CN116881819A - Stay cable working state monitoring method based on isolated forest - Google Patents
Stay cable working state monitoring method based on isolated forest Download PDFInfo
- Publication number
- CN116881819A CN116881819A CN202311151305.8A CN202311151305A CN116881819A CN 116881819 A CN116881819 A CN 116881819A CN 202311151305 A CN202311151305 A CN 202311151305A CN 116881819 A CN116881819 A CN 116881819A
- Authority
- CN
- China
- Prior art keywords
- cable
- model
- cable force
- abnormal
- stay
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 70
- 238000012544 monitoring process Methods 0.000 title claims abstract description 37
- 230000002159 abnormal effect Effects 0.000 claims abstract description 93
- 238000012549 training Methods 0.000 claims abstract description 43
- 230000008878 coupling Effects 0.000 claims abstract description 15
- 238000010168 coupling process Methods 0.000 claims abstract description 15
- 238000005859 coupling reaction Methods 0.000 claims abstract description 15
- 230000009467 reduction Effects 0.000 claims abstract description 13
- 230000036541 health Effects 0.000 claims abstract description 12
- 239000000523 sample Substances 0.000 claims description 57
- 238000009826 distribution Methods 0.000 claims description 34
- 239000011159 matrix material Substances 0.000 claims description 30
- 238000013527 convolutional neural network Methods 0.000 claims description 21
- 230000008569 process Effects 0.000 claims description 21
- 238000013256 Gubra-Amylin NASH model Methods 0.000 claims description 20
- 239000013074 reference sample Substances 0.000 claims description 15
- 230000004044 response Effects 0.000 claims description 14
- 239000000463 material Substances 0.000 claims description 11
- 238000004088 simulation Methods 0.000 claims description 11
- 238000013016 damping Methods 0.000 claims description 10
- 230000001133 acceleration Effects 0.000 claims description 9
- 238000006073 displacement reaction Methods 0.000 claims description 9
- 230000007613 environmental effect Effects 0.000 claims description 9
- 230000000694 effects Effects 0.000 claims description 8
- 101100001676 Emericella variicolor andK gene Proteins 0.000 claims description 6
- 238000012614 Monte-Carlo sampling Methods 0.000 claims description 6
- 238000000605 extraction Methods 0.000 claims description 5
- 230000006870 function Effects 0.000 claims description 5
- 238000013508 migration Methods 0.000 claims description 5
- 230000005012 migration Effects 0.000 claims description 5
- 238000010200 validation analysis Methods 0.000 claims description 5
- 238000012795 verification Methods 0.000 claims description 5
- 238000011156 evaluation Methods 0.000 claims description 4
- 230000006399 behavior Effects 0.000 claims description 3
- 238000013145 classification model Methods 0.000 claims description 3
- 230000005284 excitation Effects 0.000 claims description 3
- 230000010354 integration Effects 0.000 claims description 3
- 230000011218 segmentation Effects 0.000 claims description 3
- 239000013598 vector Substances 0.000 claims description 3
- FGUUSXIOTUKUDN-IBGZPJMESA-N C1(=CC=CC=C1)N1C2=C(NC([C@H](C1)NC=1OC(=NN=1)C1=CC=CC=C1)=O)C=CC=C2 Chemical compound C1(=CC=CC=C1)N1C2=C(NC([C@H](C1)NC=1OC(=NN=1)C1=CC=CC=C1)=O)C=CC=C2 FGUUSXIOTUKUDN-IBGZPJMESA-N 0.000 claims description 2
- 238000001514 detection method Methods 0.000 abstract description 10
- 238000013136 deep learning model Methods 0.000 abstract description 3
- 238000010586 diagram Methods 0.000 description 9
- 230000008859 change Effects 0.000 description 6
- 238000005457 optimization Methods 0.000 description 6
- 238000012545 processing Methods 0.000 description 6
- 238000005516 engineering process Methods 0.000 description 5
- 230000008901 benefit Effects 0.000 description 4
- 238000013135 deep learning Methods 0.000 description 4
- 239000000284 extract Substances 0.000 description 4
- 239000004973 liquid crystal related substance Substances 0.000 description 4
- 230000009286 beneficial effect Effects 0.000 description 3
- 238000010276 construction Methods 0.000 description 3
- 238000007689 inspection Methods 0.000 description 3
- 230000005856 abnormality Effects 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 2
- 238000004422 calculation algorithm Methods 0.000 description 2
- 238000012937 correction Methods 0.000 description 2
- 230000007797 corrosion Effects 0.000 description 2
- 238000005260 corrosion Methods 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 238000003745 diagnosis Methods 0.000 description 2
- 230000004927 fusion Effects 0.000 description 2
- 238000012423 maintenance Methods 0.000 description 2
- 239000007787 solid Substances 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- 238000005299 abrasion Methods 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000000295 complement effect Effects 0.000 description 1
- 238000010924 continuous production Methods 0.000 description 1
- 230000006378 damage Effects 0.000 description 1
- 238000013480 data collection Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000012938 design process Methods 0.000 description 1
- 230000005674 electromagnetic induction Effects 0.000 description 1
- 230000003628 erosive effect Effects 0.000 description 1
- 230000007274 generation of a signal involved in cell-cell signaling Effects 0.000 description 1
- 230000003862 health status Effects 0.000 description 1
- 238000009434 installation Methods 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- JEIPFZHSYJVQDO-UHFFFAOYSA-N iron(III) oxide Inorganic materials O=[Fe]O[Fe]=O JEIPFZHSYJVQDO-UHFFFAOYSA-N 0.000 description 1
- 238000012804 iterative process Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000004643 material aging Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000011946 reduction process Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 238000013526 transfer learning Methods 0.000 description 1
- 238000005303 weighing Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/24323—Tree-organised classifiers
-
- 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
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
-
- 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
- G06F30/23—Design optimisation, verification or simulation using finite element methods [FEM] or finite difference methods [FDM]
-
- 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
- G06N3/0455—Auto-encoder networks; Encoder-decoder 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
- G06N3/0464—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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0475—Generative 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
- G06N3/094—Adversarial learning
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Geometry (AREA)
- Computer Hardware Design (AREA)
- Bridges Or Land Bridges (AREA)
Abstract
The invention relates to the technical field of axle coupling systems, and provides a stay cable working state monitoring method based on an isolated forest, which comprises the steps of firstly, establishing a finite element model of a cable-stayed bridge to simulate axle coupling vibration when a vehicle passes through the bridge; then arranging a cable force sensor on an actual stay cable to acquire cable force signals, and correcting cable force samples in normal and abnormal states generated by a finite element model according to the cable force signals; then adopting a convolution variation self-coding deep learning model to perform dimension reduction treatment on input data, and extracting representative features from the input data; the dimension-reduced data are used for training an isolated forest model to output the health state or abnormal state of the stay cable; after training, the isolated forest model is used for monitoring the working state of the stay cable of the cable-stayed bridge in real time. The invention not only improves the accuracy of stay cable state monitoring, but also realizes early detection and early warning of the abnormal state of the stay cable, and has important application value.
Description
Technical Field
The invention relates to the technical field of axle coupling systems, in particular to a stay cable working state monitoring method based on an isolated forest.
Background
The cable-stayed bridge is a special bridge structure type, and is widely applied to bridge engineering due to the unique structural form and high-efficiency economical performance. However, with the widespread use of cable-stayed bridges and the gradual expansion of the cable-stayed bridge scale, how to effectively maintain and manage the operation state of the cable-stayed bridge, especially the operation state of the stay cable, has become a problem to be solved.
The stay cable is used as a main bearing component of the cable-stayed bridge, and the state of the stay cable directly influences the operation safety of the whole bridge. However, the stay cable is affected by many factors including wind load, rain erosion, material aging, and overweight vehicle load during operation, and these factors may cause the mechanical properties of the stay cable to change, thereby affecting the overall performance of the bridge. Therefore, the state of the stay cable is monitored in real time so as to discover and process possible problems in time, and the stay cable is important to ensuring the safe operation of the bridge.
However, the conventional stay cable state monitoring methods mainly rely on manual inspection and visual observation, and the methods can discover surface defects of the stay cable, such as rust, abrasion and the like to a certain extent, but cannot discover deep problems which are not easy to discover, such as damage, microcracks and the like in the cable. In addition, the manual inspection method also has the problems of low efficiency, long time consumption, incapability of realizing continuous monitoring and the like.
In order to solve these problems, in recent years, some advanced technologies are gradually introduced into the state monitoring of the stay cable, including acoustic emission technology, vibration monitoring technology, electromagnetic induction technology, and the like. However, these methods, while improving the effectiveness of stay cable condition monitoring to some extent, still present some problems. For example, most of these techniques require the installation of a large number of sensors and equipment, which is costly to install and maintain; moreover, these techniques often require specialized operators and complex data processing procedures, which are difficult to use and operate.
Based on the reasons, the stay cable state monitoring method is developed, can realize real-time monitoring of the stay cable state, can early warn the change of the stay cable state, is simple and convenient to operate and low in cost, and has important practical application value.
Therefore, the invention provides a stay cable working state monitoring method based on an isolated forest. The method can realize real-time monitoring and early warning of the state of the stay cable by using deep learning and machine learning technologies so as to solve the problems and provide powerful technical support for the operation safety of the cable-stayed bridge.
Disclosure of Invention
Aiming at the abnormal diagnosis of the operation state of the stay cable, the invention provides the stay cable working state monitoring method based on the isolated forest, which not only improves the accuracy of the stay cable state monitoring, but also realizes the early detection and early warning of the abnormal state of the stay cable and has important application value. The technical proposal is as follows:
a stay cable working state monitoring method based on an isolated forest comprises the following steps:
step 1: establishing a finite element model of a cable-stayed bridge structure under a normal stay cable condition, wherein the finite element model comprises a bridge model and a stay cable model, and describing the mechanical characteristics of the cable-stayed bridge;
step 2: selecting a specific vehicle model as a simulation object, constructing a vehicle model, and importing the vehicle model and a finite element model of a cable-stayed bridge structure into finite element analysis software; setting vehicle running parameters, influence of bridge floor unevenness and environmental factors, and finally performing axle coupling vibration simulation generated when the vehicle passes through the cable-stayed bridge, observing the power effect of the cable-stayed cable and influence of vehicle load on the cable-stayed cable, obtaining solution of an axle coupling dynamic model, and generating a simulated cable force signal;
step 3: a cable sensor is arranged on a stay cable of an actual cable-stayed bridge to collect cable force signals in an actual running state;
step 4: utilizing cable force signal data in an actual running state to establish a pre-trained CNN model so as to capture the inherent characteristics and dynamic behaviors of cable force of the cable-stayed bridge;
step 5: on the basis of the pre-trained CNN model, correcting and optimizing a simulated cable force signal generated by a finite element model by using a migration learning strategy to obtain a normal cable force sample;
step 6: constructing a finite element model of the cable-stayed bridge under the abnormal condition of the stay cable, generating a preliminary abnormal cable force reference sample, inputting the preliminary abnormal cable force reference sample and the normal cable force sample into a GAN model for training, and then generating a corresponding abnormal cable force sample by introducing a preset abnormal condition into the input of the GAN model;
step 7: performing characteristic dimension reduction on a normal cable force sample and an abnormal cable force sample by adopting a convolution variation self-coding model: the convolution variation self-coding model comprises an encoder and a decoder, wherein the encoder is used for partially learning potential distribution of cable force response data, and new data is generated from the potential distribution through the decoder, so that distribution characteristic information of normal cable force samples and abnormal cable force samples is obtained;
step 8: and taking the distribution characteristic information of the normal cable force sample and the abnormal cable force sample as the input of the isolated forest model, and taking the health state value or the abnormal state value of the stay cable as the output to complete the training of the model.
The selection of feature data of the convolutional variational self-coding model after dimension reduction as input to the isolated forest model is an innovative attempt, because the convolutional variational self-coder can effectively extract key and representative features from high-dimensional data and perform dimension reduction, thereby enabling the isolated forest model to perform anomaly detection more accurately and efficiently. The combination mode enables the advantages of the two models to be fully exerted, and deep fusion of feature extraction and anomaly detection is achieved.
Step 9: real-time monitoring is carried out by applying a trained isolated forest model: and (3) inputting cable force signals of the stay cables in actual operation into a trained isolated forest model, and judging whether the working state of the stay cables is abnormal or not according to the health state value or the abnormal state value output by the isolated forest model.
Further, the bridge model is:
（1）
in the formula (i),M b 、C b 、K b respectively representing a bridge mass matrix, a bridge damping matrix and a bridge stiffness matrix;respectively representing acceleration, speed and displacement response of the bridge;F b representing the external force applied to the bridge;
the vehicle model is as follows:
（2）
in the formula (i),M v 、C v andK v a mass matrix, a damping matrix and a stiffness matrix of the vehicle respectively,acceleration, speed and displacement vectors respectively representing the movement of the vehicle;f v for external excitation forces or loads acting on the vehicle;
the stay cable model is as follows:
（3）
wherein, the liquid crystal display device comprises a liquid crystal display device,M s 、C s andK s respectively representA mass matrix, a damping matrix and a rigidity matrix of the stay cable,andU s respectively representing the acceleration, the speed and the displacement of the stay cable,F s is the force acting on the stay cable.
Further, the step 4 specifically includes: pre-training based on cable force signals of an actual cable-stayed bridge, and constructing a five-layer convolutional neural network model which comprises three convolutional layers and two full-connection layers, wherein the aim is to grasp key characteristics of the cable force signals; firstly, collecting cable force signal data as input data of a convolutional neural network model, performing feature extraction on the input data through three convolutional layers, performing feature integration through two fully connected layers, and finally outputting a pre-trained convolutional neural network model.
Further, the step 6 specifically includes:
step 6.1: constructing a finite element model of the cable-stayed bridge under the abnormal condition of the stay cable based on the physical model and the material attribute, performing grid division and boundary condition setting, and then simulating the abnormal state of the stay cable by changing the material attribute or the boundary condition to generate a preliminary abnormal cable force reference sample;
step 6.2: taking the normal cable force sample generated in the step 5 and the preliminary abnormal cable force reference sample generated in the step 6.1 as inputs of a GAN model, wherein the GAN model adopts deep convolution to generate an countermeasure network, and a generator part consists of a convolution deconvolution network, so that a new cable force signal which accords with data distribution is generated through learning the data distribution; the discriminator part is composed of a convolution network and is used for judging whether an input cable force signal sample is real or generated by a generator;
step 6.3: in the model training process, firstly, a normal cable force sample and a preliminary abnormal cable force reference sample are used for pre-training, and then a preset abnormal condition is introduced into a cable force signal by controlling noise input; training the generator model for a set number of rounds, each round of training comprising two phases: firstly, fixing parameters of a discriminator, optimizing parameters of a generator, and enabling a generated false sample to pass through judgment of the discriminator as far as possible; then fixing the generator parameters, optimizing the parameters of the discriminator, and enabling the discriminator to accurately distinguish true samples from false samples as far as possible; the two phases are alternately carried out until the model converges; after training is finished, the generator generates an abnormal cable force sample under the preset abnormal condition.
Further, the step 7 specifically includes:
step 7.1: in the encoding stage, the CVAE model learns to map the original data of the cable force response into a low-dimensional potential space; in the process, the CVAE model captures key information of original data by optimizing Kullback-Leibler divergence between posterior distribution and prior distribution of hidden variables;
step 7.2: in the decoding stage, the CVAE model generates new data from the potential space so that the generated data is as close as possible to the original data; in the process, the CVAE model optimizes the reconstruction loss between the generated data and the original data, so that the model reconstructs data close to the original data from the low-dimensional potential space, and the distribution characteristic information of the data is captured.
Further, before the step 8 and the step 9, the method further includes: the model is evaluated by using a separate validation dataset, in particular:
step a: collecting actual cable force signals of N cable-stayed bridges in various operation states;
step b: generating a cable force signal sample under a normal working condition and a preset abnormal working condition according to the finite element model; the Monte Carlo sampling method is applied, and 2N normal working condition samples and 2N abnormal working condition samples are extracted from the samples;
step c: generating additional preset abnormal cable force signal samples by adopting a trained generated GAN model, and extracting 2N samples by adopting a Monte Carlo sampling method;
step d: integrating the four types of data to form an independent verification data set containing 7N samples; in the data set, the cable force signal is taken as an input, and the corresponding health state value or abnormal state value is taken as an output;
step e: selecting four classification model evaluation indexes of precision, recall ratio and F1 score to evaluate the prediction performance and generalization capability of the GAN model in the step 6 or the isolated forest model in the step 8;
step f: if the model prediction effect is not ideal, tuning the generated countermeasure network model or the isolated forest model; tuning the GAN model, including adjusting the architecture of the GAN model, the setting of an optimizer, the learning rate, and simultaneously trying different loss functions or introducing regularization techniques to improve the performance of the model; tuning the isolated forest model includes adjusting the sample segmentation mode and the number of trees.
Compared with the prior art, the invention has the beneficial effects that:
1) The invention provides an effective real-time monitoring and managing tool for the cable stayed bridge, which realizes real-time and accurate judgment of the working state of the cable stayed based on the self-coding method for the abnormal operation state of the cable stayed based on the isolated forest and the convolution variation, and provides powerful support for the operator to discover and process potential problems in time.
2) The method applies the deep learning method to data dimension reduction and feature extraction, so that the model training efficiency and the real-time monitoring speed are greatly improved; a large number of simulated and corrected stay cable response result samples comprise normal states and various abnormal states, so that training accuracy and generalization capability of the model are enhanced.
3) The independent verification data set is used, so that the accuracy and generalization capability of the model are further verified, and the method has a remarkable effect on ensuring the safe operation of the cable-stayed bridge.
4) The method is beneficial to the early detection and early warning of the abnormal state of the stay cable, and the isolated forest model can immediately send out an early warning signal when detecting the possible abnormal working state of the stay cable, so that enough time is provided for operation management personnel to check and process, and the method is very important for preventing accidents of the cable-stayed bridge.
5) The stay cable working state monitoring method based on the isolated forest not only improves the accuracy of stay cable state monitoring, realizes early detection and early warning of the abnormal state of the stay cable, but also shows practical value. The method provides powerful technical support for the safe operation of the cable-stayed bridge, and has remarkable social and economic benefits.
Drawings
Fig. 1 is a diagram of a cable-stayed bridge.
Fig. 2 is a schematic diagram of a normal cable force sample.
FIG. 3 is a schematic diagram of an abnormal cable force sample.
FIG. 4 is a flow chart of a stay cable anomaly simulation and training based on a deep convolution generated countermeasure network (DCGAN).
Fig. 5 is a characteristic diagram of a convolved self-encoded reduced-dimension cable force signal.
Fig. 6 is a diagram of a convolutional self-encoding structure.
FIG. 7 is a convolution variance self-coding-isolated forest combination early warning model.
Fig. 8 is an isolated forest recognition result.
Fig. 9 is a flowchart of a stay cable isolated forest model verification.
Fig. 10 is a flow chart of real-time stay cable operating condition monitoring.
Detailed Description
The invention will now be described in further detail with reference to the drawings and to specific examples.
The invention provides a stay cable working state monitoring method based on an isolated forest. Firstly, by establishing a finite element model of a cable-stayed bridge, axle coupling vibration during vehicle passing is simulated. And then, acquiring a cable force signal in an actual running state through a cable force sensor arranged on an actual stay cable, and correcting cable force samples in a normal state and an abnormal state generated by the finite element model so as to be closer to a real sample. And the corrected cable force samples in the normal state and the abnormal state are used for training a subsequent isolated forest model.
Then, a deep learning model called convolutional variational self-coding (Convolutional Variational Autoencoder, CVAE) is used to perform a dimension reduction process on the input data and extract representative features therefrom. The dimension-reduced data is used for training an isolated forest model, and the model takes the cable force response of the stay cable as input and the health state or abnormal state of the stay cable as output.
After training, the isolated forest model is used for monitoring the working state of the stay cable of the cable-stayed bridge in real time. According to the output of the model, whether the working state of the stay cable is abnormal or not and the degree of the abnormality can be judged, and powerful decision support is provided for maintenance and management of the cable-stayed bridge. The accuracy and generalization ability of the model was further confirmed by using a separate validation dataset. The specific process is as follows:
step 1: and establishing a finite element model of the cable-stayed bridge structure under the condition of a normal stay cable, and describing the mechanical characteristics of the cable-stayed bridge.
The implementation of the method starts with the creation of a finite element model of the cable-stayed bridge structure. Firstly, according to the design and construction drawing of the cable-stayed bridge, the geometric information of main structural parts, such as a main girder, a stay cable, a tower and the like, is extracted. Second, based on the actual performance parameters of the materials (e.g., concrete, rebar, etc.) used in the cable-stayed bridge, the material characteristics of the various portions, including modulus of elasticity, poisson's ratio, density, etc., are determined. Next, a finite element model of the cable-stayed bridge is created using existing commercial finite element analysis software (e.g., ANSYS). In the process of model construction, a three-dimensional solid model of the bridge is firstly established according to the extracted geometric information by utilizing the CAD function of software, then grid division is carried out, and common division types comprise solid division and shell division. The size of the grid should be balanced between ensuring the computational accuracy and the computational efficiency. The previously determined material properties are then assigned to the corresponding grid elements.
In the setting of the boundary conditions of the model, the constraint conditions of the bridge-stayed bridge, such as bridge piers and foundations, are usually fixed supports. In the finite element model design process, complex environmental factors such as wind load, temperature change, earthquake and the like possibly existing in the cable-stayed bridge under the actual working condition need to be considered. These environmental factors can be modeled by setting corresponding loads and boundary conditions in the finite element model. For example, wind loading may be simulated by setting fluid-structure interaction conditions, temperature changes may be simulated by setting thermal loading, and earthquakes may be simulated by setting seismic dynamic loading.
Finally, a solver of finite element analysis software is used for solving the model, if the calculation result has larger deviation from the design or actual measurement result, the model parameters need to be adjusted, such as grid size modification, grid quality optimization, material performance parameter adjustment and the like, which is an iterative process, and the model parameters need to be adjusted for multiple times to meet engineering requirements. Fig. 1 is a diagram of a cable-stayed bridge.
Step 2: selecting a specific vehicle model as a simulation object, constructing a vehicle model, and importing the vehicle model and a finite element model of a cable-stayed bridge structure into finite element analysis software; and setting vehicle running parameters, influence of bridge floor unevenness and environmental factors, and finally performing axle coupling vibration simulation generated when the vehicle passes through the cable-stayed bridge, observing the power effect of the cable-stayed cable and influence of vehicle load on the cable-stayed cable, and obtaining a solution of an axle coupling dynamic model to obtain a simulated cable force signal.
And (3) applying the finite element model of the cable-stayed bridge in the normal stay cable condition established in the step (1) to accurately simulate the axle coupling vibration generated in the process of crossing the cable-stayed bridge by the vehicle. The process mainly focuses on the power effect of the stay cable, and particularly considers the influence of the vehicle load on the stay cable.
First, simulating this process requires selecting a vehicle model that matches the actual vehicle model. The selected simulation object is a common commercial truck weighing 10 tons, 6 meters long and 2.5 meters wide, and a vehicle model of the truck is built. In this vehicle model, the four tires of the vehicle are considered as four separate mass points, while the vehicle body is considered as a rigid body. Thus, a quality matrix of the vehicle model can be obtainedM v Damping matrixC v And stiffness matrixK v The parameters may be determined based on real vehicle testing or related literature.
The vehicle model described above is then imported into ANSYS software along with the normal stay cable condition cable-stayed bridge finite element model created in step 1, including the bridge model and the stay cable model. In a software environment, a comprehensive coupling vibration model covering vehicles, bridges and stay cables is constructed based on an axle coupling dynamics theory.
Secondly, setting vehicle driving parameters, and influence of bridge floor unevenness and environmental factors. The vehicle is set to traverse the deck at a certain speed (e.g., 60 km/h) and the travel trajectory of the vehicle is set from one end of the bridge to the other. In addition, the influence of the unevenness of the bridge deck and the environmental factors such as wind power, temperature and the like are considered, and the factors can be set in corresponding parameters in software.
Then, running the simulation function of ANSYS software, simulating axle coupling vibration generated when the vehicle passes through the cable-stayed bridge, and obtaining solutions of an axle coupling dynamics model to obtain cable force response data of the vehicle, the bridge and the stay cable.
The bridge model is as follows:
（1）
in the formula (i),M b 、C b 、K b respectively representing a bridge mass matrix, a bridge damping matrix and a bridge rigidity matrix,respectively representing acceleration, speed and displacement response of the bridge,F b indicating the external force applied to the bridge.
The vehicle model is as follows:
（2）
in the formula (i),M v 、C v andK v a mass matrix, a damping matrix and a stiffness matrix of the vehicle respectively,acceleration, speed and displacement vectors respectively representing the movement of the vehicle;f v for external excitation forces or loads acting on the vehicle.
The stay cable model is as follows:
（3）
wherein, the liquid crystal display device comprises a liquid crystal display device,M s 、C s andK s respectively representing a mass matrix, a damping matrix and a rigidity matrix of the stay cable,andU s respectively representing the acceleration, the speed and the displacement of the stay cable,F s is the force acting on the stay cable.
Step 3: in an actual cable-stayed bridge structure, cable force sensors are arranged to collect cable force signals.
Based on the operation state of the actual cable-stayed bridge, cable force signals in a normal working state are obtained by arranging cable force sensors on the stay cables. The cable force sensor is arranged in consideration of the structural characteristics and cable force distribution condition of the cable-stayed bridge, and is usually arranged at important parts of the stayed cable, such as two ends or the middle part. The collection of cable force signals is a continuous process, and cable force change conditions of the cable-stayed bridge under different operation states (such as different vehicle flow rates and environmental conditions) need to be recorded. Specific data collection content comprises real-time signals of each cable sensor, time stamps, environmental condition data (such as temperature and wind power), traffic flow and the like. These data will be used for subsequent analysis and optimization.
Step 4: and establishing a pre-trained CNN model by utilizing actual cable-stayed bridge cable force signal data so as to capture the inherent characteristics and dynamic behaviors of the cable-stayed bridge cable force.
The cable force signal sample data of the actual cable-stayed bridge collected in the step 3 is used for training a CNN model, and the purpose of the cable force signal sample data is to correct and optimize the cable force signal generated by the finite element model by learning complex modes and features in the real data. The cable force signal of an actual cable-stayed bridge reflects a more realistic situation, and contains key information and characteristics which can be ignored by a plurality of analog signals. By learning these characteristics, the CNN model can more accurately model the cable force signal, thereby improving the accuracy of the finite element model. The specific technical means are as follows:
and (3) pre-training based on a cable force signal (source task) of the actual cable-stayed bridge, and constructing a five-layer Convolutional Neural Network (CNN) model. The CNN model, as an application of the prior art, comprises three convolution layers and two full connection layers, with the goal of grabbing the key characteristics of the stay cable force signal. The process of model construction is as follows: firstly, taking cable force data (comprising cable force change values, change rates, time points and the like) collected in the step 3 as input data of a CNN model, extracting characteristics of the input data through a three-layer convolutional neural network, then carrying out characteristic integration through two full-connection layers, and finally outputting the pre-trained CNN model.
Step 5: on the basis of the pre-trained CNN model, a migration learning strategy is used for correcting and optimizing a simulated cable force signal generated by a finite element model to obtain a normal cable force sample, and the normal cable force sample is used as a training sample set of an isolated forest model.
On the basis of a five-layer convolutional neural network model which is pre-trained through actual cable-stayed bridge cable force signal data, a migration learning strategy is implemented to process cable force signals (target tasks) generated by a finite element model. The adopted migration learning strategy is a Fine-Tuning (Fine-Tuning) method commonly used in the field of deep learning, which is a prior art method, and particularly can effectively improve the performance of the model on a target task under the condition that source task data (actual cable-stayed bridge cable force signal data) are relatively less and target task data (cable force signals generated by a finite element model) are more.
In the fine tuning process, 10 rounds of iterative optimization are performed, and a mode of calculating a loss function (such as cross entropy loss) and updating parameters by using a back propagation algorithm is adopted to improve the prediction accuracy of the simulated cable force signal. By means of learning knowledge of source task data, a cable force signal generated by the finite element model is closer to an actual stay cable force signal, and therefore model prediction accuracy is improved.
The simulated cable force signal after the optimization of the pre-training model is regarded as a normal cable force sample. These normal cable force samples will be used as a training sample set for the isolated forest model. Fig. 2 is a schematic diagram of a normal cable force sample.
Step 6: constructing a finite element model of the cable-stayed bridge under the abnormal condition of the stay cable, generating a preliminary abnormal cable force reference sample, inputting the preliminary abnormal cable force reference sample and the normal cable force sample into a GAN model for training, and then generating a corresponding abnormal cable force sample by introducing a preset abnormal condition into the input of the GAN model.
First, a finite element model of the cable-stayed bridge is constructed under the abnormal conditions of the stay cable (such as abnormal cable force, local corrosion, anchor failure, breakage and the like). Specifically, the finite element model is based on a physical model and material properties, performs grid division and boundary condition setting, and then simulates abnormal states of the stay cable, such as anchor failure, local corrosion and the like, by changing the material properties (such as material elastic modulus, yield strength and the like) or the boundary conditions (such as fixation, freedom and the like). Through such simulation, a preliminary abnormal cable force reference sample is generated that will be used to generate a preliminary learning reference against a network of challenge (GAN) model. FIG. 3 is a schematic diagram of an abnormal cable force sample.
And then, taking the normal cable force sample subjected to transfer learning correction and optimization in the step 5 and the preliminary abnormal cable force reference sample as the input of the GAN model. In the selection of the GAN model, a deep convolutional generation countermeasure network (DCGAN) was employed, which has proven to perform well in many generation tasks. Specifically, in the DCGAN model, the generator part is composed of a convolution deconvolution network, and the aim is to generate a new cable force signal conforming to the data distribution by learning the data distribution; the discriminator section is formed by a convolutional network and the task is to determine whether the input cable force signal sample is authentic or generated by a generator.
In the model training process, the normal cable force sample after correction and optimization in the step 5 and the preliminary abnormal cable force reference sample are used for pre-training, and then a preset abnormal condition is introduced into the cable force signal through controlling noise input (for example, noise distribution can be assumed to be Gaussian distribution with a mean value of 0 and a variance of 1). The generator model was trained 500 times, using 10000 cable force signal samples per round. Each wheel training comprises two phases: firstly, fixing parameters of a discriminator, optimizing parameters of a generator, and enabling a generated false sample to "cheat" the discriminator as far as possible; and then fixing the generator parameters, optimizing the parameters of the discriminator, and enabling the discriminator to correctly distinguish true samples from false samples as far as possible. These two phases alternate until the model converges. After training is finished, the generator can generate an abnormal cable force sample under the preset abnormal condition. FIG. 3 is a schematic diagram of an abnormal cable force sample. A stay cable anomaly simulation and training flow chart based on deep convolution to generate a countermeasure network (DCGAN) is shown in fig. 4.
And finally, taking the generated abnormal samples and the normal samples which are corrected and optimized in the step 5 as a training sample set of the isolated forest model. Through the step, the cable force signal generation based on deep learning is realized, and a richer and more representative sample is provided for training of an isolated forest model.
Step 7: performing characteristic dimension reduction on a normal cable force sample and an abnormal cable force sample by adopting a convolution variation self-coding model: the convolution variation self-coding model comprises an encoder and a decoder, wherein the encoder is used for partially learning potential distribution of the cable force response data, and new data is generated from the potential distribution through the decoder, so that distribution characteristic information of normal cable force samples and abnormal cable force samples is obtained.
The innovative practice in step 7 is mainly focused on applying the convolution variation self-coding model to perform feature dimension reduction on the samples generated in steps 5 and 6. The convolution variation self-coding model plays a key role in effectively capturing complex characteristics of the cable force signals and achieving dimension reduction. As shown in fig. 5, the distribution information of the cable force signal characteristics after the convolution variation self-coding model is subjected to dimension reduction can provide more concise and visual data input for the subsequent model.
The convolutional variational self-coding model is a deep learning model that has proven to be capable of extracting key features of data in many tasks. The application of the model is a major innovation of the procedure. Because the distribution and the form of the cable force signals are complex, direct analysis and processing often cause difficulties, and the dimension reduction processing of the convolution variation self-coding model reduces the complexity of data, and simultaneously extracts the most representative characteristics, so that the subsequent model can learn and predict more accurately. As shown in fig. 6, the structure of the convolutional variational self-coding model includes two main phases: an encoding stage and a decoding stage.
This model is based on the principle of variance inference, with the encoder part learning the potential distribution of data, and generating new data from this distribution by the decoder part. In the process, the model can effectively acquire the distribution characteristic information of the data, and the dimension-reduced data can be used for training a more efficient isolated forest model.
In the encoding phase, model learning maps raw data (cable force response) into a low-dimensional potential space. In this process, the model captures key information of the raw data mainly by optimizing the Kullback-Leibler divergence between the posterior and prior distributions of the hidden variables. In the decoding stage, the model generates new data from the potential space such that the generated data is as close as possible to the original data. In this process, the model optimizes the reconstruction loss between the generated data and the original data so that the model can reconstruct data similar to the original data from the low-dimensional potential space. In this way, the convolution variation self-coding model can effectively capture the distribution characteristic information of the data, and further provides powerful support for subsequent model training and prediction.
Step 8: and taking the distribution characteristic information of the normal cable force sample and the abnormal cable force sample as the input of the isolated forest model, and taking the health state value or the abnormal state value of the stay cable as the output to complete the training of the model.
The feature data after the convolution variation is reduced from the coding model is input into the isolated forest model, which is a prospective attempt with innovation. The cable force signal data of the stay cable is generally high-dimensional, and the traditional isolated forest model can be low in efficiency and limited in accuracy when directly processing the high-dimensional data. Convolution variants have the ability to accurately extract key, representative features from these high-dimensional data and perform dimension reduction to obtain a more compact data representation. Such low-dimensional features are more focused, which facilitates more efficient and accurate anomaly detection of the isolated forest model. In addition, the combination form of the method also provides a platform for the complementary advantages of the two models, and the deep fusion of the feature extraction and the anomaly detection is realized.
The isolated forest model is an unsupervised anomaly detection algorithm, and the model takes the cable force signal characteristics after the convolutional variation self-coding model dimension reduction treatment as input, and the output is the health state (set as 1) or the anomaly state (set as 0) of the stay cable. The application mode not only fully utilizes the advantages of the isolated forest model in the field of anomaly detection, but also greatly improves the capability of the model for processing complex cable force signals. The object of the model is to minimize the difference between the predicted and the actual values, so that a high prediction accuracy can be maintained even when complex, multiple cable force signals are processed. FIG. 7 is a convolution variance self-coding-isolated forest combination early warning model. Fig. 8 is a recognition result of an isolated forest model.
In addition, the relationship between the training set size and the identification precision of the isolated forest shown in table 1 also reveals that the isolated forest model has better stability and scalability when processing a large-scale data set. Along with the increase of the capacity of the training sample, the recognition precision of the isolated forest model shows an improved trend under various abnormal rates, which indicates that the model can adapt to large-scale and complex cable force signal data, thereby playing an important role in health monitoring and early warning of the stay cable.
TABLE 1 isolated forest identification accuracy for different training set sizes
。
Step 9: the model is evaluated by using a separate validation dataset to ensure accuracy and generalization ability of the model.
First, actual cable force signals of 2000 cable-stayed bridges in various operation states are collected. And then, generating a large number of cable force signal samples under normal working conditions and preset abnormal working conditions according to the finite element model. The Monte Carlo sampling method is applied, and 4000 normal working condition samples and 4000 abnormal working condition samples are extracted from the samples. Then, a trained generation countermeasure network model (GAN) is adopted to generate additional preset abnormal cable force signal samples, and 4000 samples are extracted by using a Monte Carlo sampling method. Finally, the four types of data are integrated to form an independent validation data set containing 14000 samples. In this dataset, the cable force signal is taken as input, while the corresponding health state (set to 1) or abnormal state (set to 0) is taken as output.
For model evaluation, four commonly used classification model evaluation indexes of Precision (Accuracy), precision (Precision), recall (Recall) and F1 score are selected. When the model prediction precision exceeds 90%, and the precision, recall ratio and F1 fraction are all higher than 0.9, the model can be considered to have good prediction performance and generalization capability.
If the model prediction effect is not ideal, the generation of the countermeasure network model (GAN) in step 6 or the isolated forest model in step 8 needs to be optimized. For generating the tuning of the countermeasure network model (GAN), mainly comprising the framework of the tuning model, the setting of an optimizer, the learning rate and the like; at the same time, different loss functions may be tried or regularization techniques introduced to improve the performance of the model. Tuning of the isolated forest model mainly comprises adjusting a sample segmentation mode, the number of trees and the like. Combining these measures will help to improve the prediction accuracy of the model. The main goal of the model in practical application is to accurately predict the working state of the stay cable, which includes both normal states and various possible abnormal states. Fig. 9 is a flow chart for verification of a stay cable isolated forest model.
Step 10: the trained and optimized isolated forest-convolution variation self-coding model is deployed to monitor the working state of the stay cable in real time. In actual operation, the sensors collect the cable force responses of the stay cables and then input these cable force response values into the isolated forest-convolution variance self-encoding model. The model will directly output the health status (set to 1) or abnormal status (set to 0) of the stay cable. If the output result shows an abnormal state, the model generates corresponding alarm information so as to remind relevant staff to carry out timely inspection and maintenance. The method can realize real-time and accurate monitoring of the working state of the stayed cable, thereby greatly improving the operation safety and efficiency of the cable-stayed bridge. Fig. 10 shows a real-time stay cable operating condition monitoring flow.
In conclusion, the method is beneficial to the early discovery and early warning of the abnormal state of the stay cable, and the isolated forest model can immediately send out an early warning signal when detecting the possible abnormal working state of the stay cable, so that enough time is provided for operation management staff to check and process, which is important for preventing accidents of the cable-stayed bridge. The stay cable operation state abnormality diagnosis method based on the isolated forest and the convolution variation self-coding not only improves the accuracy of stay cable state monitoring, realizes early discovery and early warning of the stay cable abnormal state, but also shows practical value.
Claims (6)
1. The stay cable working state monitoring method based on the isolated forest is characterized by comprising the following steps of:
step 1: establishing a finite element model of a cable-stayed bridge structure under a normal stay cable condition, wherein the finite element model comprises a bridge model and a stay cable model, and describing the mechanical characteristics of the cable-stayed bridge;
step 2: selecting a vehicle model as a simulation object, constructing a vehicle model, and importing the vehicle model and a finite element model of a cable-stayed bridge structure into finite element analysis software; setting vehicle running parameters, influence of bridge floor unevenness and environmental factors, and finally performing axle coupling vibration simulation generated when the vehicle passes through the cable-stayed bridge, observing the power effect of the cable-stayed cable and influence of vehicle load on the cable-stayed cable, obtaining solution of an axle coupling dynamic model, and generating a simulated cable force signal;
step 3: a cable sensor is arranged on a stay cable of an actual cable-stayed bridge to collect cable force signals in an actual running state;
step 4: utilizing cable force signal data in an actual running state to establish a pre-trained CNN model so as to capture the inherent characteristics and dynamic behaviors of cable force of the cable-stayed bridge;
step 5: on the basis of the pre-trained CNN model, correcting and optimizing a simulated cable force signal generated by a finite element model by using a migration learning strategy to obtain a normal cable force sample;
step 6: constructing a finite element model of the cable-stayed bridge under the abnormal condition of the stay cable, generating a preliminary abnormal cable force reference sample, inputting the preliminary abnormal cable force reference sample and the normal cable force sample into a GAN model for training, and then generating a corresponding abnormal cable force sample by introducing a preset abnormal condition into the input of the GAN model;
step 7: performing characteristic dimension reduction on a normal cable force sample and an abnormal cable force sample by adopting a convolution variation self-coding model: the convolution variation self-coding model comprises an encoder and a decoder, wherein the encoder is used for partially learning potential distribution of cable force response data, and new data is generated from the potential distribution through the decoder, so that distribution characteristic information of normal cable force samples and abnormal cable force samples is obtained;
step 8: taking the distribution characteristic information of the normal cable force sample and the abnormal cable force sample as the input of the isolated forest model, and taking the health state value or the abnormal state value of the stay cable as the output to complete the training of the model;
step 9: real-time monitoring is carried out by applying a trained isolated forest model: and (3) inputting cable force signals of the stay cables in actual operation into a trained isolated forest model, and judging whether the working state of the stay cables is abnormal or not according to the health state value or the abnormal state value output by the isolated forest model.
2. The stay cable working state monitoring method based on the isolated forest according to claim 1, wherein the bridge model is as follows:
（1）
in the formula (i),M b 、C b 、K b the bridge mass matrix, the bridge damping matrix and the bridge rigidity matrix are respectively adopted;bridge acceleration, speed and displacement response respectively;F b the bridge is subjected to external force;
the vehicle model is as follows:
（2）
in the formula (i),M v 、C v andK v a mass matrix, a damping matrix and a stiffness matrix of the vehicle respectively,acceleration, speed and displacement vectors of vehicle movement respectively;f v for external excitation forces or loads acting on the vehicle;
the stay cable model is as follows:
（3）
in the formula (i),M s 、C s andK s respectively a mass matrix, a damping matrix and a rigidity matrix of the stay cable,andU s acceleration, speed and displacement of the stay cable are respectively,F s is the force acting on the stay cable.
3. The stay cable working state monitoring method based on the isolated forest according to claim 1, wherein the step 4 is specifically: pre-training based on cable force signals of an actual cable-stayed bridge, and constructing a five-layer convolutional neural network model which comprises three convolutional layers and two full-connection layers, wherein the aim is to grasp key characteristics of the cable force signals; firstly, collecting cable force signal data as input data of a convolutional neural network model, performing feature extraction on the input data through three convolutional layers, performing feature integration through two fully connected layers, and finally outputting a pre-trained convolutional neural network model.
4. The method for monitoring the working state of the stay cable based on the isolated forest according to claim 1, wherein the step 6 specifically comprises:
step 6.1: constructing a finite element model of the cable-stayed bridge under the abnormal condition of the stay cable based on the physical model and the material attribute, performing grid division and boundary condition setting, and then simulating the abnormal state of the stay cable by changing the material attribute or the boundary condition to generate a preliminary abnormal cable force reference sample;
step 6.2: taking the normal cable force sample generated in the step 5 and the preliminary abnormal cable force reference sample generated in the step 6.1 as inputs of a GAN model, wherein the GAN model adopts deep convolution to generate an countermeasure network, and a generator part consists of a convolution deconvolution network, so that a new cable force signal which accords with data distribution is generated through learning the data distribution; the discriminator part is composed of a convolution network and is used for judging whether an input cable force signal sample is real or generated by a generator;
step 6.3: in the model training process, firstly, a normal cable force sample and a preliminary abnormal cable force reference sample are used for pre-training, and then a preset abnormal condition is introduced into a cable force signal by controlling noise input; training the generator model for a set number of rounds, each round of training comprising two phases: firstly, fixing parameters of a discriminator, optimizing parameters of a generator, and enabling a generated false sample to pass through judgment of the discriminator; then fixing the generator parameters, optimizing the parameters of the discriminator, and enabling the discriminator to correctly distinguish true samples from false samples; the two phases are alternately carried out until the model converges; after training is finished, the generator generates an abnormal cable force sample under the preset abnormal condition.
5. The method for monitoring the working state of the stay cable based on the isolated forest according to claim 1, wherein the step 7 specifically comprises:
step 7.1: in the encoding stage, the CVAE model learns to map the original data of the cable force response into a low-dimensional potential space; in the process, the CVAE model captures key information of original data by optimizing Kullback-Leibler divergence between posterior distribution and prior distribution of hidden variables;
step 7.2: in the decoding stage, the CVAE model optimizes the reconstruction loss between the generated data and the original data, and reconstructs new data from the low-dimensional potential space, thereby capturing the distribution characteristic information of the data.
6. The method for monitoring the working state of the stay cable based on the isolated forest according to claim 1, wherein the steps 8 and 9 further comprise: the model is evaluated by using a separate validation dataset, in particular:
step a: collecting actual cable force signals of N cable-stayed bridges in various operation states;
step b: generating a cable force signal sample under a normal working condition and a preset abnormal working condition according to the finite element model; the Monte Carlo sampling method is applied, and 2N normal working condition samples and 2N abnormal working condition samples are extracted from the samples;
step c: generating additional preset abnormal cable force signal samples by adopting a trained generated GAN model, and extracting 2N samples by adopting a Monte Carlo sampling method;
step d: integrating the four types of data to form an independent verification data set containing 7N samples; in the data set, the cable force signal is taken as an input, and the corresponding health state value or abnormal state value is taken as an output;
step e: selecting four classification model evaluation indexes of precision, recall ratio and F1 score to evaluate the prediction performance and generalization capability of the GAN model in the step 6 or the isolated forest model in the step 8;
step f: if the model prediction effect is not ideal, tuning the generated countermeasure network model or the isolated forest model; tuning the GAN model, including adjusting the architecture of the GAN model, the setting of an optimizer, the learning rate, and simultaneously trying different loss functions or introducing regularization techniques to improve the performance of the model; tuning the isolated forest model includes adjusting the sample segmentation mode and the number of trees.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311151305.8A CN116881819B (en) | 2023-09-07 | 2023-09-07 | Stay cable working state monitoring method based on isolated forest |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311151305.8A CN116881819B (en) | 2023-09-07 | 2023-09-07 | Stay cable working state monitoring method based on isolated forest |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116881819A true CN116881819A (en) | 2023-10-13 |
| CN116881819B CN116881819B (en) | 2023-11-14 |
Family
ID=88262626
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311151305.8A Active CN116881819B (en) | 2023-09-07 | 2023-09-07 | Stay cable working state monitoring method based on isolated forest |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116881819B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117332341A (en) * | 2023-11-28 | 2024-01-02 | 贵州空港智能科技有限公司 | Intelligent management method and system for airplane wheel guard data based on Internet of things |
| CN117647367A (en) * | 2024-01-29 | 2024-03-05 | 四川航空股份有限公司 | Machine learning-based method and system for positioning leakage points of aircraft fuel tank |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20100238027A1 (en) * | 2007-11-16 | 2010-09-23 | Filippo Bastianini | Device for monitoring the health status of structures |
| CN110619345A (en) * | 2019-07-22 | 2019-12-27 | 重庆交通大学 | Cable-stayed bridge monitoring data validity-oriented label reliability comprehensive verification method |
| CN111638077A (en) * | 2020-05-29 | 2020-09-08 | 中铁大桥科学研究院有限公司 | Stay cable intelligent monitoring method and system based on front-end processing |
| CN111967185A (en) * | 2020-08-10 | 2020-11-20 | 哈尔滨工业大学 | Cable-stayed bridge state evaluation method based on cable force and displacement distribution correlation modeling |
| CN113255035A (en) * | 2021-05-17 | 2021-08-13 | 西安电子科技大学 | Method for fusing monitoring data of small and medium-sized bridges |
| CN113553749A (en) * | 2021-09-22 | 2021-10-26 | 深圳市城市交通规划设计研究中心股份有限公司 | Bridge health monitoring method, system, computer and storage medium |
| CN114169374A (en) * | 2021-12-10 | 2022-03-11 | 湖南工商大学 | Cable-stayed bridge stay cable damage identification method and electronic equipment |
| CN115048998A (en) * | 2022-06-13 | 2022-09-13 | 大连理工大学 | Cable-stayed bridge group cable force abnormity identification and positioning method based on monitoring data |
| CN115456104A (en) * | 2022-09-27 | 2022-12-09 | 大连理工大学 | Bridge inhaul cable abnormity identification and positioning method insensitive to sensor fault |
| CN115577436A (en) * | 2022-11-17 | 2023-01-06 | 成都理工大学 | Combined deep learning method for solving wind-induced vibration response of uncertain structure |
| CN116043659A (en) * | 2022-07-08 | 2023-05-02 | 中南大学 | High-speed railway cable-stayed bridge with bridge deck line shape adjusting function |
-
2023
- 2023-09-07 CN CN202311151305.8A patent/CN116881819B/en active Active
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20100238027A1 (en) * | 2007-11-16 | 2010-09-23 | Filippo Bastianini | Device for monitoring the health status of structures |
| CN110619345A (en) * | 2019-07-22 | 2019-12-27 | 重庆交通大学 | Cable-stayed bridge monitoring data validity-oriented label reliability comprehensive verification method |
| CN111638077A (en) * | 2020-05-29 | 2020-09-08 | 中铁大桥科学研究院有限公司 | Stay cable intelligent monitoring method and system based on front-end processing |
| CN111967185A (en) * | 2020-08-10 | 2020-11-20 | 哈尔滨工业大学 | Cable-stayed bridge state evaluation method based on cable force and displacement distribution correlation modeling |
| CN113255035A (en) * | 2021-05-17 | 2021-08-13 | 西安电子科技大学 | Method for fusing monitoring data of small and medium-sized bridges |
| CN113553749A (en) * | 2021-09-22 | 2021-10-26 | 深圳市城市交通规划设计研究中心股份有限公司 | Bridge health monitoring method, system, computer and storage medium |
| CN114169374A (en) * | 2021-12-10 | 2022-03-11 | 湖南工商大学 | Cable-stayed bridge stay cable damage identification method and electronic equipment |
| CN115048998A (en) * | 2022-06-13 | 2022-09-13 | 大连理工大学 | Cable-stayed bridge group cable force abnormity identification and positioning method based on monitoring data |
| CN116043659A (en) * | 2022-07-08 | 2023-05-02 | 中南大学 | High-speed railway cable-stayed bridge with bridge deck line shape adjusting function |
| CN115456104A (en) * | 2022-09-27 | 2022-12-09 | 大连理工大学 | Bridge inhaul cable abnormity identification and positioning method insensitive to sensor fault |
| CN115577436A (en) * | 2022-11-17 | 2023-01-06 | 成都理工大学 | Combined deep learning method for solving wind-induced vibration response of uncertain structure |
Non-Patent Citations (4)
| Title |
|---|
| ZHEN SUN: "Cumulative displacement-based detection of damper malfunction in bridges using data-driven isolation forest algorithm", 《ENGINEERING FAILURE ANALYSIS》, vol. 143, pages 1 - 15 * |
| 朱思宇等: "分片响应面法在随机振动响应分析中的应用", 《振动工程学报》, vol. 33, no. 1, pages 105 - 110 * |
| 李小珍;张黎明;张洁;: "公路桥梁与车辆耦合振动研究现状与发展趋势", 工程力学, no. 03, pages 230 - 240 * |
| 马天立: "基于深度学习的桥梁健康监测数据有效性诊断研究", 《中国优秀硕士学位论文全文数据库 工程科技Ⅱ辑》, pages 034 - 493 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117332341A (en) * | 2023-11-28 | 2024-01-02 | 贵州空港智能科技有限公司 | Intelligent management method and system for airplane wheel guard data based on Internet of things |
| CN117332341B (en) * | 2023-11-28 | 2024-02-02 | 贵州空港智能科技有限公司 | Intelligent management method and system for airplane wheel guard data based on Internet of things |
| CN117647367A (en) * | 2024-01-29 | 2024-03-05 | 四川航空股份有限公司 | Machine learning-based method and system for positioning leakage points of aircraft fuel tank |
| CN117647367B (en) * | 2024-01-29 | 2024-04-16 | 四川航空股份有限公司 | Machine learning-based method and system for positioning leakage points of aircraft fuel tank |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116881819B (en) | 2023-11-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN116881819B (en) | Stay cable working state monitoring method based on isolated forest | |
| Jiang et al. | Two-stage structural damage detection using fuzzy neural networks and data fusion techniques | |
| CN111274737A (en) | Method and system for predicting remaining service life of mechanical equipment | |
| CN111859724A (en) | Hybrid-driven ballastless track fatigue life prediction method and system | |
| CN110704801A (en) | Bridge cluster structure operation safety intelligent monitoring and rapid detection complete technology | |
| Bao et al. | A deep transfer learning network for structural condition identification with limited real-world training data | |
| Mantawy et al. | Convolutional neural network based structural health monitoring for rocking bridge system by encoding time‐series into images | |
| CN115146842A (en) | Multivariate time series trend prediction method and system based on deep learning | |
| Son et al. | Deep learning-based anomaly detection to classify inaccurate data and damaged condition of a cable-stayed bridge | |
| Rosso et al. | Review on deep learning in structural health monitoring | |
| CN113935384B (en) | Bridge health monitoring method and system for self-adaptive decomposition and recognition of signals | |
| Huang et al. | Physics-guided deep neural network for structural damage identification | |
| CN113742983A (en) | Long-span structural damage identification method based on depth self-encoder neural network | |
| CN116050190B (en) | Product performance and degradation state analysis method based on digital twinning | |
| CN116663425A (en) | Bridge vehicle-induced vibration prediction method based on modal expansion and residual CNN-LSTM network | |
| CN116776267A (en) | Unsupervised data processing method and system for bridge construction control | |
| CN114550460B (en) | Rail transit anomaly detection method and device and storage medium | |
| CN114723042B (en) | Bridge damage position and damage degree identification method based on optimized neural network | |
| CN114118249A (en) | Structure damage diagnosis method based on optimized stacked self-encoder and multi-signal fusion | |
| Shen et al. | A boom damage prediction framework of wheeled cranes combining hybrid features of acceleration and Gaussian process regression | |
| Chang et al. | Intelligent diagnosis model of traction seat of urban rail vehicle based on harris hawks optimization | |
| Desjardins | Structural health assessment of bridges by long-term vibration monitoring and automated operational modal analysis | |
| CN117494026B (en) | Method, system and storage medium for positioning icing fault of power transmission line under cold and tidal weather | |
| CN114861458B (en) | Bridge traffic safety rapid evaluation method based on mapping relation proxy model | |
| Zhu et al. | Structural damage identification considering uncertainties in nonuniform measurement conditions based on convolution neural networks |
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 |