CN112528365B - Method for predicting healthy evolution trend of underground infrastructure structure - Google Patents
Method for predicting healthy evolution trend of underground infrastructure structure Download PDFInfo
- Publication number
- CN112528365B CN112528365B CN202011376296.9A CN202011376296A CN112528365B CN 112528365 B CN112528365 B CN 112528365B CN 202011376296 A CN202011376296 A CN 202011376296A CN 112528365 B CN112528365 B CN 112528365B
- Authority
- CN
- China
- Prior art keywords
- data
- date
- time sequence
- monitoring
- model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 38
- 238000012544 monitoring process Methods 0.000 claims abstract description 83
- 230000007613 environmental effect Effects 0.000 claims abstract description 36
- 230000036541 health Effects 0.000 claims abstract description 31
- 238000013528 artificial neural network Methods 0.000 claims abstract description 20
- 230000004927 fusion Effects 0.000 claims abstract description 19
- 230000008859 change Effects 0.000 claims abstract description 18
- 238000012360 testing method Methods 0.000 claims abstract description 11
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims abstract description 10
- 230000004044 response Effects 0.000 claims abstract description 8
- 239000011159 matrix material Substances 0.000 claims description 20
- 238000012549 training Methods 0.000 claims description 18
- 230000000306 recurrent effect Effects 0.000 claims description 12
- 238000012795 verification Methods 0.000 claims description 8
- 230000000694 effects Effects 0.000 claims description 7
- 238000002474 experimental method Methods 0.000 claims description 7
- 230000003542 behavioural effect Effects 0.000 claims description 3
- 230000006870 function Effects 0.000 claims description 3
- 238000011478 gradient descent method Methods 0.000 claims description 2
- 238000010801 machine learning Methods 0.000 description 6
- 230000008569 process Effects 0.000 description 4
- 230000009191 jumping Effects 0.000 description 3
- 125000004122 cyclic group Chemical group 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000012417 linear regression Methods 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 230000005856 abnormality Effects 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 239000011435 rock Substances 0.000 description 1
- 238000012706 support-vector machine Methods 0.000 description 1
- 238000010200 validation analysis Methods 0.000 description 1
- 238000012800 visualization Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/10—Geometric CAD
- G06F30/13—Architectural design, e.g. computer-aided architectural design [CAAD] related to design of buildings, bridges, landscapes, production plants or roads
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01D—MEASURING NOT SPECIALLY ADAPTED FOR A SPECIFIC VARIABLE; ARRANGEMENTS FOR MEASURING TWO OR MORE VARIABLES NOT COVERED IN A SINGLE OTHER SUBCLASS; TARIFF METERING APPARATUS; MEASURING OR TESTING NOT OTHERWISE PROVIDED FOR
- G01D21/00—Measuring or testing not otherwise provided for
- G01D21/02—Measuring two or more variables by means not covered by a single other subclass
-
- 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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
- G06Q10/06393—Score-carding, benchmarking or key performance indicator [KPI] analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/08—Construction
-
- 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
- G06F2119/12—Timing analysis or timing optimisation
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Human Resources & Organizations (AREA)
- Evolutionary Computation (AREA)
- Strategic Management (AREA)
- Economics (AREA)
- Geometry (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Health & Medical Sciences (AREA)
- Computer Hardware Design (AREA)
- General Business, Economics & Management (AREA)
- Tourism & Hospitality (AREA)
- Software Systems (AREA)
- General Health & Medical Sciences (AREA)
- Marketing (AREA)
- Development Economics (AREA)
- Entrepreneurship & Innovation (AREA)
- Operations Research (AREA)
- Computing Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Quality & Reliability (AREA)
- Mathematical Physics (AREA)
- Data Mining & Analysis (AREA)
- Educational Administration (AREA)
- Molecular Biology (AREA)
- Game Theory and Decision Science (AREA)
- Medical Informatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Primary Health Care (AREA)
- Architecture (AREA)
- Civil Engineering (AREA)
- Structural Engineering (AREA)
- Computational Mathematics (AREA)
Abstract
The invention relates to an evolution trend prediction method for the structural health of underground infrastructure, which comprises the following steps: step one, acquiring structural health monitoring data: deploying a sensor array in the underground facility, and acquiring dynamic response data reflecting the internal change of the engineering structure for a long time so as to detect and identify the structural damage and the property change of the underground infrastructure; acquiring external load data, wherein the external load refers to environmental data of the position where the underground infrastructure is located, such as climate and hydrological indexes of air temperature, river water level and the like; combining historical monitoring time sequence data with environmental data to construct a load-time sequence fusion prediction model; the time sequence dependence and the environment dependence are combined through a full-connection neural network, so that the future change trend of the monitoring index is predicted; and step four, predicting the monitoring indexes of the underground infrastructure under various boundary conditions to realize the coverage test of the safety of the underground infrastructure.
Description
Technical Field
The invention belongs to the field of intersection of civil engineering structure health monitoring and machine learning technologies, and relates to an intelligent civil engineering structure health monitoring technology.
Background
With the vigorous development of the economy of China, the travel demand of people is greatly increased, and a large amount of underground infrastructure is built. Underground infrastructure located in complex geological sections inevitably causes stress redistribution and deformation of the surrounding rock and may last months or even years. Therefore, developing a reliable method to predict the evolution trend of the tunnel structure is crucial to maintaining the stability of the tunnel structure. Over the past few years, a number of empirical and semi-empirical models have been proposed to address this problem. However, due to the limitations of many internal and external factors such as complex geological conditions and complex internal structure, the engineering has certain challenges. Therefore, a reasonable model is urgently needed to be established so as to adapt to complex working conditions and carry out high-precision mechanical property prediction on the complex working conditions.
In recent years, with the rapid development of computer science, real-time monitoring of the structure of an underground infrastructure by using a structural health monitoring system has attracted much attention. Various machine learning algorithms are gradually used to analyze monitoring data obtained from the system, and currently used machine learning prediction methods are largely divided into two major categories, that is, prediction methods based on time series and prediction methods based on external loads. The time series-based prediction method is a method for predicting a future structural change trend or a potential risk event by using a time series formed by structural health monitoring data observed in the past, and a time series prediction model based on deep learning represented by a recurrent neural network has been excellent in this work in recent years. The model based on the external load predicts the change trend of the structure according to the influence of external factors on the mechanical change of the structure, and the previous research shows that the mechanical response under the action of the external load is reliable by adopting a multiple linear regression model, so that the model is widely applied to tunnel and bridge engineering.
However, both models have their own disadvantages. The time series based prediction model ignores environmental impact factors and assumes that the observed time series repeats periodically, the effect of dealing with long term time dependence is poor. The external load based prediction model ignores the influence of historical monitoring data. Therefore, there is an urgent need for a new prediction model, which combines the advantages of the two models and overcomes their disadvantages, thereby realizing more accurate prediction.
Disclosure of Invention
The invention solves the problems: the method aims at the problems that the existing method for predicting the evolution trend of the underground infrastructure structure in the field of civil engineering structure health monitoring is low in accuracy and cannot accurately capture the trend. The invention provides a load-time sequence fusion prediction method. The method can improve the accuracy of predicting the structural change trend of the underground infrastructure and the correlation coefficient of the predicted value and the true value, so that the predicted trend is closer to the field situation.
The technical solution of the invention is as follows: aiming at the problems of various geological conditions and complex internal structure of the position of the underground infrastructure, the evolution trend prediction method combines structural health monitoring data in the underground infrastructure with external environmental data to construct a load-time sequence fusion prediction method, and predicts the evolution trend of the structural behavior of the underground infrastructure so as to improve the prediction precision. The prediction method comprises the steps of firstly extracting an original monitoring data matrix from a structural health monitoring system, obtaining environmental data in the same period and forming the matrix. And then extracting corresponding data from the original data matrix and recombining the data and the environmental data into an input matrix aiming at each sensor, processing time sequence data by using a Recurrent Neural Network (RNN) to obtain the representation of the hidden layer state of the data, splicing the hidden state at the last moment and the environmental data at the corresponding moment, and inputting the spliced hidden state and the environmental data into a fully-connected neural network to obtain output which is a predicted value. The prediction method mainly has two applications, namely predicting the structural health monitoring index change trend of the facility in the future time period according to given historical monitoring data and environment forecast data of the place where the underground infrastructure is located, and researching whether the change range of the monitoring index exceeds a given threshold value under different boundary conditions by changing the input value of the environmental factor. In the task of predicting the structural health monitoring indexes of the underground infrastructure, the prediction method can achieve the effects of lower prediction error and higher prediction correlation compared with the traditional prediction method.
The technical scheme of the invention is as follows: an evolution trend prediction method for the structural health of underground infrastructure comprises the following steps:
step one, acquiring structural health monitoring data: deploying a sensor array in an underground facility, and acquiring dynamic response data reflecting the internal change of an engineering structure for a long time, wherein the dynamic response data comprises monitoring indexes: stress, strain, water pressure, temperature; thereby detecting and identifying structural damage and behavioral changes of the underground infrastructure; forming time sequence data by the structural health monitoring data observed in a period of time;
acquiring external load data, wherein the external load refers to environmental data of the position where the underground infrastructure is located, such as climate and hydrological indexes of air temperature, river water level and the like;
combining historical monitoring time sequence data with environmental data to construct a load-time sequence fusion prediction model; the load-time sequence fusion prediction model is characterized in that a cyclic neural network is used for learning a time sequence dependence relationship in historical monitoring data, and the time sequence dependence and the environment dependence are combined through a fully-connected neural network, so that the future change trend of a monitoring index is predicted;
and fourthly, predicting the monitoring indexes of the underground infrastructure under various boundary conditions, wherein the various boundary conditions refer to the combination of various different environmental factors, and cover various possible conditions so as to realize the coverage test of the safety of the underground infrastructure.
Further, in the third step, the implementation of the load-time sequence fusion prediction model includes:
step 1.1: deriving original monitoring data from a structural health monitoring system, organizing the original monitoring data into an mxn matrix X, wherein m is the length of acquired data, n is the number of sensors, if data of a certain sensor on a certain date is missing, replacing the average value of the sensor on an adjacent date with the average value of the sensor on the adjacent date, and organizing the acquired environmental data into an mxk matrix E, wherein k is the number of environmental factors;
step 1.2: selecting two hyper-parameters of the model, namely an observation window, namely the length p of an event sequence used by the model for learning each time and the number q of future days predicted by the model;
step 1.3: selecting a sensor number i to be predicted; taking out the ith column of the matrix X, namely all data of the sensor i, and constructing an input matrix; for any date t, forming a time sequence by monitoring data of p previous dates including the date, extracting environment data of a matrix E at the date t + q, and combining the environment data and the time sequence into an input vector; performing the above operations every date t, and finally combining the dates t into an input matrix X' of (m-p-q) × (p + k);
step 1.4: dividing an input matrix X' into a training set and a verification set, wherein the first 70% of data items are used as the training set, and the last 30% of data items are used as the verification set;
step 1.5: training the model by using data in a training set, learning time sequence dependence in a time sequence by using a Recurrent Neural Network (RNN), outputting a hidden layer representation of a monitoring index by the RNN for each date t, connecting the hidden layer representation with k environmental factors when the date is t + q, and outputting a predicted value of the monitoring index when the date is t + q as input of a full-connection layer;
step 1.6: calculating the error between the predicted value and the true value in the training set by using a loss function, updating each parameter of the model by using a gradient descent method, repeatedly training until the parameters are converged, and ending the training at the moment;
step 1.7: and (3) testing the performance of the model on the verification set, skipping to the step 1.2, selecting other hyper-parameters for experiment until the hyper-parameter combination with the best effect is found out, and storing the parameters of the model.
Further, in the third step, the method for predicting the evolution trend includes:
step 2.1: loading the trained model F, namely loading each weight parameter of the model into a memory;
step 2.2: for the date t, predicting a monitoring index when the date is t + q, loading a time sequence formed by monitoring data from the date t-p +1 to the date t, and then loading an environmental data prediction value of the date t + q to form an input vector;
step 2.3: inputting the input vector into the model F, and outputting a predicted value of the monitoring index when the date is t + q;
step 2.4: and repeating the steps of 2.2 and 2.3 for a plurality of dates t to obtain the structural evolution trend of the underlying infrastructure within a period of time in the future, and drawing a line chart.
Further, in the fourth step, the safety coverage testing method includes:
step 3.1: loading the trained model F, namely loading each weight parameter of the model into a memory;
step 3.2: loading a time sequence formed by monitoring data from a date t-p +1 to a date t for a specific date t;
step 3.3: for each environmental factor, obtaining the possible range according to the maximum value of the years, and combining;
step 3.4: and splicing each possible environment data combination with the time series into an input vector, inputting the input vector into the model F to obtain the monitoring index predicted value under the condition, so that the range of the underground infrastructure monitoring indexes which can possibly appear under all the environment condition combinations can be obtained, and the safety of the underground infrastructure monitoring indexes is tested in a coverage manner.
Compared with the prior art, the invention has the advantages and effects that:
compared with the prior art, the prediction error of the method is reduced, and the future evolution trend of the underground infrastructure can be better learned. In addition, the invention provides a method for performing coverage test on the safety provided by the underground infrastructure on the basis, so that the safety condition of the infrastructure can be more comprehensively evaluated.
Drawings
FIG. 1 is a flow chart of the present invention for constructing a load-timing fusion prediction model;
FIG. 2 is a schematic diagram of a load-timing fusion prediction model constructed based on a recurrent neural network according to the present invention;
fig. 3 is an overall flow chart of the whole system of the invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, rather than all embodiments, and based on the embodiments of the present invention, all other embodiments obtained by a person skilled in the art without creative efforts belong to the protection scope of the present invention.
The method is suitable for predicting the structural health trend of various underground infrastructures such as tunnels, subways, underground buildings and the like, and the data to be predicted are mechanical indexes such as stress and strain commonly used in civil engineering. By combining the structural health monitoring data and the environmental data, the model can learn the time sequence dependence in the monitoring data and the relationship between the monitoring indexes and the environmental factors, and predict the structural evolution trend of the underground infrastructure.
According to one embodiment of the invention, the evolution trend prediction method facing the structural health of the underground infrastructure comprises the following steps:
step one, acquiring structural health monitoring data: deploying a sensor array in an underground facility, and acquiring dynamic response data reflecting the internal change of an engineering structure for a long time, wherein the dynamic response data comprises monitoring indexes: stress, strain, water pressure, temperature; thereby detecting and identifying structural damage and behavioral changes of the underground infrastructure; forming time sequence data by the structural health monitoring data observed in a period of time;
acquiring external load data, wherein the external load refers to environmental data of the position of the underground infrastructure, such as climate and hydrological indexes of air temperature, river water level and the like;
combining historical monitoring time sequence data with environmental data to construct a load-time sequence fusion prediction model; the load-time sequence fusion prediction model is characterized in that a cyclic neural network is used for learning a time sequence dependence relationship in historical monitoring data, and the time sequence dependence and the environment dependence are combined through a fully-connected neural network, so that the future change trend of a monitoring index is predicted;
and fourthly, predicting the monitoring indexes of the underground infrastructure under various boundary conditions, wherein the various boundary conditions refer to the combination of various different environmental factors, and cover various possible conditions so as to realize the coverage test of the safety of the underground infrastructure.
Fig. 1 shows a process for constructing a load-time series fusion prediction model according to an embodiment of the present invention: the sensors arranged in the underground infrastructure are organized together according to the structural health monitoring data collected by the day period and can be regarded as a time sequence, so that in order to predict the evolution trend of the monitoring index in a future period, some prediction models can be used for learning past historical data to obtain the time sequence dependence between the monitoring data. However, the underground infrastructure is located in a region which often comprises a complex climate and hydrological environment, and the effect of environmental factors on the structural evolution trend is not negligible, so that time sequence data and environmental data can be fused to construct a prediction model, and the prediction accuracy is improved. The steps of the process are as follows:
step 3, training the fusion model by using the data in the training set, and updating the parameters of the model;
step 4, after the training is finished, verifying the effect of the model by using a verification set, and selecting a proper hyper-parameter according to the performance of the model;
and 5, after the optimal model is obtained, storing the parameters of the model.
Fig. 2 shows the principle of the load-time sequence fusion prediction model constructed based on the recurrent neural network: the recurrent neural network has excellent performance in processing time series data, capturing the time series dependence of the monitoring data, and thus it is used to process historical monitoring data. The operation flow of the load-time sequence fusion prediction model constructed based on the recurrent neural network can be represented as the following steps:
step 3, splicing the environmental data at the time t + q and the hidden state at the time t into a vector, inputting the vector into the fully-connected neural network, and outputting the predicted value of the structural health monitoring data at the time t + q;
step 4, comparing the predicted value with the true value, calculating a prediction error by using a loss function, and jumping to step 6 if the error is converged, or jumping to step 5;
step 5, updating the parameters of the model by using a gradient descent algorithm, and jumping to the step 2;
and 6, saving the parameters of the model to the local, and exiting the training program.
Fig. 3 shows the application of the load-time series fusion prediction model proposed herein in engineering: after the model is trained and adjusted on the validation set to the best parameters, the parameters of the model will be saved for subsequent application. The first application is to predict the structural performance of the underground infrastructure in a future period of time under normal conditions, wherein recently collected monitoring data are input as a new time sequence, air temperature and water level forecast data at a future date are input as new environmental data, the new environmental data and the air temperature and water level forecast data are combined to form a new input vector, and the new input vector is input into a trained model to obtain a predicted value of a future structural health monitoring index. The second application is that the safety of the underground infrastructure is tested in a coverage mode by changing the input of the environment data, the change range of historical air temperature and water level can be obtained by searching the environment data of the past year, all possible environment factor combinations can be obtained by combining the environment data and the change range, each combination is combined with the time sequence formed by the monitoring data and input into the prediction model, the future structural performance of the underground infrastructure under various combinations can be obtained, whether the given threshold value is exceeded or not is observed, and therefore whether structural abnormality occurs under the extreme condition or not is judged.
The method is implemented based on computer science and various machine learning algorithms and needs a certain programming, machine learning and deep learning basis, and is realized based on Python programming language and an open-source machine learning library PyTorch. In order to verify the load-time sequence fusion prediction model based on the recurrent neural network shown in fig. 1 and 2, an experiment was performed using stress data of ten sensors in a structural health monitoring system installed in a certain tunnel. The sensors are distributed at different positions of the arch crown, arch waist and arch foot of ten sections of the tunnel. In order to evaluate the prediction capability of the model, two evaluation indexes, namely Root Mean Square Error (RMSE) and Pearson Correlation Coefficient (PCC), are used, the RMSE is used for measuring the error between the predicted value and the true value of the monitoring index, and the PCC is mainly used for measuring the predicted future structure trend and the fitting degree of the true trend. To find the most suitable observation window, i.e. the hyper-parameter p in the model, experiments were performed for both 5. Ltoreq. P.ltoreq.20, resulting in p =14 being optimal. In order to verify that the performance of the method is better than that of other models, other commonly used prediction models such as linear regression, support vector machines and multilayer perceptrons are used for carrying out comparison experiments, and the experiments prove that the prediction error of the method is the lowest and the prediction correlation is the highest. In order to verify that the method can predict the future structure evolution trend of the underground infrastructure under different time spans, experiments are carried out on prediction intervals of 1-14 days, and the experimental results show that the prediction error of the method is increased along with the increase of the prediction span q, but the prediction error within 14 days is maintained in a smaller range.
In the overall process of the method of the present invention shown in fig. 3, in the application scenario, some visualization means are used to visually display the prediction result, and the charts are drawn by using matplotlib library. In applications for predicting future structural evolution trends of underground infrastructure, a line graph is used to show the change in monitoring metrics for the next two weeks. In the application of coverage testing on the safety of the underground infrastructure, thermodynamic diagrams are used for showing the distribution of monitoring index prediction values of the underground infrastructure under different environmental factor combinations.
Although the illustrative embodiments of the present invention have been described in order to facilitate those skilled in the art to understand the invention, it is to be understood that the invention is not limited in scope to the specific embodiments, but rather, it will be apparent to those skilled in the art that various changes may be made without departing from the spirit and scope of the invention as defined and defined in the appended claims, and it is intended that all matter contained in the invention and created by the inventive concept be protected.
Claims (3)
1. An evolution trend prediction method for the structural health of underground infrastructure is characterized by comprising the following steps:
step one, acquiring structural health monitoring data: deploying a sensor array in an underground facility, and acquiring dynamic response data reflecting the internal change of an engineering structure for a long time, wherein the dynamic response data comprises monitoring indexes: stress, strain, water pressure, temperature; thereby detecting and identifying structural damage and behavioral changes of the underground infrastructure; forming time sequence data by the structural health monitoring data observed in a preset length of time;
acquiring external load data, wherein the external load refers to environmental data of the position of the underground infrastructure, and comprises climate and hydrological indexes including air temperature and river water level;
combining historical monitoring time sequence data with environmental data to construct a load-time sequence fusion prediction model; the load-time sequence fusion prediction model is used for learning a time sequence dependency relationship in historical monitoring data by using a recurrent neural network, and the time sequence dependency and the environment dependency are combined by using a full-connection neural network, so that the future change trend of the monitoring index is predicted;
fourthly, forecasting the monitoring indexes of the underground infrastructure under various boundary conditions, wherein the various boundary conditions refer to the combination of various different environmental factors and cover various possible conditions so as to realize the coverage test of the safety of the underground infrastructure;
the third step, the implementation of the load-time sequence fusion prediction model comprises the following steps:
step 1.1: deriving original monitoring data from a structural health monitoring system, organizing the original monitoring data into an mxn matrix X, wherein m is the length of acquired data, n is the number of sensors, if data of a certain sensor on a certain date is missing, replacing the average value of the sensor on an adjacent date with the average value of the sensor on the adjacent date, and organizing the acquired environmental data into an mxk matrix E, wherein k is the number of environmental factors;
step 1.2: selecting two hyper-parameters of the model, namely an observation window, namely the length p of a time sequence used by the model for learning each time and the number q of future days predicted by the model;
step 1.3: selecting a sensor number i to be predicted; taking out the ith column of the matrix X, namely all data of the sensor i, and constructing an input matrix; for any date t, forming a time sequence by monitoring data of p previous dates including the date, extracting environment data of a matrix E at the date t + q, and combining the environment data and the time sequence into an input vector; performing the above operations every date t, and finally combining the dates t into an input matrix X' of (m-p-q) × (p + k);
step 1.4: dividing an input matrix X' into a training set and a verification set, wherein the first 70% of data items are used as the training set, and the last 30% of data items are used as the verification set;
step 1.5: training the model by using data in the training set, learning time sequence dependence in a time sequence by using a Recurrent Neural Network (RNN), outputting a hidden layer representation of a monitoring index by the RNN for each date t, connecting the hidden layer representation with k environmental factors when the date is t + q, and outputting a predicted value of the monitoring index when the date is t + q as the input of a full connection layer;
step 1.6: calculating the error between the predicted value and the true value in the training set by using a loss function, updating each parameter of the model by using a gradient descent method, repeatedly training until the parameters are converged, and ending the training at the moment;
step 1.7: and (3) testing the performance of the model on the verification set, skipping to the step 1.2, selecting other hyper-parameters for experiment until the hyper-parameter combination with the best effect is found out, and storing the parameters of the model.
2. The method for predicting the evolution trend oriented to the structural health of the underground infrastructure as claimed in claim 1, wherein the third step is implemented by the steps of:
step 2.1: loading the trained model F, namely loading each weight parameter of the model into a memory;
step 2.2: for the date t, predicting a monitoring index when the date is t + q, loading a time sequence formed by monitoring data from the date t-p +1 to the date t, and then loading an environmental data prediction value of the date t + q to form an input vector;
step 2.3: inputting the input vector into the model F, and outputting the predicted value of the monitoring index when the date is t + q;
step 2.4: and repeating the steps of 2.2 and 2.3 for a plurality of dates t to obtain the structural evolution trend of the underlying infrastructure within a period of time in the future, and drawing a line chart.
3. The method for predicting the evolution trend of the underground infrastructure structural health, as set forth in claim 1, wherein the step four, the safety coverage testing method is implemented by the steps of:
step 3.1: loading the trained model F, namely loading each weight parameter of the model into a memory;
step 3.2: loading a time sequence formed by monitoring data from a date t-p +1 to a date t for a specific date t;
step 3.3: for each environmental factor, obtaining the possible range according to the maximum value of the years, and combining;
step 3.4: and splicing each possible environment data combination with the time series into an input vector, inputting the input vector into the model F to obtain a monitoring index predicted value under the condition, and thus obtaining the possible occurrence range of the monitoring index of the underground infrastructure under all the environment condition combinations, and thereby testing the safety of the underground infrastructure.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011376296.9A CN112528365B (en) | 2020-11-30 | 2020-11-30 | Method for predicting healthy evolution trend of underground infrastructure structure |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011376296.9A CN112528365B (en) | 2020-11-30 | 2020-11-30 | Method for predicting healthy evolution trend of underground infrastructure structure |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112528365A CN112528365A (en) | 2021-03-19 |
| CN112528365B true CN112528365B (en) | 2022-10-14 |
Family
ID=74995381
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011376296.9A Active CN112528365B (en) | 2020-11-30 | 2020-11-30 | Method for predicting healthy evolution trend of underground infrastructure structure |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112528365B (en) |
Families Citing this family (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113377744B (en) * | 2021-06-16 | 2023-07-18 | 北京建筑大学 | Reconstruction method and device for structural anomaly monitoring data with environmental temperature correlation |
| CN115098561B (en) * | 2022-06-28 | 2024-03-26 | 浙江大学 | Method for extracting alarm threshold of cultural relic ontology monitoring data |
| CN115638831B (en) * | 2022-12-21 | 2023-04-25 | 四川九通智路科技有限公司 | Highway facility risk monitoring method and system based on MEMS sensor |
| CN116150680B (en) * | 2023-04-18 | 2023-07-21 | 中铁第四勘察设计院集团有限公司 | Underground transportation junction damage state identification method and system based on big data |
| CN116913416B (en) * | 2023-09-12 | 2023-12-26 | 中国地质大学（北京） | Quantitative evaluation and prediction method and device for weathering degree of sea rock |
| CN117196893B (en) * | 2023-11-06 | 2024-01-05 | 中交第四航务工程勘察设计院有限公司 | Infrastructure foundation pit real-time monitoring and early warning management method and system |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108168926A (en) * | 2018-01-22 | 2018-06-15 | 浙江大学城市学院 | A kind of environmental suitability test device of structural member and health monitoring equipment |
| CN110070172A (en) * | 2019-03-13 | 2019-07-30 | 西安理工大学 | The method for building up of sequential forecasting models based on two-way independent loops neural network |
| CN110502820A (en) * | 2019-08-14 | 2019-11-26 | 浙江精工钢结构集团有限公司 | A kind of structural steelwork Monitoring and forecasting system in real-time method based on BIM |
| CN111258297A (en) * | 2020-01-17 | 2020-06-09 | 北京大学 | Equipment health index construction and service life prediction method based on data fusion network |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8521443B2 (en) * | 2008-10-16 | 2013-08-27 | Oxfordian | Method to extract parameters from in-situ monitored signals for prognostics |
| CN111242292B (en) * | 2020-04-26 | 2020-09-08 | 北京全路通信信号研究设计院集团有限公司 | OD data prediction method and system based on deep space-time network |
-
2020
- 2020-11-30 CN CN202011376296.9A patent/CN112528365B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108168926A (en) * | 2018-01-22 | 2018-06-15 | 浙江大学城市学院 | A kind of environmental suitability test device of structural member and health monitoring equipment |
| CN110070172A (en) * | 2019-03-13 | 2019-07-30 | 西安理工大学 | The method for building up of sequential forecasting models based on two-way independent loops neural network |
| CN110502820A (en) * | 2019-08-14 | 2019-11-26 | 浙江精工钢结构集团有限公司 | A kind of structural steelwork Monitoring and forecasting system in real-time method based on BIM |
| CN111258297A (en) * | 2020-01-17 | 2020-06-09 | 北京大学 | Equipment health index construction and service life prediction method based on data fusion network |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112528365A (en) | 2021-03-19 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112528365B (en) | Method for predicting healthy evolution trend of underground infrastructure structure | |
| Avci et al. | A review of vibration-based damage detection in civil structures: From traditional methods to Machine Learning and Deep Learning applications | |
| EP3809220B1 (en) | Method and system for semi-supervised deep anomaly detection for large-scale industrial monitoring systems based on time-series data utilizing digital twin simulation data | |
| Lei et al. | Mutual information based anomaly detection of monitoring data with attention mechanism and residual learning | |
| Malekzadeh et al. | Performance-based structural health monitoring through an innovative hybrid data interpretation framework | |
| Niyirora et al. | Intelligent damage diagnosis in bridges using vibration-based monitoring approaches and machine learning: A systematic review | |
| CN111178758A (en) | Concrete dam monitoring data intelligent management and real-time evaluation system based on BIM | |
| 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 | |
| CN114357594A (en) | Bridge abnormity monitoring method, system, equipment and storage medium based on SCA-GRU | |
| CN116341272A (en) | Construction safety risk management and control system for digital distribution network engineering | |
| Wang et al. | Combined digital twin and hierarchical deep learning approach for intelligent damage identification in cable dome structure | |
| CN116861694A (en) | Cable temperature prediction method based on artificial intelligence | |
| Giglioni et al. | Deep autoencoders for unsupervised damage detection with application to the Z24 benchmark bridge | |
| CN117035739A (en) | Work order path planning method and device for photovoltaic power station, electronic equipment and medium | |
| CN111026790A (en) | Structure safety assessment and forecasting method based on data mining | |
| Li et al. | Structural health monitoring data anomaly detection by transformer enhanced densely connected neural networks | |
| D'Urso et al. | Predictive maintenance of standalone steel industrial components powered by a dynamic reliability digital twin model with artificial intelligence | |
| Song et al. | A novel outlier detection method of long-term dam monitoring data based on SSA-NAR | |
| CN113516179B (en) | Method and system for identifying water leakage performance of underground infrastructure | |
| Rao | A machine learning-based damage prediction techniques for structural health monitoring | |
| Muin et al. | Human-machine collaboration framework for bridge health monitoring | |
| Chinforoush et al. | A novel method for forecasting surface wind speed using wind-direction based on hierarchical markov model | |
| Raphael et al. | Sensor data interpretation in bridge monitoring—A case study | |
| Kumar et al. | Multiple layer radial basis neural network with remora regression tree optimum feature extraction for structural health monitoring | |
| Tran | A new framework for damage detection of steel frames using burg autoregressive and stacked autoencoder-based deep 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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |