CN117345342A - Intelligent monitoring system and early warning method for tunnel water inflow - Google Patents
Intelligent monitoring system and early warning method for tunnel water inflow Download PDFInfo
- Publication number
- CN117345342A CN117345342A CN202311417662.4A CN202311417662A CN117345342A CN 117345342 A CN117345342 A CN 117345342A CN 202311417662 A CN202311417662 A CN 202311417662A CN 117345342 A CN117345342 A CN 117345342A
- Authority
- CN
- China
- Prior art keywords
- tunnel
- early warning
- monitoring
- data
- sensor
- 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
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 title claims abstract description 132
- 238000012544 monitoring process Methods 0.000 title claims abstract description 98
- 238000000034 method Methods 0.000 title claims abstract description 41
- 238000010276 construction Methods 0.000 claims abstract description 30
- 239000011435 rock Substances 0.000 claims abstract description 22
- 238000004458 analytical method Methods 0.000 claims abstract description 15
- 238000010801 machine learning Methods 0.000 claims abstract description 15
- 238000013500 data storage Methods 0.000 claims abstract description 8
- 239000011159 matrix material Substances 0.000 claims description 19
- 238000012549 training Methods 0.000 claims description 18
- 239000003673 groundwater Substances 0.000 claims description 17
- 230000007613 environmental effect Effects 0.000 claims description 10
- 230000003204 osmotic effect Effects 0.000 claims description 9
- 238000007781 pre-processing Methods 0.000 claims description 8
- 238000009412 basement excavation Methods 0.000 claims description 3
- 238000004364 calculation method Methods 0.000 claims description 3
- 150000001875 compounds Chemical class 0.000 claims description 3
- 238000003066 decision tree Methods 0.000 description 6
- 238000005457 optimization Methods 0.000 description 6
- 238000013528 artificial neural network Methods 0.000 description 3
- 238000012545 processing Methods 0.000 description 3
- 238000012795 verification Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000011478 gradient descent method Methods 0.000 description 1
- 238000009434 installation Methods 0.000 description 1
- 238000011835 investigation Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000004377 microelectronic Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 230000008054 signal transmission Effects 0.000 description 1
Classifications
-
- E—FIXED CONSTRUCTIONS
- E21—EARTH OR ROCK DRILLING; MINING
- E21F—SAFETY DEVICES, TRANSPORT, FILLING-UP, RESCUE, VENTILATION, OR DRAINING IN OR OF MINES OR TUNNELS
- E21F17/00—Methods or devices for use in mines or tunnels, not covered elsewhere
- E21F17/18—Special adaptations of signalling or alarm devices
-
- E—FIXED CONSTRUCTIONS
- E21—EARTH OR ROCK DRILLING; MINING
- E21F—SAFETY DEVICES, TRANSPORT, FILLING-UP, RESCUE, VENTILATION, OR DRAINING IN OR OF MINES OR TUNNELS
- E21F16/00—Drainage
- E21F16/02—Drainage of tunnels
-
- E—FIXED CONSTRUCTIONS
- E21—EARTH OR ROCK DRILLING; MINING
- E21F—SAFETY DEVICES, TRANSPORT, FILLING-UP, RESCUE, VENTILATION, OR DRAINING IN OR OF MINES OR TUNNELS
- E21F17/00—Methods or devices for use in mines or tunnels, not covered elsewhere
-
- 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
- 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
- G06Q50/02—Agriculture; Fishing; Forestry; Mining
Landscapes
- Engineering & Computer Science (AREA)
- Mining & Mineral Resources (AREA)
- Life Sciences & Earth Sciences (AREA)
- Physics & Mathematics (AREA)
- General Life Sciences & Earth Sciences (AREA)
- Geochemistry & Mineralogy (AREA)
- Geology (AREA)
- Theoretical Computer Science (AREA)
- Business, Economics & Management (AREA)
- Evolutionary Computation (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Geometry (AREA)
- Software Systems (AREA)
- Medical Informatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Agronomy & Crop Science (AREA)
- Animal Husbandry (AREA)
- Marine Sciences & Fisheries (AREA)
- Computer Hardware Design (AREA)
- Health & Medical Sciences (AREA)
- Economics (AREA)
- General Health & Medical Sciences (AREA)
- Human Resources & Organizations (AREA)
- Marketing (AREA)
- Primary Health Care (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Testing Or Calibration Of Command Recording Devices (AREA)
Abstract
The invention provides an intelligent monitoring system and an early warning method for tunnel water inflow, comprising a multi-source real-time monitoring unit and an early warning unit, wherein the multi-source real-time monitoring unit comprises a humidity sensor, a flow sensor, a water level sensor, a stress sensor, an underground water pressure sensor and a rainfall sensor; the early warning unit comprises a data monitoring module, a data storage module, a prediction analysis module and an early warning module; the humidity sensor is used for monitoring humidity of different mileage of the tunnel; the flow sensor is used for monitoring the flow change of water burst in the tunnel; the water level sensor is used for monitoring the change of the underground water level in the tunnel; the stress sensor is used for monitoring the stress change of surrounding rock in the tunnel; the underground water pressure sensor is used for monitoring the underground water level change in the tunnel; the rainfall sensor is used for monitoring the change of the water inflow in the tunnel. And predicting the change trend of the water inflow data in the tunnel through a machine learning algorithm, and carrying out grading early warning to ensure the orderly normal operation of tunnel construction.
Description
Technical Field
The invention relates to the technical field of tunnel engineering, in particular to an intelligent monitoring system and an early warning method for tunnel water inflow.
Background
Along with the expansion of the construction range of railways and highways, the number of tunnels under bad geological conditions is gradually increased, such as fault fracture zones, karst and the like. The construction and operation of these tunnels present a high safety risk, since the influence of groundwater may cause problems such as gushing water and blockage of drainage pipelines, in order to reduce the safety risk, it is necessary to monitor the groundwater discharge of the tunnels in real time and analyze the trend of the groundwater quantity to identify possible risks. However, how to realize real-time monitoring of the underground water volume of a tunnel is a difficult problem, and needs to be solved by tunnel technicians.
In the tunnel construction process, the tunnel water burst not only can delay the construction process, but also can increase the construction cost, meanwhile, the stability and the safety of the tunnel can be influenced, the condition can lead to the decrease of the underground water level, and then the adjacent ecological environment is negatively influenced, the most serious condition is that the tunnel water burst possibly damages the normal operation environment of the tunnel, so that the tunnel cannot be normally used, the safety risk of a construction site is also continuously changed due to the continuous change of the water burst amount in the tunnel construction process, and constructors need to continuously take safety measures to ensure the safety of the constructors.
At present, the tunnel water inflow is monitored manually to carry out manual early warning, and the manual monitoring has the following problems: manual monitoring is time-consuming and labor-consuming; the emergency can not be dealt with in time, and the monitoring data is scattered; the data obtained by the monitoring personnel has hysteresis and the real-time condition of the construction tunnel can not be mastered in time.
Disclosure of Invention
The invention provides an intelligent monitoring system and an early warning method for the water inflow of a tunnel aiming at the technical problems, which are used for monitoring the water inflow of tunnel construction in real time for a long time, predicting the change trend of water inflow data in the tunnel through a machine learning algorithm, and carrying out hierarchical early warning so as to ensure the orderly normal operation of tunnel construction.
The technical scheme adopted for solving the technical problems is as follows:
the intelligent monitoring system for the tunnel water inflow comprises a multi-source real-time monitoring unit and an early warning unit, wherein the multi-source real-time monitoring unit comprises a humidity sensor, a flow sensor, a water level sensor, a stress sensor, an underground water pressure sensor and a rainfall sensor; the early warning unit comprises a data monitoring module, a data storage module, a prediction analysis module and an early warning module;
the humidity sensor is used for monitoring humidity of different mileage of the tunnel;
the flow sensor is used for monitoring the flow change of water burst in the tunnel;
the water level sensor is used for monitoring the change of the underground water level in the tunnel;
the stress sensor is used for monitoring the stress change of surrounding rock in the tunnel;
the underground water pressure sensor is used for monitoring the underground water level change in the tunnel;
the rainfall sensor is used for monitoring the change of the water inflow in the tunnel;
the data monitoring module is used for monitoring that the sensor transmits signals to the base station of the hole in a wireless way through an antenna, and the base station uploads information to the data storage module of the computer end;
the prediction analysis module is used for preprocessing sample data through a CRITIC algorithm; and carrying the pretreated tunnel water inflow variable data and geological environment information codes into an LSTM model and a LightGBM model, training a prediction model, weighting output values of the two models, and finally obtaining a predicted value of the tunnel environment time sequence data at the next moment.
The early warning module carries out grading early warning aiming at data of different water inflow, and the grading early warning comprises blue early warning, orange early warning, yellow early warning and red early warning.
The CRITIC algorithm comprises the following steps:
the j index of the i object of S1, u samples and v indexes takes x as the value ij Form an original sample data matrix x= (X) ij ) uxv ；
S2, normalizing each index value in the X through a Z-score method to obtain a normalized matrix as follows:
in the formula (I), the components of the compound,sum s j The mean value and the standard deviation of the j index are respectively;
S3、sum s j The calculation formula of (2) is as follows:
s4, calculating a variation system v of each index j The following are provided:
in the formula (II), v j A coefficient of variation for the j-th index;
s5, calculating a standardized matrix X * The correlation coefficient of each index is obtained as follows:
R＝(r kl ) v×v (k=1, 2, …, v; l=1, 2, …, v) (three)
In the formula (III), r kl Is the correlation coefficient between the K and the l index;
s6, calculating coefficients reflecting the degree of independence of the indexes:
and finally, calculating the weight of each index:
obtaining a final sample matrix X f ：
The pretreated tunnel water burst variable data and geological environment information codes are brought into an LSTM model and a LightGBM model, and the prediction model is trained as follows:
h1, analyzing geological environment characteristics of tunnel face excavation, and performing coding operation on geological environment information;
h2, encoding the preprocessed feature matrix variable data and geological environment information, converting the feature matrix variable data and geological environment information into supervised learning data through a sliding window algorithm, setting an early warning time step length to be 30 minutes according to the actual construction condition of the site, and respectively predicting tunnel water inflow data at the time t+1 (namely after 30 minutes) through data at the time t;
h3, taking the surrounding rock grade, geological structure and whether water burst of the tunnel face occurs or not where the tunnel face is currently excavated as constraint conditions, and performing coding operation so that the early warning model converts geological environment information into data so as to facilitate analysis of the prediction model;
and H4, respectively solving errors of the LightGBM model and the LSTM model, and carrying out weighted combination on the prediction results of the two models by using an error reciprocal method to finally obtain the prediction value of the tunnel environment time sequence data at the next moment, wherein the weighted formula is as follows:
f＝m 1 n 1 +m 2 n 2
in the formula (IV), n 1 As LSTMPredicted value, n 2 Is the predicted value of the LightGBM model, a 1 Is the error of LSTM model, a 2 Error of the LightGBM model, m 1 And m 2 The weight coefficients of the LSTM model and the LightGBM model are respectively;
h5, calculating the water inflow change rate as follows:
in the formula (five), X True sense Is the true value of the tunnel water inflow at the moment t,the predicted value of the tunnel water inflow at the time t+1;
and H6, carrying out hierarchical early warning according to a threshold value set in a construction environment, wherein the method comprises the following steps:
when v t ＜v min When the method is used for blue early warning, early warning is not carried out, and workers normally construct the method;
when v min ＜v t ＜v med When the building is in orange early warning, workers normally construct the building and are ready to evacuate at any time;
when v med ＜v t ＜v max When the system is yellow, early warning is carried out, the personnel in the face are immediately evacuated, and the subsequent personnel are evacuated orderly;
when v max ＜v t And when the system is in red early warning, all constructors immediately withdraw and immediately take emergency measures.
An intelligent monitoring and early warning method for tunnel water inflow, which utilizes the intelligent monitoring system for tunnel water inflow as claimed in any one of claims 1-3, and comprises the following steps:
p1, preprocessing relevant water burst data of a tunnel based on a CRITIC algorithm;
p2, carrying the pretreated tunnel water inflow variable data and geological environment information codes into a machine learning algorithm, and training a prediction model;
p3, weighting the output values of the two models to finally obtain a predicted value of the tunnel environment time sequence data at the next moment
And P4, carrying out hierarchical early warning based on a prediction result, setting four early warning levels aiming at water inflow data in different ranges, carrying out blue early warning, orange early warning, yellow early warning and red early warning, setting different early warning values for different early warning levels, and guiding site construction.
In step P1, the tunnel related water gushing data includes the environmental humidity in the tunnel, the water gushing flow, the groundwater level, the surrounding rock stress, the groundwater osmotic pressure, and the rainfall of the ground and the surrounding area of the tunnel.
The beneficial effects of the invention are as follows:
1. the advantage of processing the sequence data through a machine learning method, the historical data of the factors related to the water burst of the tunnel and the geological environment information are taken as training data, the water burst of the working face of the tunnel is predicted accurately and stably, the construction tunnel is monitored in real time through the intelligent early warning system of the water burst tunnel, early warning is timely sent, and operators are evacuated, so that the accident rate is reduced, and the safety of the operators is guaranteed.
2. The method can be suitable for harsh construction environments, and all-weather multi-source monitoring of the tunnel is realized through multi-source monitoring of environmental humidity, groundwater level, surrounding rock stress, groundwater osmotic pressure, and rainfall of the ground and surrounding areas of the tunnel.
Drawings
FIG. 1 is a flow chart of intelligent early warning of tunnel water inflow in the invention;
FIG. 2 is a schematic diagram of an early warning unit according to the present invention;
FIG. 3 is a geological environment information encoding;
FIG. 4 is a schematic diagram of a machine learning model;
fig. 5 is a flow chart of the tunnel water inflow intelligent monitoring and early warning method in the invention.
Detailed Description
The invention will be further described with reference to the drawings and examples.
Example 1
As shown in fig. 1, the intelligent monitoring system for tunnel water inflow comprises a multi-source real-time monitoring unit and an early warning unit, wherein the multi-source real-time monitoring unit comprises a humidity sensor, a flow sensor, a water level sensor, a stress sensor, an underground water pressure sensor and a rainfall sensor; as shown in fig. 2, the early warning unit includes a data monitoring module, a data storage module, a prediction analysis module and an early warning module;
the humidity sensor is used for monitoring humidity of different mileage of the tunnel;
the flow sensor is used for monitoring the flow change of water burst in the tunnel;
the water level sensor is used for monitoring the change of the underground water level in the tunnel;
the stress sensor is used for monitoring the stress change of surrounding rock in the tunnel;
the underground water pressure sensor is used for monitoring the underground water level change in the tunnel;
the rainfall sensor is used for monitoring the change of the water inflow in the tunnel;
the device comprises an environmental humidity monitoring instrument, an underground water level monitoring instrument, a surrounding rock stress monitoring instrument, an underground water osmotic pressure monitoring instrument, a water inflow monitoring instrument and a rainfall monitoring instrument, wherein the instrument is respectively arranged around a tunnel and near a tunnel face construction surface;
a wireless base station is installed at a tunnel portal, the data monitoring module is used for monitoring the wireless transmission of signals to the tunnel portal base station through an antenna by the sensor, and the base station uploads information to a data storage module at a computer end;
the prediction analysis module is used for preprocessing sample data through a CRITIC algorithm; and carrying the pretreated tunnel water inflow variable data and geological environment information codes into an LSTM model and a LightGBM model, training a prediction model, weighting output values of the two models, and finally obtaining a predicted value of the tunnel environment time sequence data at the next moment.
The early warning module carries out grading early warning aiming at data of different water inflow, and the grading early warning comprises blue early warning, orange early warning, yellow early warning and red early warning.
The tunnel water burst variable data comprises: the environmental humidity, the water inflow, the ground water level, the surrounding rock stress, the ground water osmotic pressure and the rainfall of the ground and surrounding areas of the tunnel in the tunnel;
collecting the tunnel water burst factors, and preprocessing tunnel water burst data by utilizing an improved CRITIC algorithm;
the procedure for improving CRITIC algorithm is as follows:
the j index of the i object of S1, u samples and v indexes takes x as the value ij Form an original sample data matrix x= (X) ij ) uxv ；
S2, normalizing each index value in the X through a Z-score method to obtain a normalized matrix as follows:
in the formula (I), the components of the compound,sum s j The mean value and the standard deviation of the j index are respectively;
S3、sum s j The calculation formula of (2) is as follows:
s4, calculating a variation system v of each index j The following are provided:
in the formula (II), v j A coefficient of variation for the j-th index;
s5, calculating a standardized matrix X * The correlation coefficient of each index is obtained as follows:
R＝(r kl ) v×v (k=1, 2, …, v; l=1, 2, …, v) (three)
In the formula (III), r kl Is the correlation coefficient between the K and the l index;
s6, calculating coefficients reflecting the degree of independence of the indexes:
and finally, calculating the weight of each index:
obtaining a final sample matrix X f ：
The pretreated tunnel water burst variable data and geological environment information codes are brought into an LSTM model and a LightGBM model, and the prediction model is trained as follows:
h1, analyzing geological environment characteristics of tunnel face excavation, and performing coding operation on geological environment information;
the surrounding rock grade, geological structure and whether water burst occurs on the tunnel face of the tunnel currently excavated are taken as constraint conditions, coding operation is carried out, so that the early warning model converts geological environment information into data so as to predict the model for analysis, the coding information is shown in figure 3, xi is the rock property of the tunnel stratum, the geological structure and whether water burst occurs on the tunnel face, the processed feature matrix variable data and geological information (x 1, x2, x3 … xi) are respectively brought into a LSTM, lightGBM model, the relation between the data is excavated and learned through a neural network and a machine learning algorithm, and the machine learning model is built through python to continuously bring real-time monitoring data into model training. Carrying the environment data into a trained model for predictive analysis;
the machine learning algorithm is as follows: as shown in fig. 4, a multi-layer LSTM network is combined with a LightGBM machine learning algorithm to learn a trend model of variation among different time series data;
h2, encoding the preprocessed feature matrix variable data and geological environment information, converting the feature matrix variable data and geological environment information into supervised learning data through a sliding window algorithm, setting an early warning time step length to be 30 minutes according to the actual construction condition of the site, and respectively predicting tunnel water inflow data at the time t+1 (namely after 30 minutes) through data at the time t;
the LSTM model is built through a neural network learning library keras in python;
the LightGBM model is built through a machine learning library in python;
h3, taking the surrounding rock grade, geological structure and whether water burst of the tunnel face occurs or not where the tunnel face is currently excavated as constraint conditions, and performing coding operation so that the early warning model converts geological environment information into data so as to facilitate analysis of the prediction model;
and H4, respectively solving errors of the LightGBM model and the LSTM model, and carrying out weighted combination on the prediction results of the two models by using an error reciprocal method to finally obtain the prediction value of the tunnel environment time sequence data at the next moment, wherein the weighted formula is as follows:
f＝m 1 n 1 +m 2 n 2
in the formula (IV), n 1 Is the predicted value of LSTM, n 2 Is the predicted value of the LightGBM model, a 1 Is the error of LSTM model, a 2 Error of the LightGBM model, m 1 And m 2 The weight coefficients of the LSTM model and the LightGBM model are respectively;
h5, calculating the water inflow change rate as follows:
in the formula (five), X True sense Is the true value of the tunnel water inflow at the moment t,the predicted value of the tunnel water inflow at the time t+1;
and H6, carrying out hierarchical early warning according to a threshold value set in a construction environment, wherein the method comprises the following steps:
when v t ＜v min When the method is used for blue early warning, early warning is not carried out, and workers normally construct the method;
when v min ＜v t ＜v med When the building is in orange early warning, workers normally construct the building and are ready to evacuate at any time;
when v med ＜v t ＜v max When the system is yellow, early warning is carried out, the personnel in the face are immediately evacuated, and the subsequent personnel are evacuated orderly;
when v max ＜v t And when the system is in red early warning, all constructors immediately withdraw and immediately take emergency measures.
As shown in FIG. 5, the intelligent monitoring and early warning method for the tunnel water inflow comprises the following steps:
p1, installation monitoring module includes: the device comprises an environmental humidity monitoring instrument, an underground water level monitoring instrument, a surrounding rock stress monitoring instrument, an underground water osmotic pressure monitoring instrument, a water inflow monitoring instrument and a rainfall monitoring instrument, wherein the instrument is respectively arranged around a tunnel and near a tunnel face construction surface;
installing a monitoring module, arranging instruments around a tunnel and near a tunnel face construction surface respectively, installing a wireless base station at a tunnel opening, wirelessly transmitting signals to the opening base station by a monitoring sensor through an antenna, and uploading information to a computer end by the base station to store data;
a wireless base station is installed at a tunnel portal, a monitoring sensor wirelessly transmits signals to the portal base station through an antenna, and the base station uploads information to a computer end for data storage;
p2, based on an improved CRITIC algorithm, preprocessing the related water burst data of the tunnel; the problem of large numerical value difference among indexes due to different dimensions is solved, so that the sample data can more clearly reflect the importance degree of the indexes;
the tunnel water burst data includes: the environmental humidity, the water inflow, the ground water level, the surrounding rock stress, the ground water osmotic pressure and the rainfall of the ground and surrounding areas of the tunnel in the tunnel;
collecting the tunnel water burst factors, and preprocessing tunnel water burst data by utilizing an improved CRITIC algorithm;
p3, carrying the pretreated tunnel water inflow variable data and geological environment information codes into a machine learning algorithm, and training a prediction model;
the data converted into supervised learning is brought into the LSTM model for predictive analysis, and when learning to train the neural network model, a training process and a verification process are generally included, so that batch data divided by step size needs to be further divided into a training data set and a verification data set. In all model training processes of this embodiment, the training data set accounts for 80% of the total batch data, and the verification data set accounts for 20%;
the data converted into supervised learning is brought into a LightGBM model for predictive analysis, a gradient lifting decision tree model is trained, a test network is obtained, and the training process is as follows: firstly determining a primary optimal division point of the LightGBM model according to the training set and a gain formula, secondly generating a leaf division point of an initial LightGBM index model according to the primary optimal division point, then determining a leaf division point gain maximum point according to a preset division threshold value, generating a decision tree of the initial LightGBM model according to the leaf division point gain maximum point and the leaf division point, and finally configuring the initial LightGBM model by utilizing the decision tree to obtain the trained LightGBM model.
Gradient-lifting decision tree model (LightGBM) learning includes: definition of the Weak model F p (x) The model meaning is a decision tree model, which represents a model generated by p iterations, and consists of p weak models, wherein the weak learner for defining the integrated p iterations is a strong learner:
F p (x)＝f 1 (x)+f 2 (x)+…+f p (x)
in machine learning, training of a model is mainly composed of two steps, namely, selection of super parameters and training tuning. The super-parameters have a very large impact on the performance of the machine learning model, so in practice these two steps are usually iterated continuously until the super-parameters most suitable for the task are found.
The optimization of the super parameters mainly has two difficulties, namely, the combination optimization problem cannot be optimized by a gradient descent method like the general parameters, and a general effective optimization method is not available; secondly, the time cost of evaluating the hyper-parameter configuration is very high, so that some optimization methods are difficult to apply in hyper-parameter optimization. In view of the above difficulties, generally initializing hyper-parameters uses relatively simple and efficient methods such as grid search, random search, bayesian optimization, and the like. Each method has respective advantages and disadvantages and application ranges, the LSTM super-parameters considered by the model of the invention have a neural network layer and the number of neurons, the LightGBM super-parameters considered by the model of the invention have a learning rate and the number of decision trees, and the embodiment of the invention adopts a grid search method to optimize the super-parameters.
The grid search is simply to manually give out the parameters which are required to be changed in the model, automatically train the model by using a program, and finally output the accuracy of the model under different super parameters. Because the LSTM and the LightGBM model reach better effect under the condition of super-parameter defaulting, the super-parameter which is not far away from the default is searched by taking the default value of the super-parameter as the center.
Predicting data by root mean square errorEvaluating, and when training iterates for a certain number of times, taking the test set into the model for verificationAnd if the RMSE meets the requirements, completing model training.
Determining geological background information of the face according to geological environment investigation information finished in the earlier stage, determining surrounding rock grade of the face, determining whether the face passes through a geological structure, and performing coding operation according to the information;
p4, weighting the output values of the two models to finally obtain a predicted value of the tunnel environment time sequence data at the next moment;
p5, carrying out hierarchical early warning based on a prediction result, setting four early warning levels aiming at water inflow data in different ranges, and setting different early warning values for blue early warning, orange early warning, yellow early warning and red early warning according to different early warning levels to guide site construction;
calculating the water inflow change rate V at the moment t by the predicted value of the tunnel water inflow at the moment t+1 t+1 V is set up t+1 And comparing the detected value with a set early warning threshold value, and carrying out hierarchical early warning.
When V is t+1 ＞500m 3 At/h, red early warning is carried out, 50m 3 /h≤V t+1 ≤500m 3 Yellow early warning at/h, 5m 3 /h≤V t+1 ≤50m 3 Orange early warning at/h, V t+1 ＜5m 3 And (3) blue early warning is carried out at the time of/h.
In step P1, the tunnel related water gushing data includes the environmental humidity in the tunnel, the water gushing flow, the groundwater level, the surrounding rock stress, the groundwater osmotic pressure, and the rainfall of the ground and the surrounding area of the tunnel.
Humidity transducer integrates in the box body, and convenient to carry need not external power supply by battery power, lays near the face and tunnel each important construction operation face monitoring tunnel environment humidity change, by portable power source power supply, can dismantle at any time and be applied to the different mileage environment monitoring in tunnel, along with the construction is developed, monitors operation face and the nearby environmental data of face at any time.
The rainfall sensor is high in measurement accuracy, particularly suitable for monitoring strong rainfall (> 10 mm/min), convenient to install, free from being influenced by rain, high in stability in severe weather such as rain and snow, high in reliability, used for monitoring rainfall change of the ground and surrounding areas of a tunnel, and capable of being used for monitoring the open area in a certain distance around the tunnel and the position close to an entrance and an exit of the tunnel.
The water level sensor has high control precision, is internally provided with the emitting diode and the phototransistor, has long service time, can be installed in any direction and is simple to install, the direction can be adjusted according to the construction monitoring requirement, and the water level sensor is used for monitoring the change of the underground water level in the tunnel and is installed at the tunnel portal, the tunnel roof, the tunnel bottom and the two sides of the tunnel to monitor the change of the underground water level of the tunnel.
The stress sensor does not need to be in direct contact with an object to be tested, can acquire the internal stress condition through external monitoring, has a simple structure, few parts and multiple microelectronic technologies, can stably operate for a long time in a severe environment, is used for monitoring the stress change of surrounding rock, and is installed at the fracture of a rock layer and the deformation of a rock body in a tunnel.
The flow sensor can monitor the flow condition of water in the tunnel in real time, and outputs data to a computer or other data acquisition equipment for analysis and processing, so that automatic control is realized, the flow sensor has a simple structure and higher reliability, is used for monitoring the change of the flow quantity in the tunnel, is arranged in drainage grooves on two sides near the tunnel outlet, and monitors the integral water inflow of the tunnel.
The underground water pressure sensor is used for monitoring the pressure change condition of underground water around the tunnel and is arranged at the position where underground water leakage and water burst occur in the tunnel.
Each monitoring instrument is powered by a mobile power supply, and data is transmitted to a base station through wireless transmission and then transmitted to a computer through the base station.
The advantage of processing the sequence data through a machine learning method, the historical data of the factors related to the water burst of the tunnel and the geological environment information are taken as training data, the water burst of the working face of the tunnel is predicted accurately and stably, the construction tunnel is monitored in real time through the intelligent early warning system of the water burst tunnel, early warning is timely sent, and operators are evacuated, so that the accident rate is reduced, and the safety of the operators is guaranteed.
The method can be suitable for harsh construction environments, and all-weather multi-source monitoring of the tunnel is realized through multi-source monitoring of environmental humidity, groundwater level, surrounding rock stress, groundwater osmotic pressure, and rainfall of the ground and surrounding areas of the tunnel.
The above description is not intended to limit the invention to the particular embodiments disclosed, but on the contrary, the intention is to cover all modifications, equivalents, and alternatives falling within the spirit and principles of the invention.
Claims (5)
1. The utility model provides a tunnel water inflow intelligent monitoring system, includes multisource real-time monitoring unit and early warning unit, its characterized in that: the multi-source real-time monitoring unit comprises a humidity sensor, a flow sensor, a water level sensor, a stress sensor, an underground water pressure sensor and a rainfall sensor; the early warning unit comprises a data monitoring module, a data storage module, a prediction analysis module and an early warning module;
the humidity sensor is used for monitoring humidity of different mileage of the tunnel;
the flow sensor is used for monitoring the flow change of water burst in the tunnel;
the water level sensor is used for monitoring the change of the underground water level in the tunnel;
the stress sensor is used for monitoring the stress change of surrounding rock in the tunnel;
the underground water pressure sensor is used for monitoring the underground water level change in the tunnel;
the rainfall sensor is used for monitoring the change of the water inflow in the tunnel;
the data monitoring module is used for monitoring that the sensor transmits signals to the base station of the hole in a wireless way through an antenna, and the base station uploads information to the data storage module of the computer end;
the prediction analysis module is used for preprocessing sample data through a CRITIC algorithm; the pretreated tunnel water inflow variable data and geological environment information codes are brought into an LSTM model and a LightGBM model, the prediction model is trained, the output values of the LSTM model and the LightGBM model are weighted, and finally the prediction value of the tunnel environment time sequence data at the next moment is obtained;
the early warning module carries out grading early warning aiming at data of different water inflow, and the grading early warning comprises blue early warning, orange early warning, yellow early warning and red early warning.
2. The intelligent monitoring system for tunnel water inflow of claim 1, wherein: the CRITIC algorithm comprises the following steps:
the j index of the i object of S1, u samples and v indexes takes x as the value ij Form an original sample data matrix x= (X) ij ) uxv ；
S2, normalizing each index value in the X through a Z-score method to obtain a normalized matrix as follows:
in the formula (I), the components of the compound,sum s j The mean value and the standard deviation of the j index are respectively;
S3、sum s j The calculation formula of (2) is as follows:
s4, calculating a variation system v of each index j The following are provided:
in the formula (II), v j A coefficient of variation for the j-th index;
s5, calculating a standardized matrix X * The correlation coefficient of each index is obtained as follows:
R＝(r kl ) v×v (k=1, 2, …, v; l=1, 2, …, v) (three)
In the formula (III), r kl Is the correlation coefficient between the K and the l index;
s6, calculating coefficients reflecting the degree of independence of the indexes:
and finally, calculating the weight of each index:
obtaining a final sample matrix X f ：
3. The intelligent monitoring system for tunnel water inflow of claim 1, wherein: the pretreated tunnel water burst variable data and geological environment information codes are brought into an LSTM model and a LightGBM model, and the prediction model is trained as follows:
h1, analyzing geological environment characteristics of tunnel face excavation, and performing coding operation on geological environment information;
h2, encoding the preprocessed feature matrix variable data and geological environment information, converting the feature matrix variable data and geological environment information into supervised learning data through a sliding window algorithm, setting an early warning time step length to be 30 minutes according to the actual construction condition of the site, and respectively predicting tunnel water inflow data at the time t+1 (namely after 30 minutes) through data at the time t;
h3, taking the surrounding rock grade, geological structure and whether water burst of the tunnel face occurs or not where the tunnel face is currently excavated as constraint conditions, and performing coding operation so that the early warning model converts geological environment information into data so as to facilitate analysis of the prediction model;
and H4, respectively solving errors of the LightGBM model and the LSTM model, and carrying out weighted combination on the prediction results of the two models by using an error reciprocal method to finally obtain the prediction value of the tunnel environment time sequence data at the next moment, wherein the weighted formula is as follows:
in the formula (IV), n 1 Is the predicted value of LSTM, n 2 Is the predicted value of the LightGBM model, a 1 Is the error of LSTM model, a 2 Error of the LightGBM model, m 1 And m 2 The weight coefficients of the LSTM model and the LightGBM model are respectively;
h5, calculating the water inflow change rate as follows:
in the formula (five), X True sense Is the true value of the tunnel water inflow at the moment t,the predicted value of the tunnel water inflow at the time t+1;
and H6, carrying out hierarchical early warning according to a threshold value set in a construction environment, wherein the method comprises the following steps:
when v t ＜v min When the method is used for blue early warning, early warning is not carried out, and workers normally construct the method;
when v min ＜v t ＜v med When the building is in orange early warning, workers normally construct the building and are ready to evacuate at any time;
when v med ＜v t ＜v max When the system is yellow, early warning is carried out, the personnel in the face are immediately evacuated, and the subsequent personnel are evacuated orderly;
when v max ＜v t And when the system is in red early warning, all constructors immediately withdraw and immediately take emergency measures.
4. An intelligent monitoring and early warning method for tunnel water inflow is characterized by comprising the following steps of: a tunnel inflow intelligent monitoring system according to any one of claims 1-3, the method comprising the steps of:
p1, preprocessing relevant water burst data of a tunnel based on a CRITIC algorithm;
p2, carrying the pretreated tunnel water inflow variable data and geological environment information codes into a machine learning algorithm, and training a prediction model;
p3, weighting the output values of the two models to finally obtain a predicted value of the tunnel environment time sequence data at the next moment
And P4, carrying out hierarchical early warning based on a prediction result, setting four early warning levels aiming at water inflow data in different ranges, carrying out blue early warning, orange early warning, yellow early warning and red early warning, setting different early warning values for different early warning levels, and guiding site construction.
5. The intelligent monitoring and early warning method for tunnel water inflow is characterized in that: in step P1, the tunnel related water gushing data includes the environmental humidity in the tunnel, the water gushing flow, the groundwater level, the surrounding rock stress, the groundwater osmotic pressure, and the rainfall of the ground and the surrounding area of the tunnel.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311417662.4A CN117345342A (en) | 2023-10-30 | 2023-10-30 | Intelligent monitoring system and early warning method for tunnel water inflow |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311417662.4A CN117345342A (en) | 2023-10-30 | 2023-10-30 | Intelligent monitoring system and early warning method for tunnel water inflow |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117345342A true CN117345342A (en) | 2024-01-05 |
Family
ID=89361150
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311417662.4A Pending CN117345342A (en) | 2023-10-30 | 2023-10-30 | Intelligent monitoring system and early warning method for tunnel water inflow |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117345342A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117786822A (en) * | 2024-02-23 | 2024-03-29 | 北京交通大学 | Waterproof treatment management system suitable for shield tunnel emergency |
| CN118296929A (en) * | 2024-02-28 | 2024-07-05 | 山东大学 | Multi-source information fusion tunnel dynamic water grouting plugging optimization design method and system |
| CN118333433A (en) * | 2024-06-14 | 2024-07-12 | 中国电建集团贵阳勘测设计研究院有限公司 | Target seepage prediction method based on BO-GBRT strategy |
| CN118504282A (en) * | 2024-07-15 | 2024-08-16 | 云南交投集团公路建设有限公司 | Metering data acquisition and processing method based on tunnel counter-slope drainage |
-
2023
- 2023-10-30 CN CN202311417662.4A patent/CN117345342A/en active Pending
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117786822A (en) * | 2024-02-23 | 2024-03-29 | 北京交通大学 | Waterproof treatment management system suitable for shield tunnel emergency |
| CN117786822B (en) * | 2024-02-23 | 2024-05-03 | 北京交通大学 | Waterproof treatment management system suitable for shield tunnel emergency |
| CN118296929A (en) * | 2024-02-28 | 2024-07-05 | 山东大学 | Multi-source information fusion tunnel dynamic water grouting plugging optimization design method and system |
| CN118333433A (en) * | 2024-06-14 | 2024-07-12 | 中国电建集团贵阳勘测设计研究院有限公司 | Target seepage prediction method based on BO-GBRT strategy |
| CN118504282A (en) * | 2024-07-15 | 2024-08-16 | 云南交投集团公路建设有限公司 | Metering data acquisition and processing method based on tunnel counter-slope drainage |
| CN118504282B (en) * | 2024-07-15 | 2024-09-13 | 云南交投集团公路建设有限公司 | Metering data acquisition and processing method based on tunnel counter-slope drainage |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN117345342A (en) | Intelligent monitoring system and early warning method for tunnel water inflow | |
| Zhou et al. | Dynamic prediction for attitude and position in shield tunneling: A deep learning method | |
| Wu et al. | Rock burst prediction probability model based on case analysis | |
| US20220308568A1 (en) | System and method for monitoring soil gas and performing responsive processing on basis of result of monitoring | |
| CN117870609B (en) | Soft rock tunnel face deformation monitoring method based on incomplete arch effect | |
| MX2011004640A (en) | Apparatus for analysis and control of a reciprocating pump system by determination of a pump card. | |
| CN115324650A (en) | Early warning grading system and method based on intelligent recognition | |
| CN113685188A (en) | TBM tunneling optimization method based on physical characteristics of rock slag | |
| US20230392498A1 (en) | Emissions estimations at a hydrocarbon operation location using a data-driven approach | |
| CN117191147A (en) | Flood discharge dam water level monitoring and early warning method and system | |
| CN116341272A (en) | Construction safety risk management and control system for digital distribution network engineering | |
| CN117711140A (en) | Tunnel water bursting disaster timing early warning method and system based on multi-source data fusion | |
| CN118094379A (en) | Oil well fault diagnosis prediction method and system for multi-factor fusion analysis | |
| CN1896458B (en) | System and methods of deriving fluid properties of downhole fluids and uncertainty thereof | |
| CN115619585A (en) | Intelligent mine environment monitoring and management method and system | |
| CN113835138B (en) | Method for predicting total organic carbon content of shale based on deep coding decoding network | |
| CN116090322A (en) | Intelligent decision visualization system for pipeline integrity | |
| CN117634308A (en) | Decision system and analysis method for analyzing rail transit engineering data | |
| CN111340275B (en) | Tunnel support mode selection real-time prediction method based on detection while drilling technology | |
| CN117474340A (en) | Risk evaluation method and system for subway shield construction settlement | |
| CN112084553A (en) | Surveying method for tunnel planning | |
| CN116164241A (en) | Intelligent detection method for leakage faults of gas extraction pipe network | |
| CN114066271A (en) | Tunnel water inrush disaster monitoring and management system | |
| CN117189249A (en) | Tunnel multielement information on-site monitoring method based on wireless transmission of Internet of things | |
| Yu et al. | Predicting structure performance of urban critical infrastructure: an augmented attention-based LSTM model |
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 |