CN116663126A - Bridge temperature effect prediction method based on channel attention BiLSTM model - Google Patents
Bridge temperature effect prediction method based on channel attention BiLSTM model Download PDFInfo
- Publication number
- CN116663126A CN116663126A CN202310766544.8A CN202310766544A CN116663126A CN 116663126 A CN116663126 A CN 116663126A CN 202310766544 A CN202310766544 A CN 202310766544A CN 116663126 A CN116663126 A CN 116663126A
- Authority
- CN
- China
- Prior art keywords
- bridge
- temperature
- output
- bilstm
- layer
- 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
- 230000002277 temperature effect Effects 0.000 title claims abstract description 50
- 238000000034 method Methods 0.000 title claims abstract description 41
- 238000012544 monitoring process Methods 0.000 claims abstract description 34
- 238000012549 training Methods 0.000 claims abstract description 20
- 230000007246 mechanism Effects 0.000 claims abstract description 14
- 238000013528 artificial neural network Methods 0.000 claims abstract description 8
- 238000004364 calculation method Methods 0.000 claims abstract description 8
- 230000008569 process Effects 0.000 claims abstract description 6
- 230000006870 function Effects 0.000 claims description 18
- 230000004044 response Effects 0.000 claims description 18
- 238000001914 filtration Methods 0.000 claims description 8
- 238000003062 neural network model Methods 0.000 claims description 8
- 230000004913 activation Effects 0.000 claims description 7
- 230000015654 memory Effects 0.000 claims description 7
- 230000002457 bidirectional effect Effects 0.000 claims description 6
- 238000006073 displacement reaction Methods 0.000 claims description 6
- 238000012952 Resampling Methods 0.000 claims description 5
- 230000009471 action Effects 0.000 claims description 5
- 238000007781 pre-processing Methods 0.000 claims description 5
- 238000005259 measurement Methods 0.000 claims description 4
- 238000005457 optimization Methods 0.000 claims description 4
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 claims description 3
- 238000004422 calculation algorithm Methods 0.000 claims description 3
- 239000000470 constituent Substances 0.000 claims description 2
- 238000009826 distribution Methods 0.000 claims description 2
- 230000009467 reduction Effects 0.000 claims description 2
- 238000004458 analytical method Methods 0.000 abstract description 9
- 238000004088 simulation Methods 0.000 abstract description 7
- 230000007774 longterm Effects 0.000 abstract description 6
- 238000013507 mapping Methods 0.000 abstract description 5
- 230000006399 behavior Effects 0.000 abstract description 4
- 230000009286 beneficial effect Effects 0.000 abstract description 4
- 238000013135 deep learning Methods 0.000 abstract description 3
- 238000004514 thermodynamic simulation Methods 0.000 abstract description 2
- 238000012360 testing method Methods 0.000 description 6
- 230000000694 effects Effects 0.000 description 5
- 238000012795 verification Methods 0.000 description 4
- 230000007787 long-term memory Effects 0.000 description 3
- 230000002159 abnormal effect Effects 0.000 description 2
- 230000006403 short-term memory Effects 0.000 description 2
- 230000008878 coupling Effects 0.000 description 1
- 238000010168 coupling process Methods 0.000 description 1
- 238000005859 coupling reaction Methods 0.000 description 1
- 238000005336 cracking Methods 0.000 description 1
- 238000013136 deep learning model Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000002542 deteriorative effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 238000005728 strengthening Methods 0.000 description 1
- 238000012731 temporal analysis Methods 0.000 description 1
- 238000000700 time series analysis Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/10—Geometric CAD
- G06F30/13—Architectural design, e.g. computer-aided architectural design [CAAD] related to design of buildings, bridges, landscapes, production plants or roads
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
- G06F2119/02—Reliability analysis or reliability optimisation; Failure analysis, e.g. worst case scenario performance, failure mode and effects analysis [FMEA]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
- G06F2119/08—Thermal analysis or thermal optimisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
- G06F2119/14—Force analysis or force optimisation, e.g. static or dynamic forces
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Evolutionary Computation (AREA)
- Geometry (AREA)
- General Physics & Mathematics (AREA)
- Computer Hardware Design (AREA)
- General Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- Structural Engineering (AREA)
- Software Systems (AREA)
- Mathematical Optimization (AREA)
- Mathematical Analysis (AREA)
- Computational Mathematics (AREA)
- Civil Engineering (AREA)
- Medical Informatics (AREA)
- Pure & Applied Mathematics (AREA)
- Architecture (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a bridge temperature effect prediction method based on a channel attention BiLSTM model, which adopts a sequence-to-sequence deep learning method to directly extract a nonlinear mapping relation between a temperature field and temperature strain from monitoring data of a bridge, and replaces the modeling process of the traditional numerical simulation. The channel attention mechanism introduced into the time sequence improves the BiLSTM model, so that the BiLSTM model has good generalization performance and prediction accuracy in temperature effect prediction. The method overcomes the calculation time-consuming problem of bridge thermodynamic simulation analysis, and is beneficial to analyzing thermodynamic behavior of a large bridge under long-term service conditions; the mapping relation between the temperature field and the temperature effect is directly constructed according to the bridge monitoring data, so that the utilization rate of the bridge monitoring data is improved, and the current situation of large-scale popularization and application of the bridge monitoring system is met; the used deep neural network has good generalization capability and applicability in temperature effect prediction, does not need a large amount of monitoring data in a complete period for training, and is simple and convenient to use.
Description
Technical Field
The invention relates to the crossing field of bridge temperature effect analysis and computer science, in particular to a bridge temperature effect prediction method based on a channel attention BiLSTM model.
Background
The bridge is used as a pivot engineering for transportation, and the operation safety of the bridge is ensured, so that the bridge has strategic significance for ensuring smooth regional transportation and ordered social and economic development. However, the bridge needs to face the temperature effect of reciprocating circulation in the operation stage, and the generated isothermal effects of deformation and stress even exceed the load effects of automobile load, wind load and the like, so that the bridge is one of important factors of concrete cracking, stress fatigue and support deflection. Therefore, in order to prevent the bridge from significantly deteriorating under the action of temperature, it is necessary to track and evaluate the temperature field and the temperature effect of the bridge.
Various scholars have proposed a series of bridge temperature effect analysis methods based on numerical simulation, which cover various aspects such as material thermal performance calculation, environment-structure thermodynamic coupling analysis, sunlight intensity and angle transformation tracking simulation, and can accurately simulate structural mechanical behaviors under time-varying and three-dimensional temperature fields. Numerical simulation, however, involves complex modeling procedures and parameter settings, and the reliability of the results depends on the expertise and analysis experience of the user. Moreover, the numerical model of the large bridge has the characteristics of numerous degrees of freedom, complex thermodynamic boundary conditions and the like, so that the simulation analysis is time-consuming seriously. Therefore, there are still many difficulties and challenges in evaluating the long-term temperature effects of bridge structures through numerical simulation.
The fast-developed machine learning and deep learning method is hopeful to directly extract the mapping relation between the bridge temperature effect and the temperature field from the measured data, can avoid the problems of modeling errors, parameter uncertainty and the like of numerical simulation, and provides a new tool for analyzing the long-term temperature effect of the bridge. In particular, in recent years, bridge health monitoring systems are advocated and developed greatly by the countries, and massive data accumulated by the monitoring systems bring wide application prospects to the data driving method.
Currently, only a few students attempt to apply deep learning to proxy modeling of bridge temperature effects. Existing studies typically employ extremely large monitoring data (more than one year of complete data) to analyze long-term temperature effects; alternatively, less monitoring data is used to model the temperature effect, but the temperature difference between the training set data and the test set data is not obvious. Moreover, the monitoring data often have a plurality of abnormal values due to various interference factors. After abnormal values are removed, the monitored data lacks part of temperature intervals and corresponding temperature effects, so that the learnable characteristics of the deep learning model are lost. However, the existing researches generally adopt classical time series analysis models such as RNN, LSTM and the like, and do not have the capability of acquiring essential characteristics of temperature effects from incomplete monitoring data, so that the actual prediction effect in the temperature effect modeling is poor.
Disclosure of Invention
Technical problems: aiming at the defects of the prior art, the invention provides a bridge temperature effect prediction method based on a channel attention BiLSTM model, which can directly establish a mapping relation between a temperature field and a temperature effect by using monitoring data in a limited and incomplete period, thereby assisting in predicting a long-term evolution rule of the bridge temperature effect and greatly reducing analysis time consumption of the thermodynamic behavior of the bridge.
The technical scheme is as follows: in order to achieve the above purpose, the present invention provides a bridge temperature effect prediction method based on a channel attention BiLSTM model, which comprises the following steps:
step S1, preprocessing monitoring data of a temperature field and structural response, wherein the structural response comprises stress strain and displacement;
s2, determining input and output dimensions of a deep neural network according to the input temperature field and the channel number of temperature effect data, and establishing a two-way long-short-term memory network of a channel attention mechanism;
step S3, a data set is established according to the preprocessing result of the step S1, wherein the constituent elements of the data set are temperature and corresponding structural response;
s4, training the deep neural network model by using the established data set, wherein during training, data input into the neural network model is the temperature measured on the surface or in the bridge structure, and output is stress strain and displacement generated by the bridge under the action of the temperature;
and S5, re-collecting the temperature measured on the surface or in the bridge structure, and inputting the temperature into the neural network model trained in the step S4 to obtain the stress strain and displacement generated under the corresponding temperature action.
Further, in step 1, the pretreatment process includes: the method comprises the steps of (1) filtering and denoising monitoring data, (2) aligning a temperature field with the signal length of the structural response monitoring data, (3) extracting a temperature effect, namely structural response, from the structural response monitoring data, and (4) resampling the temperature field data and the temperature effect data.
Furthermore, a low-pass filter with the cutoff frequency of 0.5Hz is adopted for filtering and noise reduction, the signal length of the two types of monitoring data of the temperature field and the structural response is aligned in a cut-off or zero-filling mode, and the temperature effect is extracted from the monitoring data of the structural response by a median filtering method of a moving window.
Further, the interval of resampling is 10min.
Further, the temperature is obtained by a measuring element which is tightly attached to the surface of the bridge structure or embedded in the bridge structure.
Further, in step S2, the bidirectional long-short-term memory network of the channel attention mechanism includes an input layer, two bidirectional long-short-term memory layers, i.e. a BiLSTM layer, a channel attention mechanism module, i.e. a CA module, and a full connection layer as output layers, and the activation function between the hidden layers is a ReLU function;
during training, bridge temperature field data input into a network are respectively sent to a first BiLSTM layer and a CA module; after the output of the first BiLSTM layer is subjected to a ReLU activation function, the output is respectively sent to the second BiLSTM layer and the CA module; the CA module calculates attention weight according to the initial input of the network and the output of the first BiLSTM layer; and after the output of the second BiLSTM layer is subjected to a ReLU activation function, adding the output with initial input temperature data of the network, weighting according to the attention weight output by the CA module, and sending the obtained result into the FC layer to obtain a prediction result of the output network model, namely the structural response of the bridge.
Further, the channel attention mechanism module comprises two FC layers and a weight calculation module based on cosine similarity, wherein the distance between the output of the first biplstm layer and the output of the second FC layer in the CA module is measured by the cosine similarity, the attention weight is calculated according to the distance measurement, the degree of the channel bias is enhanced by the weight distribution strengthening network model, and the attention weight of the CA module is calculated according to the formula (1) -formula (3):
α i ＝σ(D i )(1)
wherein alpha is i Attention mechanism weight for the ith channel, D i For the average cosine similarity corresponding to the ith channel, d ij For the cosine distance between the ith hidden layer output of the second BiLSTM layer and the jth hidden layer output of the FC layer in the CA module, σ () is a Sigmoid function, n f Is the number of channels output by the first BiLSTM layer; f is the output of the first BiLSTM layer, f i The output of the ith channel of the layer; h is the output of the second FC layer in the CA module, h i The output of the ith channel of the layer.
Further, the Mean Square Error (MSE) of the predicted value and the true value is selected as the loss function, that is, equation (4) is adopted as the optimization target of the network parameters:
where J is a loss function, y i Is the predicted value of the temperature effect of the method provided by the invention,is the actual measurement value of the temperature effect, and B is the batch size input by the network model in the training process.
Further, an Adam optimization algorithm is selected to train the neural network model.
The beneficial effects are that: compared with the prior art, the technical scheme of the invention has the following beneficial technical effects:
firstly, the problem of time consumption in calculation of bridge thermodynamic simulation analysis is solved, and the thermodynamic behavior of the large bridge under the long-term service condition is analyzed.
Secondly, the mapping relation between the temperature field and the temperature effect can be directly constructed according to the bridge monitoring data, so that the utilization rate of the bridge monitoring data is effectively improved, and the current situation of large-scale popularization of the bridge monitoring system is met.
And thirdly, compared with the classical time sequence deep neural networks such as an RNN model, an LSTM model and the like, the adopted deep neural network has better generalization capability and applicability in temperature effect prediction, does not need a large amount of monitoring data in a complete period for training, and is simple and convenient to use.
Drawings
FIG. 1 is temperature monitoring data used in the preferred case;
FIG. 2 is strain monitoring data (single day) for the preferred case;
FIG. 3 is a two-way long and short term memory network of the channel attention mechanism of the present invention;
FIG. 4 is a diagram of a two-way long and short term memory network layer (BiLSTM layer) output according to the present invention;
FIG. 5 is a flow chart of attention weight calculation by the channel attention mechanism module;
FIG. 6 is a comparison of time course results of the method of the present invention with a classical network model;
FIG. 7 is a comparison of the prediction error of the method of the present invention with a classical network model.
Detailed Description
The objects, procedures and effects of the present invention will be more clearly and clearly understood from the following detailed description of the present invention with reference to the accompanying drawings and preferred embodiments. It should be noted that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the invention.
The method is characterized in that a large-span bridge is used as a preferable case, and the monitoring data period is 1 month to 9 months of a certain year. In the case, the network input is temperature field data of a bridge; the network output is temperature strain. The temperature field measuring points are selected from T1-T10 positioned on the middle section of the main span, namely, the characteristic dimension of network input is 10 (the number of channels is 10); and the strain measuring points are selected from S1-S10 positioned on the middle section of the main span and S21-S30 positioned on the support section of the main span, namely, the characteristic dimension of network output is 20. (1) step S1: the temperature field and the original strain monitoring data are preprocessed. Firstly, filtering by adopting a low-pass filter with cutoff frequency of 0.5Hz to remove high-frequency noise; secondly, extracting temperature strain by adopting median filtering with a sliding window length of 20 min; then, aligning the temperature field data with the strain data in a cutting-off, zero filling and other modes; and finally, resampling the monitoring data, wherein the new sampling interval is 10min. After pretreatment, the temperature field and temperature strain data of the bridge are shown in fig. 1 and 2, respectively.
(2) Step S2: according to the characteristic dimension of network input and output, a bidirectional long-short-term memory network of a channel attention mechanism shown in figure 3 is established, and the bidirectional long-term memory network comprises an input layer, two BiLSTM layers, a CA module and an FC layer as output layers, wherein an activation function between hidden layers is a ReLU function. Wherein the batch size of the sample is denoted as n b The method comprises the steps of carrying out a first treatment on the surface of the The time sequence length of the network input is recorded as n t The feature dimension is denoted as n i The method comprises the steps of carrying out a first treatment on the surface of the The characteristic dimension of the network output is recorded as n o The method comprises the steps of carrying out a first treatment on the surface of the The hidden layer output dimension of the first BiLSTM layer is denoted as n f 。
Output of the BiLSTM layer as shown in fig. 4, bi-directional output of the BiLSTM layer is averaged for each time position,thus obtaining BiLSTM layer output in the network model; meanwhile, the hidden layer output dimension of the second BiLSTM layer is consistent with the characteristic dimension of the network input, namely n in the case i 10. The CA module comprises two FC layers and a cosine similarity calculation module; the cosine similarity calculation module is shown in fig. 5, and calculates cosine similarity of the output of the second layer BiLSTM layer of the backbone part of the network and the output of the second layer FC layer in the module, and further calculates average cosine similarity of each channel, so as to solve attention weights of each channel. Specifically, the attention weight of each channel is calculated according to the formula (1) -formula (3):
α i ＝σ(D i )(1)
(3) Step S3: a database required for the deep neural network is prepared. Based on the preprocessed data of step S1, a window length S is used w All monitoring data were split into 448 equal length sequences for a fixed window of 200. After that, the divided sequence is as follows: 1: the proportion of 2 is divided into a training set, a verification set and a test set. The training set and the verification set are mainly 1-7 month data, and the test set is 8-9 month data. Therefore, the temperature interval corresponding to the training set and the verification set is lower than the temperature interval corresponding to the test set.
(4) Step S4: model training of deep neural networks. And selecting a Mean Square Error (MSE) shown in a formula (4) as a function, selecting an Adam algorithm by an optimizer, wherein the initial learning rate is 0.001, and the batch processing size is the number of samples of the whole training set. Before training, the temperature field and the temperature strain data are normalized to the [ -1,1] interval respectively.
Where J is a loss function, y i Is the predicted value of the temperature effect of the method provided by the invention,is the actual measurement value of the temperature effect, and B is the batch size input by the network model in the training process.
(5) Step S5: model verification of deep neural network. The prediction accuracy of the network model is analyzed by adopting Root Mean Square Error (RMSE) between the predicted value and the measured value as an evaluation index, and the network model provided by the invention is compared with a classical LSTM network. In order to strictly prove the beneficial effects of the method, the scheme example repeatedly trains the two network models for 10 times, adopts the RMSE median of the 10 training results to evaluate the prediction precision, and eliminates the influence caused by random initialization of the network parameters. Fig. 6 and 7 show the comparison of strain time course prediction and prediction accuracy: it can be seen that the classical LSTM model fails on the test set because the temperature interval of the training set data is lower than the test set. In contrast, the network model provided by the invention still shows good prediction precision and can be used as a proxy model for bridge temperature effect analysis.
Claims (9)
1. A bridge temperature effect prediction method based on a channel attention BiLSTM model is characterized by comprising the following steps:
step S1, preprocessing monitoring data of a temperature field and structural response, wherein the structural response comprises stress strain and displacement;
s2, determining input and output dimensions of a deep neural network according to the input temperature field and the channel number of temperature effect data, and establishing a two-way long-short-term memory network of a channel attention mechanism;
step S3, a data set is established according to the preprocessing result of the step S1, wherein the constituent elements of the data set are temperature and corresponding structural response;
s4, training the deep neural network model by using the established data set, wherein during training, data input into the neural network model is the temperature measured on the surface or in the bridge structure, and output is stress strain and displacement generated by the bridge under the action of the temperature;
and S5, re-collecting the temperature measured on the surface or in the bridge structure, and inputting the temperature into the neural network model trained in the step S4 to obtain the stress strain and displacement generated under the corresponding temperature action.
2. The bridge temperature effect prediction method based on the channel attention BiLSTM model according to claim 1, wherein in step 1, the preprocessing process comprises: the method comprises the steps of (1) filtering and denoising monitoring data, (2) aligning a temperature field with the signal length of the structural response monitoring data, (3) extracting a temperature effect, namely structural response, from the structural response monitoring data, and (4) resampling the temperature field data and the temperature effect data.
3. The bridge temperature effect prediction method based on the channel attention BiLSTM model according to claim 2, wherein a low-pass filter with a cutoff frequency of 0.5Hz is adopted for filtering and noise reduction, the signal lengths of two types of monitoring data of a temperature field and a structural response are aligned in a cut-off or zero-filling mode, and the temperature effect is extracted from the monitoring data of the structural response through a median filtering method of a moving window.
4. A bridge temperature effect prediction method based on a channel attention BiLSTM model according to claim 2 or 3, wherein the resampling interval is 10min.
5. The bridge temperature effect prediction method based on the channel attention BiLSTM model according to claim 1, wherein the temperature is obtained by a measuring element which is tightly attached to the surface of the bridge structure or embedded in the bridge structure.
6. The bridge temperature effect prediction method based on the channel attention BiLSTM model according to claim 1, wherein in step S2, the bidirectional long-short-term memory network of the channel attention mechanism comprises an input layer, two bidirectional long-short-term memory layers, namely a BiLSTM layer, a channel attention mechanism module, namely a CA module, and a full connection layer as output layers, and an activation function between hidden layers is a ReLU function;
during training, bridge temperature field data input into a network are respectively sent to a first BiLSTM layer and a CA module; after the output of the first BiLSTM layer is subjected to a ReLU activation function, the output is respectively sent to the second BiLSTM layer and the CA module; the CA module calculates attention weight according to the initial input of the network and the output of the first BiLSTM layer; and after the output of the second BiLSTM layer is subjected to a ReLU activation function, adding the output with initial input temperature data of the network, weighting the added data according to the attention weight output by the CA module, and sending the obtained result into the FC layer to obtain a prediction result of the output network model, namely the structural response of the bridge.
7. The bridge temperature effect prediction method based on the channel attention BiLSTM model according to claim 6, wherein the channel attention mechanism module comprises two FC layers and a weight calculation module based on cosine similarity, wherein the distance between the output of the first BiLSTM layer and the output of the second FC layer in the CA module is measured by the cosine similarity, the attention weight is calculated according to the distance, the degree of the bias of the network model to the channel is enhanced by weight distribution, and the attention weight of the CA module is calculated according to formula (1) -formula (3):
α i ＝σ(D i )(1)
wherein alpha is i Attention mechanism weight for the ith channel, D i For the average cosine similarity corresponding to the ith channel, d ij For the cosine distance between the ith hidden layer output of the second BiLSTM layer and the jth hidden layer output of the FC layer in the CA module, σ () is a Sigmoid function, n f Is the number of channels output by the first BiLSTM layer; f (f) i The output of the ith channel of the layer; h is a i The output of the ith channel of the layer.
8. The bridge temperature effect prediction method based on the channel attention BiLSTM model according to claim 1 or 7, wherein the Mean Square Error (MSE) of the predicted value and the true value is selected as the loss function, namely, equation (4) is adopted as the optimization target of the network parameters:
where J is a loss function, y i Is the predicted value of the temperature effect of the method provided by the invention,is the actual measurement value of the temperature effect, and B is the batch size input by the network model in the training process.
9. The bridge temperature effect prediction method based on the channel attention BiLSTM model according to claim 1 or 7, wherein an Adam optimization algorithm is selected to train the neural network model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310766544.8A CN116663126A (en) | 2023-06-27 | 2023-06-27 | Bridge temperature effect prediction method based on channel attention BiLSTM model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310766544.8A CN116663126A (en) | 2023-06-27 | 2023-06-27 | Bridge temperature effect prediction method based on channel attention BiLSTM model |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116663126A true CN116663126A (en) | 2023-08-29 |
Family
ID=87726181
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310766544.8A Pending CN116663126A (en) | 2023-06-27 | 2023-06-27 | Bridge temperature effect prediction method based on channel attention BiLSTM model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116663126A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117807854A (en) * | 2024-02-29 | 2024-04-02 | 四川华腾公路试验检测有限责任公司 | Bridge monitoring deflection and temperature separation method based on physical constraint neural network |
-
2023
- 2023-06-27 CN CN202310766544.8A patent/CN116663126A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117807854A (en) * | 2024-02-29 | 2024-04-02 | 四川华腾公路试验检测有限责任公司 | Bridge monitoring deflection and temperature separation method based on physical constraint neural network |
| CN117807854B (en) * | 2024-02-29 | 2024-05-28 | 四川华腾公路试验检测有限责任公司 | Bridge monitoring deflection and temperature separation method based on physical constraint neural network |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112101480B (en) | Multivariate clustering and fused time sequence combined prediction method | |
| CN113723010B (en) | Bridge damage early warning method based on LSTM temperature-displacement correlation model | |
| CN111881495A (en) | Prestress steel structure safety assessment method based on digital twinning | |
| CN114048790B (en) | Road base strain analysis method based on coupling bidirectional LSTM and convolution structure network | |
| CN114548375B (en) | Cable-stayed bridge girder dynamic deflection monitoring method based on two-way long-short-term memory neural network | |
| CN116663126A (en) | Bridge temperature effect prediction method based on channel attention BiLSTM model | |
| CN112507479B (en) | Oil drilling machine health state assessment method based on manifold learning and softmax | |
| CN115654381A (en) | Water supply pipeline leakage detection method based on graph neural network | |
| CN112163669A (en) | Pavement subsidence prediction method based on BP neural network | |
| CN111784065A (en) | Oil well productivity intelligent prediction method based on grey correlation | |
| CN117521512A (en) | Bearing residual service life prediction method based on multi-scale Bayesian convolution transducer model | |
| CN114217025B (en) | Analysis method for evaluating influence of meteorological data on air quality concentration prediction | |
| CN116881819A (en) | Stay cable working state monitoring method based on isolated forest | |
| CN115982141A (en) | Characteristic optimization method for time series data prediction | |
| CN113988210A (en) | Method and device for restoring distorted data of structure monitoring sensor network and storage medium | |
| CN109406076A (en) | A method of beam bridge structure damage reason location is carried out using the mobile principal component of displacement sensor array output | |
| CN110245370A (en) | A kind of high CFRD multiple target mechanics parameter inversion method | |
| CN117454100A (en) | Missing data reconstruction method for engineering structure | |
| CN116842323A (en) | Abnormal detection method for operation data of water supply pipeline | |
| CN111160490A (en) | Deep learning dangerous rock deformation prediction method and device based on multiple time sequences | |
| CN113973403B (en) | Temperature-induced strain field redistribution intelligent sensing method based on structure discrete measurement point topology | |
| CN113570111B (en) | Bridge health state on-chip monitoring method based on lightweight network | |
| CN114004138A (en) | Building monitoring method and system based on big data artificial intelligence and storage medium | |
| CN111291490B (en) | Nonlinear mapping intelligent modeling method for structure multi-scale heterogeneous response | |
| CN111210147A (en) | Sintering process operation performance evaluation method and system based on time sequence feature extraction |
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 |