CN116542146A - Bridge monitoring temperature field-strain field space-time correlation model and health diagnosis method - Google Patents
Bridge monitoring temperature field-strain field space-time correlation model and health diagnosis method Download PDFInfo
- Publication number
- CN116542146A CN116542146A CN202310507996.4A CN202310507996A CN116542146A CN 116542146 A CN116542146 A CN 116542146A CN 202310507996 A CN202310507996 A CN 202310507996A CN 116542146 A CN116542146 A CN 116542146A
- Authority
- CN
- China
- Prior art keywords
- strain
- temperature
- field
- bridge
- data
- 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
- 238000012544 monitoring process Methods 0.000 title claims abstract description 52
- 238000000034 method Methods 0.000 title claims abstract description 35
- 230000036541 health Effects 0.000 title claims abstract description 27
- 238000003745 diagnosis Methods 0.000 title claims abstract description 22
- 239000000835 fiber Substances 0.000 claims abstract description 22
- 238000007781 pre-processing Methods 0.000 claims abstract description 5
- 238000004364 calculation method Methods 0.000 claims description 21
- 238000012549 training Methods 0.000 claims description 21
- 230000005540 biological transmission Effects 0.000 claims description 15
- 230000006870 function Effects 0.000 claims description 13
- 238000012360 testing method Methods 0.000 claims description 11
- 238000004458 analytical method Methods 0.000 claims description 9
- 239000013598 vector Substances 0.000 claims description 9
- 238000010276 construction Methods 0.000 claims description 8
- 230000006378 damage Effects 0.000 claims description 8
- 230000000644 propagated effect Effects 0.000 claims description 6
- 238000003062 neural network model Methods 0.000 claims description 5
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 claims description 4
- CIWBSHSKHKDKBQ-JLAZNSOCSA-N Ascorbic acid Chemical compound OC[C@H](O)[C@H]1OC(=O)C(O)=C1O CIWBSHSKHKDKBQ-JLAZNSOCSA-N 0.000 claims description 3
- 230000005856 abnormality Effects 0.000 claims description 3
- 230000003287 optical effect Effects 0.000 claims description 3
- 238000005457 optimization Methods 0.000 claims description 3
- 238000005070 sampling Methods 0.000 claims description 3
- 238000002405 diagnostic procedure Methods 0.000 claims 1
- 238000013507 mapping Methods 0.000 claims 1
- 238000007405 data analysis Methods 0.000 abstract description 5
- 230000008901 benefit Effects 0.000 abstract description 2
- 230000035882 stress Effects 0.000 description 17
- 230000008859 change Effects 0.000 description 14
- 201000010099 disease Diseases 0.000 description 9
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 9
- 230000009471 action Effects 0.000 description 7
- 238000005516 engineering process Methods 0.000 description 7
- 239000004567 concrete Substances 0.000 description 6
- 238000011161 development Methods 0.000 description 6
- 230000018109 developmental process Effects 0.000 description 5
- 238000012423 maintenance Methods 0.000 description 4
- 230000011218 segmentation Effects 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 230000007797 corrosion Effects 0.000 description 2
- 238000005260 corrosion Methods 0.000 description 2
- 230000001808 coupling effect Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000008030 elimination Effects 0.000 description 2
- 238000003379 elimination reaction Methods 0.000 description 2
- 230000007613 environmental effect Effects 0.000 description 2
- 238000004880 explosion Methods 0.000 description 2
- 239000004973 liquid crystal related substance Substances 0.000 description 2
- 238000005259 measurement Methods 0.000 description 2
- 239000011513 prestressed concrete Substances 0.000 description 2
- 238000010008 shearing Methods 0.000 description 2
- 229910001294 Reinforcing steel Inorganic materials 0.000 description 1
- 238000009825 accumulation Methods 0.000 description 1
- 230000032683 aging Effects 0.000 description 1
- 238000003763 carbonization Methods 0.000 description 1
- 239000002131 composite material Substances 0.000 description 1
- 238000010219 correlation analysis Methods 0.000 description 1
- 230000008878 coupling Effects 0.000 description 1
- 238000010168 coupling process Methods 0.000 description 1
- 238000005859 coupling reaction Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000006735 deficit Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 230000035784 germination Effects 0.000 description 1
- 230000000977 initiatory effect Effects 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 230000014759 maintenance of location Effects 0.000 description 1
- 238000004643 material aging Methods 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000003449 preventive effect Effects 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 230000002787 reinforcement Effects 0.000 description 1
- 230000008439 repair process Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—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
- 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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
- G06N3/0442—Recurrent networks, e.g. Hopfield networks characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- 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
Abstract
The invention belongs to the technical application field of bridge monitoring data analysis, and particularly discloses a bridge monitoring temperature field-strain field space-time correlation model and a health diagnosis method. Meanwhile, the influence of the vehicle load is eliminated by preprocessing the monitoring strain, the nonlinear and time-lag characteristics between the temperature field and the strain field are described by using a BiLSTM network, the prediction error is only related to bridge structural parameters, and the method can be used for bridge health diagnosis. The invention considers the influence of the temperature field on the strain, and realizes the multipoint health diagnosis of the key stress part and the vulnerable part of the bridge by utilizing the advantage of wide arrangement range of the strain measuring points of the distributed fiber bragg grating.
Description
Technical Field
The invention belongs to the technical application field of bridge monitoring data analysis, relates to a bridge monitoring temperature field-strain field space-time correlation model and a health diagnosis method, and particularly relates to a BiLSTM-based bridge monitoring temperature field-strain field space-time correlation model and a health diagnosis method.
Background
Firstly, bridges have more or less certain defects in construction, which cause various problems during the operation period; secondly, with the increasing of traffic volume, the design standard during bridge construction cannot meet the current operation requirement, and many bridges are in overload operation states; in addition, under the continuous action of load and environmental conditions, the problems of fatigue, corrosion, material aging and the like are increasingly prominent, the damage accumulation and resistance attenuation of the structural member are continuously increased, and once the damage of important parts is developed to a critical level, proper maintenance and reinforcement measures are not timely adopted, so that the integral structure is seriously damaged. In order to ensure the safety of bridges and users, the structural condition of the bridges needs to be mastered in time by mining the bridge health monitoring data so as to cope with daily operation and emergency situations.
Research on intelligent monitoring and intelligent diagnosis of diseases of bridges also has many challenges:
(1) The existing point type and distributed sensing technology cannot meet the dynamic distributed test requirement of bridge monitoring, and the high-durability long-distance dynamic distributed sensing technology suitable for bridge structures needs to be applied to bridge monitoring. For example, if the main disease of the concrete box girder bridge, namely cracks, are not found in time or maintenance measures are not in place or are not in time, the length and the width of the bridge are continuously developed due to local stress redistribution caused by the cracks, on one hand, the bridge concrete cracks can have great influence on the overall structural strength and the stability of the bridge, and further influence the normal use and the durability of the bridge, and on the other hand, the cracks can lead to carbonization of concrete and corrosion of reinforcing steel bars, thereby threatening the safety of the bridge structure and the service life of the bridge. Aiming at the characteristics of the general quantity, long service life and wide distribution and randomness of disease germination of bridge engineering, if a bridge monitoring system based on TDM and WDM long-distance distributed fiber bragg grating sensor networks can be applied to monitor the strain and the temperature field of a concrete structure, the requirements of timely finding cracks of a prestressed concrete beam bridge can be well met.
(2) Strain data implies bridge local health state information, and an effective intelligent diagnosis technology for bridge diseases under the multi-factor coupling effect based on the strain data is not available at present. Under the combined actions of the temperature change of the bridge, the randomness of the vehicle load and the like, the strain has strong time-varying characteristics, and the aging influence model of different load factor changes and the coupling effect thereof on strain data is lacking at present. How to remove the influence of the load and the temperature of the vehicle on the strain change and establish a data analysis model only related to structural parameters is a key of the bridge health diagnosis technology based on the strain.
(3) The bridge is used as a large structure for coordinating deformation, correlation analysis between strain of one channel and data of a nearby temperature channel is difficult to capture the complex influence of temperature field change on the strain. The establishment of a nonlinear time-lag model by establishing a field correspondence between a temperature field and a strain field is important to grasp the overall change rule of the bridge in the temperature field and to perform large-scale real-time health diagnosis.
Disclosure of Invention
In order to solve the technical problems in the background art, the invention provides a bridge monitoring temperature field-strain field space-time correlation model and a health diagnosis method which can realize the area monitoring of stress distribution, crack generation and development of important parts of a concrete large-span continuous box girder bridge.
In order to achieve the above purpose, the present invention adopts the following technical scheme:
a construction method of a space-time correlation model of a bridge monitoring temperature field and a strain field is characterized by comprising the following steps: the construction method of the space-time correlation model of the bridge monitoring temperature field and the strain field comprises the following steps:
1) Determining the strain and temperature measuring point positions of a distributed fiber grating sensing network and arranging the sensing network;
2) Transmitting and storing data; the data transmission mode is divided into wired transmission and wireless transmission, data acquired by the fiber bragg grating demodulator are transmitted to a data server, and the server is utilized to store the transmitted fiber bragg grating data;
3) Preprocessing the monitored temperature and strain data;
4) And establishing a temperature field-strain field space-time correlation BiLSTM model.
Preferably, the specific implementation manner of the step 3) adopted by the invention is as follows:
3.1 Extracting the temperature-induced strain; the strain sensor arranged at the wheel pressure position can capture the strain caused by the temperature and the vehicle load at the same time, and the temperature-induced strain is extracted according to the strain data characteristics;
3.2 Downsampling the temperature-induced strain data extracted in the step 3.1);
3.3 Normalized analysis; in order to improve the network training efficiency, the temperature and strain data of each channel obtained by sampling in the step 3.2) are normalized respectively, and the data are mapped into the [0,1] interval:
（1）
wherein:is the total number of temperature channels; />Is the total number of strain channels;
indicate->Time->Monitoring temperatures of the individual temperature channels;
、/>respectively +.>Maximum and minimum values of the individual temperature channels;
indicate->Time->Monitoring temperatures of the individual temperature channels;
、/>respectively +.>Maximum and minimum values of the individual temperature channels;
3.4 Data segmentation; dividing the normalized data in the step 3.3) into a training set and a testing set, taking the temperature and the temperature-induced change annual cycle characteristics into consideration, wherein at least half cycle duration including the highest temperature and the lowest temperature is required to be used as the training set, and the rest data are used as the testing set.
Preferably, the specific implementation manner of the step 4) adopted by the invention is as follows:
4.1 Building a neural network model with two BiLSTM layers and a full connection layer;
wherein the BiLSTM layer comprises forward hidden LSTM cells propagated from front to backAnd backward hidden LSTM cell propagated backward and forward +.>The method comprises the steps of carrying out a first treatment on the surface of the The calculation from input to output is as follows:
，/>（2）
，/>（3）
（4）
（5）
（6）
wherein: superscriptIndicate->Time;
and->Are respectively in the first layer of the network>A forward hidden unit and a backward hidden unit at the moment;
and->Are respectively in the second layer of the network>A forward hidden unit and a backward hidden unit at the moment;
is a full connection layer>A value of time of day;
and->The weight and the bias of the full connection layer are respectively;
is a hyperbolic tangent function;
Lrepresenting predicted valuesAnd measured value->A loss function between;
a single LSTM cell in the architecture includes state cell C t() And three gating operations (forget gate)The input gate i and the output gate o), the gating operation can avoid the problems of gradient elimination and gradient explosion in the error back propagation calculation, and the training convergence efficiency is improved; output door->The calculation is as follows:
（7）
wherein:representing vector h t(-1) Vector x t() Is spliced by (a)>And->The weights and offsets of the output gates, respectively, +.>Is a sigmoid function;
forgetting doorAnd input gate i calculates and outputs gate +.>Calculating similarity;
the state unit C t() The calculation is as follows:
（8）
（9）
wherein:and->Candidate state units->Weights and offsets of (2);
the hidden state h t() The calculation is as follows:
（10）
wherein:representing a point multiplication operation between two vectors; loss functionLTo predict the mean square error between strain and measured strain,
（11）
4.2 Inputting training set data for training, and iteratively adjusting various parameters in the network by using a back propagation algorithm, wherein the back propagation algorithm is preferably calculated by using an Adam optimizer, and the method comprises the following steps of:
（12）
wherein:general meaning ofiStep one, iterative optimization parameters; />Is the learning rate;
and iteratively optimizing each parameter until the loss function converges to obtain a BiLSTM-based temperature field-strain field correlation model.
Preferably, the wire transmission in the step 2) adopted by the invention is transmitted through an Ethernet switch and an optical cable; the wireless transmission is transmitted through a wireless 4G/5G network.
The bridge monitoring temperature field-strain field space-time correlation model is constructed based on the bridge monitoring temperature field-strain field space-time correlation model construction method.
A bridge health diagnosis method based on a space-time correlation model of a bridge monitoring temperature field and a strain field is characterized by comprising the following steps:
1) Acquiring a space-time correlation model of a bridge monitoring temperature field-strain field;
2) And diagnosing the health state of the bridge based on the space-time correlation model of the bridge monitoring temperature field and the strain field obtained in the step 1).
Preferably, the specific implementation manner of the step 2) adopted by the invention is as follows: according to the trained model, inputting temperature data of the test set, outputting predicted strain, and outputting the predicted strainAnd actually measured strain->Error betweenPerforming control chart analysis when a certain strain channel error exceeds the training set error>At the threshold value, it is diagnosed that damage has occurred in the vicinity of the strain sensor.
Preferably, the threshold value used in the present invention is a value corresponding to 99.7% of the probability of occurrence of an abnormality.
The beneficial effects of the invention are as follows:
the invention aims at the characteristics of large quantity, long service life and wide distribution and randomness of disease initiation of bridge engineering, applies the strain and temperature field monitoring of a bridge structure based on a TDM and WDM long-distance distributed fiber grating sensing network, can improve the current situation that the existing real-time monitoring technology can only monitor the structural state of individual points of a key section, can well meet the requirement of timely finding out cracks of a prestressed concrete bridge, and provides the distribution and development condition of the bridge cracks in real time and accurately so as to timely and effectively take preventive maintenance measures, inhibit further generation and development of the cracks, improve the service life of the bridge and reduce the maintenance cost of the whole life cycle. And analyzing bridge temperature and strain data monitored by the distributed fiber grating sensing network, and establishing a degree field-strain field space-time correlation model based on the BiLSTM network. The strain of the bridge in the normal state is changed under the combined action of the load of the vehicle and the temperature field, and the influence of the vehicle and the temperature is removed when the bridge is subjected to health diagnosis through the strain. The influence of vehicle load is eliminated by preprocessing the data, the nonlinear and time-lag characteristics between the temperature field and the strain field are described by using the BiLSTM network, the prediction error is only related to bridge structural parameters, and the method can be used for bridge health diagnosis. Different from the analysis between single-point temperature and strain in the past, the model considers the influence of the temperature field on strain, and utilizes the advantage of wide arrangement range of strain measuring points of the distributed fiber bragg grating to realize the multipoint health diagnosis of key stress parts and vulnerable parts of the bridge. The invention aims to provide a bridge monitoring temperature field-strain field space-time correlation model based on BiLSTM and a health diagnosis method. On one hand, the distributed fiber grating sensor network based on TDM and WDM is applied to bridge monitoring, and the sensor network adopts a wavelength division and time division composite demodulation technology, and has stronger sensitivity to crack occurrence within a range of 12cm from the crack. On the other hand, the invention establishes a nonlinear time-lag field corresponding model between a temperature field and a strain field, and the model prediction error is only related to structural parameters. The invention can realize the area monitoring of stress distribution, crack generation and development of important parts of the concrete large-span continuous box girder bridge, realizes the on-line monitoring of the environment condition normalization of the bridge response of actual measurement of bridge strain by using a BilSTM model, is beneficial to solving the problems of inconsistent parameter level or standard and the like caused by environmental factors, and establishes the intelligent diagnosis technology of bridge diseases under the effect of multi-factor coupling.
Drawings
FIG. 1 is a schematic diagram of a BiLSTM network architecture employed in the present invention;
FIG. 2 is a schematic diagram of an LSTM cell used in the present invention;
FIG. 3 is a flow chart of a BiLSTM-based bridge monitoring temperature field-strain field space-time correlation model and a health diagnosis method provided by the invention;
FIG. 4 is a schematic illustration of the placement of certain bridges based on the method provided by the present invention;
FIG. 5 is a cross-sectional strain and temperature station layout at a mid-span, 1/4-span, pier top cross-section, respectively, based on the method provided by the present invention;
FIG. 6 is an expanded view of temperature and strain gauge point locations at a midspan section, a 1/4 midspan section, a pier top section, respectively, based on the method provided by the present invention;
FIG. 7 is a graph of temperature and strain measurement point placement at a fracture site based on the method provided by the present invention;
FIG. 8 is an extracted temperature induced strain map based on the method provided by the present invention;
FIG. 9 is an example of health diagnostic impairment based on the methods provided by the present invention;
fig. 10 is a view of a bridge damage scene.
Detailed Description
The invention provides a method for constructing a bridge monitoring temperature field-strain field space-time correlation model based on BiLSTM, the model and a health diagnosis method based on the model, wherein the method for constructing the model comprises the following steps:
the first step: and determining the positions of strain and temperature measuring points of the distributed fiber bragg grating sensing network and arranging the sensing network.
(1) The fiber bragg grating strain sensing network measuring point arrangement should follow the following principle:
1) The crack generation and development in the main bridge box girder and the crack repair re-development condition can be monitored in real time;
2) It should be possible to monitor web diagonal slits and top and bottom plate longitudinal slits, as well as left and right web slits;
3) The beam longitudinal strain field can be acquired;
4) Sensors on each section should be respectively arranged according to stress characteristics and disease conditions of each plate;
and arranging strain sensors on all the sections according to the stress characteristics and disease conditions of all the plates.
(1) For the roof in the box, because the roof directly bears the wheel load and the function of the whole structure, the sensors are required to be distributed longitudinally and transversely.
(2) For the bottom plate in the box, the bottom plate basically only bears the integral action of the structure, the stress direction is relatively clear, and the sensor is required to be longitudinally arranged.
(3) For the web in the box, the web is relatively complex in stress due to the fact that the web is subjected to the action of positive stress and shearing stress, and can be monitored in a strain relief mode.
(4) And tracking and monitoring the generated cracks according to the actual conditions of the site, and arranging measuring points near the cracks for monitoring the stress change conditions near the cracks.
(2) The fiber bragg grating temperature sensing network measuring point arrangement should follow the following principle:
1) A temperature compensation sensor is arranged at the position close to the fiber bragg grating strain sensor;
2) It should be possible to monitor the difference between web, floor, and left and right web temperatures;
3) The longitudinal temperature field of the beam body can be acquired;
and arranging the temperature sensors on each section according to the characteristics of the bridge temperature field.
(1) Temperature sensors are arranged near strain sensors at the positions of the top plate in the box, the bottom plate in the box, the web plate in the box, the cracks and the like.
(2) Temperature sensors in box girders of key cross sections (such as midspan, bridge piers, 1/4 bridge spans and 3/4 bridge spans), at least 3 measuring points are arranged at the positions of a top plate, a bottom plate, a web plate and the like, and the temperature gradient of each side face in one cross section is monitored, and the specific arrangement is shown in figure 4.
(3) And a temperature sensor is uniformly arranged along the longitudinal direction to capture the temperature difference of different longitudinal positions of the bridge.
And a second step of: and (5) data transmission and storage. The data transmission mode is divided into wired transmission (through an Ethernet switch and an optical cable) and wireless transmission (wireless 4G/5G network), data acquired by the fiber grating demodulator are transmitted to a data server, and the server is used for storing the transmitted fiber grating data.
And a third step of: and preprocessing the monitored temperature and strain data.
(1) Extracting the temperature-induced strain. If the in-box roof strain sensor is arranged at the wheel pressure position, the temperature measured by the temperature sensor and the strain caused by the vehicle load can be captured at the same time, the vehicle-induced strain captured by the strain sensor at the non-wheel pressure position is weaker or not, and the temperature-induced strain is extracted according to the strain data characteristics.
(2) And (5) downsampling data. Taking the characteristic of slower temperature and temperature induced strain change into consideration, downsampling the temperature and strain data to save calculation resources.
(3) And (5) normalizing analysis. In order to improve the network training efficiency, the temperature and strain data of each channel are normalized respectively, and the data are mapped into the [0,1] interval:
（1）
wherein, the liquid crystal display device comprises a liquid crystal display device,for the total number of temperature channels (number of sensors measuring temperature), +.>For the total number of strain channels>Indicate->Time->The monitored temperature of the individual temperature channels (measured by the temperature sensor),>、/>respectively +.>Maximum, minimum of the individual temperature channels, < ->Indicate->Time->The monitored temperature of the individual temperature channels is determined,、/>respectively +.>Maximum and minimum of each temperature channel.
(4) And (5) data segmentation. The normalized data are divided into a training set and a test set, and at least half period duration (half year) including the highest temperature (the maximum value of the temperature channel) and the lowest temperature (the minimum value of the temperature channel) is required to be used as the training set and the rest data are used as the test set in consideration of the temperature and the temperature induced strain annual period characteristics.
Fourth step: and establishing a temperature field-strain field space-time correlation BiLSTM model.
(1) A neural network model is built with two BiLSTM layers and one fully connected layer. Due to the influences of factors such as specific heat capacity, sunlight and humidity of the bridge, certain hysteresis exists between temperature change and strain change (strain measured by a strain sensor) in time, and the temperature-induced strain change of a certain measuring point is influenced by the temperature change of the measuring point and the temperature distribution of the whole temperature field. Considering the space-time correlation between the temperature field and the strain field, a BiLSTM network capable of describing the space-time correlation between different variables is selected to learn the relation between the temperature field and the strain field.
BiLSTM network architecture As shown in FIG. 1, biLSTM layer contains forward hidden LSTM cells propagated from front to backAnd backward hidden LSTM cell propagated backward and forward +.>. The calculation from input to output is as follows:
，/>（2）
，/>（3）
（4）
（5）
（6）
wherein: superscriptIndicate->Time;
and->Are respectively in the first layer of the network>A forward hidden unit and a backward hidden unit at the moment;
and->Are respectively in the second layer of the network>A forward hidden unit and a backward hidden unit at the moment;
is a full connection layer>A value of time of day;
and->The weight and the bias of the full connection layer are respectively;
is a hyperbolic tangent function;
Lrepresenting predicted valuesAnd measured value->A loss function therebetween.
The LSTM cells are described in detail below.
A single LSTM unit in the BiLSTM network architecture is shown in fig. 2, and the calculation process is illustrated by taking the forward hidden LSTM unit h as an example. Forward hidden LSTM cell includes state cell C t() And three gating operations (forget gate)Input gate i and output gate o), the gating operation can avoid the gradient elimination and gradient explosion problems in the error back propagation calculation, and the training convergence efficiency is improved.
Output doorThe calculation is as follows:
（7）
wherein:representing vector h t(-1) Vector x t() Is spliced by the steps of (1);
and->The weight and bias of the output gate respectively;
is a sigmoid function.
Forgetting doorAnd input gate i calculates and outputs gate +.>The calculations are similar.
State unit C t() The calculation is as follows:
（8）
（9）
wherein the method comprises the steps ofAnd->Candidate state units->Is included in the weight and bias of (1).
Hidden state h t() The calculation is as follows:
（10）
wherein:representing a point multiplication operation between two vectors.
Loss functionLTo predict strainAnd actually measured strain->Mean square error between:
（11）
(2) Inputting training set data for training, iteratively adjusting each parameter in the network by using a back propagation algorithm, and calculating by using an Adam optimizer as follows:
（12）
wherein:general meaning ofiStep one, iterative optimization parameters; />Is the learning rate.
And iteratively optimizing each parameter until the loss function converges to obtain a BiLSTM-based temperature field-strain field correlation model.
Based on the model, the invention also provides a method for diagnosing bridge health, which specifically comprises the following steps: according to the trained model, temperature data of the test set is input, predicted strain (calculated according to the model) is output, and the predicted strain is calculatedAnd actually measured strain->Error between->Performing control chart analysis when a certain strain channel error exceeds the training set error>When a threshold value (corresponding to 99.7% of probability of occurrence of abnormality) is reached, it is diagnosed that damage has occurred in the vicinity of the strain sensor.
The following describes in detail the technical scheme provided by the invention through the attached drawings and the embodiments:
examples
The large-span continuous beam bridge monitoring and data analysis method based on the distributed fiber bragg grating sensor network is implemented on a certain oversized five-span rigid frame continuous combined system beam bridge.
The main bridge is 127m+3×232m+127m, 6 lanes are shared, and the full width of the bridge deck is 34.5m, as shown in fig. 3. The upper structure of the bridge is divided into left and right parts, the single girder is a single-box single-chamber box girder, the top width of the box girder is 16.6m, the bottom width is 8m, the length of a single-side cantilever flange plate is 4.3m, the thickness of a box girder top plate is 25-45 cm, the thickness of a bottom plate is 30-135 cm, and the thickness of a web plate is 45-100 cm. The root of the box Liang Lianggao is 12.5m and 4m in the midspan, and the lower edge of the beam bottom is changed according to a quadratic parabola.
The method comprises the following steps:
the first step: and determining the positions of strain and temperature measuring points of the bridge distributed fiber bragg grating sensing network and arranging the sensing network.
For the arrangement of the sensors on each section, the sensors are arranged according to the stress characteristics and disease conditions of each plate, as shown in fig. 5-7.
(1) For the roof in the box, the width is 6-7.1m, and because the roof directly bears the wheel load and the effect of the whole structure, the sensors are arranged in the longitudinal direction and the transverse direction, the distance between the sensors is 0.75m, 9 measuring points are arranged on the roof in the box with each section, and each measuring point is in two directions.
(2) For the bottom plate in the box, the width is 6-7.1m, the bottom plate basically only bears the integral action of the structure, and the stress direction is relatively definite, so that the sensors are arranged longitudinally, the distance between the sensors is 1.5m, 5 measuring points are arranged on the bottom plate in the box with each section, and one direction is used for each measuring point.
(3) For the inner web of the box, the height is 3.25-11.7m, and the inner web is relatively complex in stress due to the fact that the inner web is subjected to the action of positive stress and shearing stress, so that the inner web is monitored in a strain relief mode. The arrangement is that the cross section mandrel is used as the center, the array arrangement is carried out with 0.75m as the interval, 3-12 measuring points are arranged on each cross section according to the heights of webs in the box with different cross sections, and three directions of each measuring point are arranged.
(4) And for the generated cracks, tracking and monitoring the 19 th cross-span middle section roof crack (closed) of the right frame according to the actual condition of the site, and arranging measuring points near the cracks for monitoring the stress change condition near the cracks.
And a second step of: the bridge adopts a 4G network for wireless transmission, data acquired by a fiber grating demodulator is transmitted to a data server, the transmitted fiber grating data is stored by using 1 server, and the temperature and strain sampling frequency of the bridge is 10Hz.
And a third step of: the bridge temperature, strain data are preprocessed.
(1) Extracting the temperature-induced strain. Analyzing the bridge data, the vehicle induced strain ratio is small, and median filtering (window length is 1min data, window length is required to be odd, window length is taken to be 601) is carried out on the original data to extract the temperature induced strain, as shown in fig. 8.
(2) And (5) downsampling data. The bridge temperature and temperature induced strain were sampled once for 10min and the data length per day was 144 in order to reduce the amount of calculation on the basis of the retention of the temperature and temperature induced strain trend.
(3) And (5) normalizing analysis. The temperature and the strain data of each channel are normalized to the [0,1] interval according to the formula (1). The bridge has 363 temperature channels and 363 strain channels.
(4) And (5) data segmentation. Taking one day data length (144 steps) as one sample, taking account of temperature and temperature-induced change year cycle characteristics, taking 2015-2016 years data as training set (623 samples) and 2017-2019 years data as test set (702 samples).
Fourth step: and establishing a temperature field-strain field space-time correlation BiLSTM model.
A neural network model with two BiLSTM layers and one full-connection layer is built, the input step length and the output step length are 144, and the number of units of each layer is 55. And inputting training set data by using an Adam optimizer, and optimizing parameters of the BiLSTM model until convergence.
Fifth step: the bridge is diagnosed for health according to the trained model. Inputting temperature data of the test set, outputting predicted strain, and outputting the predicted strainAnd actually measured strain->Error between->The control chart analysis is performed, the damage is shown in fig. 9, the prediction error gradually exceeds the threshold range, the damage appears near the strain gauge, the bridge is removed for inspection, a new crack appears at the place, as shown in fig. 10 (the position where the new crack appears in fig. 10, and the corresponding change appears through the data analysis), and the analysis result is consistent. />
Claims (8)
1. A construction method of a space-time correlation model of a bridge monitoring temperature field and a strain field is characterized by comprising the following steps: the construction method of the space-time correlation model of the bridge monitoring temperature field and the strain field comprises the following steps:
1) Determining the strain and temperature measuring point positions of a distributed fiber grating sensing network and arranging the sensing network;
2) Transmitting and storing data; the data transmission mode is divided into wired transmission and wireless transmission, data acquired by the fiber bragg grating demodulator are transmitted to a data server, and the server is utilized to store the transmitted fiber bragg grating data;
3) Preprocessing the monitored temperature and strain data;
4) And establishing a temperature field-strain field space-time correlation BiLSTM model.
2. The method for constructing the space-time correlation model of the bridge monitoring temperature field and the strain field according to claim 1, wherein the specific implementation manner of the step 3) is as follows:
3.1 Extracting the temperature-induced strain; the strain sensor arranged at the wheel pressure position can capture the strain caused by the temperature and the vehicle load at the same time, and the temperature-induced strain is extracted according to the strain data characteristics;
3.2 Downsampling the temperature-induced strain data extracted in the step 3.1);
3.3 Normalizing the temperature and strain data of each channel obtained by sampling in the step 3.2), and mapping the data into the [0,1] interval:
（1）
wherein:is the total number of temperature channels;
is the total number of strained channels;
indicate->Time->Monitoring temperatures of the individual temperature channels;
is->Maximum value of each temperature channel;
is->Minimum value of each temperature channel;
indicate->Time->Monitoring temperatures of the individual temperature channels;
is->Maximum value of each temperature channel;
is->Minimum value of each temperature channel;
3.4 Dividing the normalized data of step 3.3) into a training set and a test set, the training set comprising at least a maximum temperature and a minimum temperature for a half-cycle duration.
3. The method for constructing the space-time correlation model of the bridge monitoring temperature field and the strain field according to claim 2, wherein the specific implementation manner of the step 4) is as follows:
4.1 Building a neural network model with two BiLSTM layers and a full connection layer;
wherein the BiLSTM layer comprises forward hidden LSTM cells propagated from front to backAnd backward hidden LSTM cell propagated backward and forward +.>The method comprises the steps of carrying out a first treatment on the surface of the The calculation from input to output is as follows:
，/>（2）
，/>（3）
（4）
（5）
（6）
wherein: superscriptIndicate->Time;
and->Are respectively in the first layer of the network>A forward hidden unit and a backward hidden unit at the moment;
and->Are respectively in the second layer of the network>A forward hidden unit and a backward hidden unit at the moment;
is a full connection layer>A value of time of day;
and->The weight and the bias of the full connection layer are respectively;
is a hyperbolic tangent function;
Lrepresentation ofPredictive valueAnd measured value->A loss function between;
a single LSTM cell in the architecture includes state cell C t() And three gating operations, including a forget gateAn input gate i and an output gate o;
wherein the output doorThe calculation method is as follows:
（7）
wherein:representing vector h t(-1) Vector x t() Is spliced by the steps of (1);
and->The weight and bias of the output gate respectively; />Is a sigmoid function;
the forgetting doorThe calculation of the input gate i and the calculation of the output gate i are equal to each other>The calculation mode of (2) is the same;
the state unit C t() The calculation mode of (2) is as follows:
（8）
（9）
wherein:and->Candidate state units->Weights and offsets of (2);
the hidden state h t() The calculation is as follows:
（10）
wherein:representing a point multiplication operation between two vectors; loss functionLTo predict the mean square error between strain and measured strain,
（11）
4.2 Inputting the training set obtained in the step 3.4) into the neural network model constructed in the step 4.1) for training, and iteratively adjusting each parameter in the network by using a back propagation algorithm, wherein the back propagation algorithm is preferably calculated by using an Adam optimizer, and the method comprises the following steps of:
（12）
wherein:general meaning ofiStep one, iterative optimization parameters; />Is the learning rate;
and iteratively optimizing each parameter until the loss function converges to obtain a BiLSTM-based temperature field-strain field correlation model.
4. The method for constructing the space-time correlation model of the bridge monitoring temperature field and the strain field according to claim 1, 2 or 3, wherein the method comprises the following steps: the wired transmission in the step 2) is transmitted through an Ethernet switch and an optical cable; the wireless transmission is transmitted through a wireless 4G/5G network.
5. A bridge monitoring temperature field-strain field space-time correlation model constructed based on the bridge monitoring temperature field-strain field space-time correlation model construction method according to claim 1 or 2 or 3 or 4.
6. A bridge health diagnosis method based on the bridge monitoring temperature field-strain field space-time correlation model according to claim 5, which is characterized in that:
1) Acquiring a bridge monitoring temperature field-strain field space-time correlation model according to claim 5;
2) And diagnosing the health state of the bridge based on the space-time correlation model of the bridge monitoring temperature field and the strain field obtained in the step 1).
7. Bridge health of the bridge monitoring temperature field-strain field spatiotemporal correlation model of claim 6A diagnostic method characterized by: the specific implementation manner of the step 2) is as follows: according to the space-time correlation model of the bridge monitoring temperature field and the strain field obtained in the step 1), inputting temperature data of a test set, outputting predicted strain, and carrying out strain predictionAnd actually measured strain->Error between->Performing control chart analysis when a certain strain channel error exceeds the training set error>At the threshold value, it is diagnosed that damage has occurred in the vicinity of the strain sensor.
8. The bridge health diagnosis method of the bridge monitoring temperature field-strain field space-time correlation model according to claim 7, wherein: the threshold is a value corresponding to 99.7% of the probability of occurrence of an abnormality.
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2023100193681 | 2023-01-06 | ||
| CN202310019368 | 2023-01-06 |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116542146A true CN116542146A (en) | 2023-08-04 |
Family
ID=87455529
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310507996.4A Pending CN116542146A (en) | 2023-01-06 | 2023-05-08 | Bridge monitoring temperature field-strain field space-time correlation model and health diagnosis method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116542146A (en) |
Citations (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103134432A (en) * | 2013-01-25 | 2013-06-05 | 南京大学 | Bridge displacement monitoring system based on optical fiber sensors |
| CN105675056A (en) * | 2016-02-22 | 2016-06-15 | 东南大学 | Long-span bridge damage recognition method |
| WO2017202139A1 (en) * | 2016-05-26 | 2017-11-30 | 东南大学 | Bridge damage identification method based on long-gauge-length strain influence envelope |
| CN108052770A (en) * | 2017-12-28 | 2018-05-18 | 大连理工大学 | A kind of Loads of Long-span Bridges girder performance method for early warning for considering time-varying effecting |
| CN108399277A (en) * | 2018-01-24 | 2018-08-14 | 华南理工大学 | A kind of bridge damnification recognition method based on temperature with strain relevance |
| CN108763763A (en) * | 2018-05-28 | 2018-11-06 | 东南大学 | A kind of bridge structure strain-responsive abnormity early warning method |
| CN110533007A (en) * | 2019-09-13 | 2019-12-03 | 东南大学 | A kind of vehicle-mounted strain of bridge influences the identification of line feature intelligent and extracting method |
| KR102197696B1 (en) * | 2019-12-24 | 2021-01-04 | 한국건설기술연구원 | Structure health monitoring system using optic fiber-based hybrid nerve network sensor, and method for the same |
| CN113639646A (en) * | 2021-05-24 | 2021-11-12 | 武汉理工大学 | Full-time global online monitoring device and method for deformation of large engineering structure |
| CN113723010A (en) * | 2021-09-08 | 2021-11-30 | 福建中锐网络股份有限公司 | Bridge damage early warning method based on LSTM temperature-displacement correlation model |
| CN113973403A (en) * | 2021-11-10 | 2022-01-25 | 东南大学 | Temperature-induced strain field redistribution intelligent sensing method based on structure discrete measuring point topology |
| CN115311246A (en) * | 2022-08-23 | 2022-11-08 | 保利长大工程有限公司 | Bridge safety management method and system based on big data |
| CN115392360A (en) * | 2022-08-11 | 2022-11-25 | 哈尔滨工业大学 | Transformer-based large bridge temperature-response related pattern recognition and health diagnosis method |
| CN115389066A (en) * | 2022-07-29 | 2022-11-25 | 安徽工业大学 | Bridge health monitoring system based on distributed fiber grating sensing |
-
2023
- 2023-05-08 CN CN202310507996.4A patent/CN116542146A/en active Pending
Patent Citations (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103134432A (en) * | 2013-01-25 | 2013-06-05 | 南京大学 | Bridge displacement monitoring system based on optical fiber sensors |
| CN105675056A (en) * | 2016-02-22 | 2016-06-15 | 东南大学 | Long-span bridge damage recognition method |
| WO2017202139A1 (en) * | 2016-05-26 | 2017-11-30 | 东南大学 | Bridge damage identification method based on long-gauge-length strain influence envelope |
| CN108052770A (en) * | 2017-12-28 | 2018-05-18 | 大连理工大学 | A kind of Loads of Long-span Bridges girder performance method for early warning for considering time-varying effecting |
| CN108399277A (en) * | 2018-01-24 | 2018-08-14 | 华南理工大学 | A kind of bridge damnification recognition method based on temperature with strain relevance |
| CN108763763A (en) * | 2018-05-28 | 2018-11-06 | 东南大学 | A kind of bridge structure strain-responsive abnormity early warning method |
| CN110533007A (en) * | 2019-09-13 | 2019-12-03 | 东南大学 | A kind of vehicle-mounted strain of bridge influences the identification of line feature intelligent and extracting method |
| KR102197696B1 (en) * | 2019-12-24 | 2021-01-04 | 한국건설기술연구원 | Structure health monitoring system using optic fiber-based hybrid nerve network sensor, and method for the same |
| CN113639646A (en) * | 2021-05-24 | 2021-11-12 | 武汉理工大学 | Full-time global online monitoring device and method for deformation of large engineering structure |
| CN113723010A (en) * | 2021-09-08 | 2021-11-30 | 福建中锐网络股份有限公司 | Bridge damage early warning method based on LSTM temperature-displacement correlation model |
| CN113973403A (en) * | 2021-11-10 | 2022-01-25 | 东南大学 | Temperature-induced strain field redistribution intelligent sensing method based on structure discrete measuring point topology |
| CN115389066A (en) * | 2022-07-29 | 2022-11-25 | 安徽工业大学 | Bridge health monitoring system based on distributed fiber grating sensing |
| CN115392360A (en) * | 2022-08-11 | 2022-11-25 | 哈尔滨工业大学 | Transformer-based large bridge temperature-response related pattern recognition and health diagnosis method |
| CN115311246A (en) * | 2022-08-23 | 2022-11-08 | 保利长大工程有限公司 | Bridge safety management method and system based on big data |
Non-Patent Citations (3)
| Title |
|---|
| HANWEI ZHAO ET AL: "Digital modeling on the nonlinear mapping between multi-source monitoring data of in-service bridges", STRUCT CONTROL HEALTH MONIT, pages 1 - 17 * |
| YADI TIAN ET AL: "Relationship modeling between vehicle-induced girder vertical deflection and cable tension by BiLSTM using field monitoring data of a cable-stayed bridge", STRUCT CONTROL HEALTH MONIT, pages 1 - 20 * |
| 第1-18页: "Support Vector Regression for Anomaly Detection from Measurement Histories", ORE OPEN RESEARCH EXETER, pages 1 - 18 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Catbas et al. | Structural health monitoring and reliability estimation: Long span truss bridge application with environmental monitoring data | |
| CN105139585A (en) | Intelligent early warning and prediction method for soil slope danger | |
| CN105956216A (en) | Finite element model correction method for large-span steel bridge based on uniform temperature response monitoring value | |
| CN114005278B (en) | Intelligent monitoring and early warning system and method for highway infrastructure group | |
| CN105760934A (en) | Bridge abnormity monitoring restoration method based on wavelet and BP neural network | |
| Ni et al. | Fast operational modal analysis of a single-tower cable-stayed bridge by a Bayesian method | |
| CN111967185A (en) | Cable-stayed bridge state evaluation method based on cable force and displacement distribution correlation modeling | |
| CN105956218A (en) | Steel bridge finite element model correction method based on non-uniform temperature response monitoring value | |
| Li et al. | Automatic uncoupling of massive dynamic strains induced by vehicle-and temperature-loads for monitoring of operating bridges | |
| Xu et al. | A Bayesian approach for site-specific extreme load prediction of large scale bridges | |
| CN107664489A (en) | A kind of bridge strain and the measuring method of deformation | |
| Yue et al. | Ultra-high precise Stack-LSTM-CNN model of temperature-induced deflection of a cable-stayed bridge for detecting bridge state driven by monitoring data | |
| CN112818444B (en) | Railway concrete bridge linear real-time control method based on operation and driving safety | |
| CN113987925A (en) | Steel aqueduct structure extreme stress early warning method based on sliding window Gaussian process model | |
| Guan et al. | Web‐based structural health monitoring of an FRP composite bridge | |
| WO2024041233A1 (en) | Method for evaluating fatigue damage and life of bridge structure under multi-factor coupling action | |
| CN113408190A (en) | Bayes-LSTM model-based surrounding rock deformation prediction method during construction of highway tunnel | |
| CN116542146A (en) | Bridge monitoring temperature field-strain field space-time correlation model and health diagnosis method | |
| Moon et al. | Application of artificial neural network to predict dynamic displacements from measured strains for a highway bridge under traffic loads | |
| CN113821952B (en) | Digital twin ballastless track optimization method based on Kalman filtering algorithm | |
| Mufti et al. | A case for structural health monitoring (SHM) and civionics enhances the evaluation of the load carrying capacity of aging bridges | |
| Wang et al. | Data interpretation and forecasting of SHM heteroscedastic measurements under typhoon conditions enabled by an enhanced Hierarchical sparse Bayesian Learning model with high robustness | |
| Ren et al. | Estimation of extreme cable forces of cable-stayed bridges based on monitoring data and random vehicle models | |
| Zhao et al. | Cloud monitoring system for assembled beam bridge based on index of dynamic strain correlation coefficient | |
| CN113673145A (en) | Surface subsidence prediction method based on PCA-LSTM neural network |
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 |