
(2) generating a confrontation network model based on a long-and-short-term memory method and conditions, and training four deep neural networks through the sample library in the step (1): generating a network G and a judging network D, processing an LSTM network of a time sequence and judging a convolutional neural network CNN of crack structure strength, wherein the LSTM network is built in the generating network G;
(3) calibrating the crack propagation path of the next time step output by the generation network G through the pixel value, and judging the structural strength of the crack propagation path according to the convolutional neural network CNN;
(4) and (3) in the testing stage, drawing the history of the structural crack propagation path measured on site into an image, inputting the image into the generation network G trained in the step (2), obtaining a predicted crack propagation path and obtaining the corresponding strength at each moment.
2. The method for predicting the residual bearing capacity and crack propagation path of a crack-containing structure according to claim 1, wherein the predicted structural strength is tensile strength, compressive strength, bending strength, shear strength and torsional strength of the structure, and the input structural response information is displacement field, strain field, stress field and phase field.
3. The method for predicting the residual bearing capacity and the crack propagation path of the structure containing the cracks based on the LSTM-cGAN as claimed in claim 1, wherein the method for obtaining the crack propagation path and the structural strength at each moment in the step (1) is specifically as follows:
when the internal condition of the structure is unknown, measuring crack propagation paths of the structure under different load conditions and corresponding structural strength during crack propagation by arranging strain gauges and combining video monitoring;
when the internal condition of the structure is known, the structure response under different load conditions is obtained through finite element phase field theory calculation, and the corresponding crack propagation path and the structure strength at each moment are obtained.
4. The method for predicting the residual bearing capacity and the crack propagation path of a crack-containing structure based on LSTM-cGAN according to claim 1, wherein the step (1) of drawing the crack-containing structure into an image according to a color gradient bar specifically comprises: and converting the structural response information of each position into a corresponding pixel value according to the color gradient bar jet, and drawing into a picture.
5. The method for predicting the residual bearing capacity and crack propagation path of a crack-containing structure based on LSTM-cGAN as claimed in claim 1, wherein in the step (2), the generated network G is used to generate the crack propagation path at the next time according to the structure crack propagation path history, and the network D is used to determine whether the generated structure crack propagation path is correct or not, and the conditionally generated countermeasure network takes the structure crack propagation path history as input, and has an objective function as:
wherein x represents an image of a structural crack propagation path from within the database; y represents the structural crack propagation path history; v (D, G) represents a cost function in the game problem, namely an objective function needing optimization; x to pdata(x) Representing x-obedient structural crack propagation path distribution p in databasedata，
Is shown in pdataDistributing and calculating expectations; x to pz(z) denotes that x obeys a prior distribution pz，pzIs [ -1,1 [ ]]Inner uniform distribution, i.e. z is [ -1,1 [)]A vector of intra-random samples is generated,
is shown in pzDistributing and calculating expectations; d (x | y) represents the output of input x through the judgment network D under the condition of the control parameter y; g (z | y) represents the output image of the input vector z passing through the generation network G under the condition of the control parameter y; d (G (z | y)) represents the output of G (z | y) through the judgment network D.
6. The method for predicting the residual bearing capacity and the crack propagation path of the crack-containing structure based on the LSTM-cGAN as claimed in claim 1, wherein in the step (2), the LSTM network is built in the generation network G, and is used for processing the convolution part in the generation network to extract the characteristics of the crack propagation path history and predict the crack propagation path at the next moment; the LSTM network structure is as follows:
inputting a layer: taking the extracted crack propagation path historical characteristics as input, converting the extracted crack propagation path historical characteristics through a weight and a bias, entering an activation function tanh, converting data into a three-dimensional array as input of entering a hidden layer, wherein the x coordinate is the number of batches of batch training, batch _ size, the y coordinate is the number of data of one batch, namely the number of data of one batch of batch training, the number of the data of one batch of long-time and short-time memory model lstm training is entered, and the z coordinate is cell _ size, namely the number of hidden layer units;
hiding the layer: inputting the processed data and the hidden layer state obtained by the last lstm unit into the lstm unit, if the training is carried out for the first time, initializing the hidden layer to be 0, if the training is not carried out for the first time, continuing to use the parameters of the previous hidden layer, and then obtaining the output data of the hidden layer and the hidden layer state at the moment through an input gate, a forgetting gate and an output gate;
output layer: performing re-deformation on the output obtained by the hidden layer, and converting the weight and the offset of the output layer to obtain the characteristic of a stress expansion path at the next moment;
fourthly, other: in the echo propagation in the training period, the adjustment of the weight, the bias and the hidden layer state is determined by a loss function, and each parameter is corrected by back propagation through constructing the loss function and carrying out gradient descent processing.
7. The method for predicting the residual bearing capacity and the crack propagation path of the crack-containing structure based on the LSTM-cGAN of claim 1, wherein the method for calibrating the structure response information by the pixel value in the step (3) is specifically as follows: and converting the pixel value of each position into corresponding structural response information according to the color gradient bar jet.
8. The method for predicting the residual bearing capacity and the crack propagation path of the crack-containing structure based on the LSTM-cGAN of claim 1, wherein in the step (3), the method for calibrating the strength of the crack-containing structure produced by the G model specifically comprises: and (5) obtaining the strength of the composite structure containing the cracks according to CNN training.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911400482.9A CN111209620B (en) | 2019-12-30 | 2019-12-30 | Method for predicting residual bearing capacity and crack propagation path of crack-containing structure |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911400482.9A CN111209620B (en) | 2019-12-30 | 2019-12-30 | Method for predicting residual bearing capacity and crack propagation path of crack-containing structure |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111209620A CN111209620A (en) | 2020-05-29 |
| CN111209620B true CN111209620B (en) | 2021-11-16 |
Family
ID=70787674
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911400482.9A Active CN111209620B (en) | 2019-12-30 | 2019-12-30 | Method for predicting residual bearing capacity and crack propagation path of crack-containing structure |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111209620B (en) |
Families Citing this family (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111783338B (en) * | 2020-06-30 | 2023-04-21 | 平安国际智慧城市科技股份有限公司 | Microstructure metal intensity distribution prediction method and device based on artificial intelligence |
| CN112051142B (en) * | 2020-08-03 | 2021-05-18 | 武汉大学 | Universal phase field method for simulating different failure modes of brittle material |
| CN112784331B (en) * | 2020-09-25 | 2022-09-13 | 汕头大学 | Soil stress-strain relation determination method based on improved LSTM deep learning method |
| CN112149356A (en) * | 2020-09-27 | 2020-12-29 | 湖南大学 | Method, device, equipment and medium for predicting structural crack propagation path |
| CN112347531B (en) * | 2020-09-27 | 2023-06-30 | 山东师范大学 | Brittle marble Dan Sanwei crack propagation path prediction method and system |
| CN114048790B (en) * | 2021-08-01 | 2024-05-24 | 北京工业大学 | Road base strain analysis method based on coupling bidirectional LSTM and convolution structure network |
| CN114065650B (en) * | 2021-11-26 | 2024-09-24 | 重庆邮电大学 | Material crack tip multi-scale strain field measurement tracking method based on deep learning |
| CN118196542B (en) * | 2024-05-16 | 2024-08-06 | 湖南工程学院 | Crack propagation path prediction method, crack propagation path prediction device, crack propagation path prediction storage medium, and crack propagation path prediction product |
Family Cites Families (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20180065498A (en) * | 2016-12-08 | 2018-06-18 | 한국항공대학교산학협력단 | Method for deep learning and method for generating next prediction image using the same |
| US11275989B2 (en) * | 2017-05-22 | 2022-03-15 | Sap Se | Predicting wildfires on the basis of biophysical indicators and spatiotemporal properties using a long short term memory network |
| CN109102493A (en) * | 2018-07-03 | 2018-12-28 | 柳州市木子科技有限公司 | A kind of automobile metal plate work flaw detection system based on CNN and LR |
| CN109166126B (en) * | 2018-08-13 | 2022-02-18 | 苏州比格威医疗科技有限公司 | Method for segmenting paint cracks on ICGA image based on condition generation type countermeasure network |
| CN109345507B (en) * | 2018-08-24 | 2021-07-13 | 河海大学 | Dam image crack detection method based on transfer learning |
| CN109241609B (en) * | 2018-08-31 | 2022-11-29 | 华东交通大学 | Bayesian dynamic prediction method based on Markov chain Monte Carlo |
| CN109447373A (en) * | 2018-11-16 | 2019-03-08 | 上海海事大学 | Haze method is predicted based on the LSTM neural network of python platform |
| CN110321815A (en) * | 2019-06-18 | 2019-10-11 | 中国计量大学 | A kind of crack on road recognition methods based on deep learning |
-
2019
- 2019-12-30 CN CN201911400482.9A patent/CN111209620B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN111209620A (en) | 2020-05-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111209620B (en) | Method for predicting residual bearing capacity and crack propagation path of crack-containing structure | |
| Huang et al. | A machine learning based plasticity model using proper orthogonal decomposition | |
| Novák et al. | ANN inverse analysis based on stochastic small-sample training set simulation | |
| Zhang et al. | Vibration-based assessment of delaminations in FRP composite plates | |
| Chojaczyk et al. | Review and application of artificial neural networks models in reliability analysis of steel structures | |
| CN114117840B (en) | Structural performance prediction method based on simulation and test data hybrid drive | |
| Sepasdar et al. | A data-driven approach to full-field nonlinear stress distribution and failure pattern prediction in composites using deep learning | |
| CN115238394B (en) | Multi-source uncertainty hybrid reliability digital twin modeling method for composite material structure | |
| CN111209695A (en) | LSTM-based structural dynamic response prediction method | |
| Anton et al. | Physics-informed neural networks for material model calibration from full-field displacement data | |
| CN112784495A (en) | Mechanical structure real-time fatigue life prediction method based on data driving | |
| CN109739178B (en) | Workpiece machining deformation online prediction method driven by monitoring data | |
| CN106568647B (en) | A kind of Strength Forecast of Concrete method neural network based | |
| Böhringer et al. | A strategy to train machine learning material models for finite element simulations on data acquirable from physical experiments | |
| Gao et al. | An artificial neural network method for probabilistic life prediction of corroded reinforced concrete | |
| CN111210409B (en) | Condition-based generation confrontation network structure damage identification method | |
| Lanning et al. | Applicability of convolutional neural networks for calibration of nonlinear dynamic models of structures | |
| Solorzano et al. | ANN-based surrogate model for predicting the lateral load capacity of RC shear walls | |
| Bilgehan et al. | Buckling load estimation of cracked columns using artificial neural network modeling technique | |
| CN111241614B (en) | Engineering structure load inversion method based on condition generation confrontation network model | |
| CN111241725B (en) | Structure response reconstruction method for generating countermeasure network based on conditions | |
| CN112818870A (en) | Method for predicting residual life of bearing based on gated neural network framework | |
| Wang et al. | Improved deep neural network for predicting structural response of stiffened cylindrical shells to far-field underwater explosion | |
| CN107145665B (en) | Roadway surrounding rock stress modeling and prediction method | |
| EP3789956A1 (en) | Method and system for analyzing the mechanical response of microstructures under investigation |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |