A tunnel construction water inflow calculation method based on machine learning

Abstract

The application discloses a tunnel construction water inflow calculation method based on machine learning and belongs to the field of high-precision prediction of tunnel water inflow disasters. Accurate prediction of tunnel water inflow is one of the focal point problems of underground engineering, which is directly related to the engineering construction progress, cavern stability and construction personnel safety. Therefore, a method for calculating the tunnel water inflow by constructing a prediction model through a plurality of machine learning algorithms is proposed. The method comprises the following steps: for the tunnel section needing to calculate the water inflow, a geological condition database is established through drilling data, field survey data and long-term observation data in the early stage; the data in the database are analyzed and screened; the influencing factors are selected and preprocessed; the selected influencing factors are used to iteratively update the prediction model; and finally, the high-precision tunnel water inflow prediction result can be calculated by inputting the field survey data of the tunnel construction area.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of high-precision prediction of tunnel gushing water disasters, and in particular to a tunnel construction gushing water quantity calculation method based on machine learning. BACKGROUND

[0002] In various types of projects such as water conservancy, transportation, underground mineral exploitation, etc., tunnel excavation is involved and may cause sudden gushing water accidents. The occurrence probability of sudden gushing water is high, the water inrush quantity is large, the water pressure is high, and it is difficult to accurately judge the position and predict the scale and dynamic characteristics because the occurrence process is sudden. The sudden gushing water problem often affects the construction, causes the surrounding rock to lose stability, falls off and blocks the tunnel, and at the same time causes ground subsidence and ground settlement and pollution of the surrounding water environment, affects long-term operation and the ecological environment of the ground surface, and even causes more serious life safety and property loss.

[0003] There are many methods for calculating gushing water quantity, which can be summarized as follows: approximate method, theoretical method, stochastic mathematical method, nonlinear theoretical method, and numerical method. The approximate method is mostly derived from the summary of engineering practice, and has good prediction accuracy for gushing water problems under simple geological conditions. The theoretical method often needs to be corrected in combination with measured data, and the theoretical accuracy may not be sufficient in actual application. The stochastic mathematical method is sensitive to abnormal data, and the results may be significantly affected by data noise or modeling errors. The nonlinear theoretical method requires a large amount of detailed observation data, especially in time series and spatial distribution. The numerical method needs to be verified by field data or theoretical models, and it may be difficult to judge the credibility of the results when there is a lack of experience. Machine learning method is a new means of quantitative analysis and calculation of groundwater, and is one of the important means, which is particularly suitable for complex geological conditions.

[0004] At present, the method for obtaining hydrogeological parameters for accurate prediction of tunnel gushing water is relatively perfect, but how to use various parameters to complete simple, fast, convenient and effective gushing water quantity prediction is still a difficulty. Therefore, the present application is based on random forest (RF), BP (Back Propagation) neural network algorithm, etc. to build an accurate prediction model for tunnel gushing water, which fills the gap in the field of high-precision prediction of tunnel gushing water disasters. SUMMARY

[0005] The purpose of the present application is to solve the problem of calculating tunnel gushing water quantity by integrating machine learning models, and to provide a tunnel construction gushing water quantity calculation method based on machine learning, which has low time cost, high accuracy and is more convenient.

[0006] Technical scheme, in order to achieve the above purpose, the present application provides a tunnel construction gushing water quantity calculation method based on machine learning, which comprises the following specific steps:

[0007] 1) Collect the geological environment data of the surrounding area of the tunnel project, including drilling data, surface exploration data, underground rock stratum data, hydrological information, and import into the heterogeneous gushing disaster big data storage;

[0008] 2) Use the large island ocean formula and the modified Goodman formula to preliminarily determine the tunnel unit length gushing quantity and the maximum gushing quantity of the calculation area, and use the unit length gushing quantity to judge the actual occurrence grade of the gushing disaster, and judge the danger grade of the gushing disaster according to the maximum unit length gushing quantity;

[0009] 3) Import the multi-source exploration information of the study section into the heterogeneous data storage based on the Hadoop system, and according to the fitting results of the field data of the section and the past gushing accident area field data in the original database, fill in the data vacancy of the prediction section and eliminate abnormal data;

[0010] 4) Input the data in the database into the simulation software Groundwater Model System (GMS) for three-dimensional geological modeling, and based on the differences between modeling and field conditions, the filled data is supplemented and corrected, and the database is further fitted and updated;

[0011] 5) Use the negative topographic type, stratum lithology, unfavorable geological type, structure surface type and development degree, karst development degree, groundwater occurrence or gushing type, rainfall, construction disturbance degree, rock stratum inclination and rock stratum dip angle, permeability coefficient, groundwater level depth, horizontal distance and vertical distance between groundwater body and tunnel, tunnel length and tunnel depth, and groundwater head information as input information of the prediction model, and iterate and update the initial model, so that the fitting degree of the tunnel prediction unit length gushing quantity output by the model and the actual unit length gushing quantity of the past gushing disaster is more than 85%, and the neural network model is trained;

[0012] 6) Collect the above data of the study area and input it into the trained neural network model to output the tunnel unit length gushing quantity, and use the unit length gushing quantity to judge the actual occurrence grade of the gushing disaster.

[0013] Further, in step 1), when collecting project example data, select a construction tunnel with complete pre-exploration data, and the missing rate of exploration information should be less than 30%, and the surrounding area data should at least include the information of tunnel water pressure, surrounding rock permeability coefficient, surrounding rock broken degree, structure surface type and development degree, tunnel length and depth.

[0014] Further, in step 2), the tunnel unit length maximum gushing quantity is calculated as follows:

[0015]

[0016] wherein q0 is the maximum water inflow per unit length of the tunnel (m / d); K is the permeability coefficient of the surrounding rock (m / d); m is an empirical correction coefficient, with a value of 0.86; h is the distance from the initial groundwater level to the tunnel floor (m); r0 is the radius of the equivalent circle of the tunnel cross section (m); and d is the diameter of the equivalent circle of the tunnel cross section (m). 3

[0017] The correspondence between the hazard level of the water inflow disaster of the tunnel and the predicted maximum water inflow is as follows: I level, q0>1(10 3 m 3 / d), extremely high risk, disastrous consequences: II level, 0.3 3 m 3 / d), high risk, serious harm: III level, 0.05 3 m 3 / d), moderate risk, delay in construction: IV level, q0<0.05(10 3 m 3 / d), low risk.

[0018] The water inflow per unit length of the tunnel is calculated as follows:

[0019]

[0020] wherein Q is the water inflow per unit length of the tunnel (m / d); H is the height of the groundwater level to the center line of the tunnel (m); r is the radius of the tunnel (m); and K is the permeability coefficient of the surrounding rock (m / d). The formula assumes that the initial groundwater level is horizontal and unchanging, that the groundwater is in a steady state and satisfies radial flow, and that the medium is homogeneous and isotropic.

[0021] The correspondence between the occurrence level of the water inflow disaster of the tunnel and the predicted water inflow is as follows: I level, Q>5(10 2 m 3 / d), extremely high possibility: II level, 1.5 2 m 3 / d), high possibility: III level, 0.25 2 m 3 / d), moderate possibility: IV level, Q<0.25(10 2 m 3 / d), low possibility.

[0022] ​Further, in the step 3), the data storage library built should contain the information in step 1), and the existing information is screened and cleaned, and the data with the relative error of the fitting result exceeding 50% is removed. When the data storage library is used to supplement the geological information of the study area, it should be further ensured that the proportion of missing data in the total data should be less than 25% to avoid the occurrence of data distortion phenomenon.

[0023] Further, in the step 4), the fitting coefficient of the corrected GMS three-dimensional data model and the actual situation should be kept above 0.9 to ensure the accuracy of the filled data.

[0024] Further, in the step 5), the selected model is any one of SVM, AdaBoost and BP.

[0025] Compared with the prior art, the technical scheme of the present application has the following beneficial technical effects:

[0026] The machine learning-based tunnel construction water inflow calculation method provided by the present application is more close to the actual situation and more accurate in calculation ability than the traditional method of simply relying on original data for calculation and modeling; compared with the traditional calculation method, the machine learning model pays more attention to the nonlinear relationship between various calculation factors, relies on deep learning of the nonlinear relationship to obtain accurate water inflow calculation results, and is simple to operate and low in cost; when multiple machine learning algorithm methods are used to calculate the tunnel water inflow, the method can quickly and accurately obtain the predicted water inflow of different sections of the underground tunnel according to the field investigation data. In actual engineering application, the average error of the water inflow calculated by the method and the actual water inflow is less than 7%, which has high economic and application value. The present application provides an innovative technical idea for constructing an accurate prediction model to calculate the tunnel water inflow by using machine learning algorithm. BRIEF DESCRIPTION OF DRAWINGS

[0027] In order to more clearly illustrate the technical scheme in the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments or prior art description will be briefly introduced as follows.

[0028] Figure 1 The flowchart for solving the tunnel construction water inflow based on machine learning involved in the present application. DETAILED DESCRIPTION

[0029] The present application will be further described below in combination with the drawings and specific embodiments.

[0030] The present application provides a machine learning-based tunnel construction water inflow calculation method, which comprises the following specific steps:

[0031] 1) Collect the geological environment data of the surrounding area of the tunnel project, including drilling data, surface exploration data, underground rock stratum data, hydrological information, and import into the heterogeneous gushing disaster big data storage;

[0032] 2) Use the large island ocean formula and the modified Goodman formula to preliminarily determine the tunnel unit length gushing quantity and the maximum gushing quantity of the calculation area, and use the unit length gushing quantity to determine the actual occurrence grade of the gushing disaster, and determine the danger grade of the gushing disaster according to the maximum unit length gushing quantity;

[0033] 3) The multi-source exploration information of the research section is imported into the heterogeneous data storage based on the Hadoop system, and according to the fitting results of the field data of the section and the past gushing accident area field data in the original database, the data gaps of the prediction section are filled and the abnormal data are eliminated;

[0034] 4) The data in the database is input into the simulation software Groundwater Model System (GMS) for three-dimensional geological modeling, and the filled data is supplemented and corrected based on the differences between the modeling and the actual situation, and the database is further fitted and updated;

[0035] 5) Use the negative topographic type, stratum lithology, unfavorable geological type, structure surface type and development degree, karst development degree, groundwater occurrence or gushing type, rainfall, construction disturbance degree, rock stratum inclination and rock stratum dip angle, permeability coefficient, groundwater level depth, horizontal distance and vertical distance between groundwater body and tunnel, tunnel length and tunnel depth, and groundwater head information as input information of the prediction model, and iterate and update the initial model to make the fitting degree of the tunnel prediction unit length gushing quantity output by the model and the actual unit length gushing quantity of the past gushing disaster reach more than 85%, and complete the training of the neural network model;

[0036] 6) Collect the above data of the research area and input them into the trained neural network model to output the tunnel unit length gushing quantity, and use the unit length gushing quantity to determine the actual occurrence grade of the gushing disaster.

[0037] Further, in step 1), when collecting project example data, select a construction tunnel with complete pre-exploration data, and the missing rate of exploration information should be less than 30%, and the surrounding area data should at least include the information of tunnel water pressure, surrounding rock permeability coefficient, surrounding rock broken degree, structure surface type and development degree, tunnel length and depth.

[0038] Further, in step 2), the tunnel unit length maximum gushing quantity is calculated as follows:

[0039]

[0040] In the formula, q0 is the maximum water inflow per unit length of the tunnel (m³). 3 / d); K is the permeability coefficient of the surrounding rock (m / d); m is the empirical correction coefficient with a value of 0.86; h is the distance from the initial groundwater level to the tunnel floor (m); r0 is the radius of the equivalent circle of the tunnel cross section (m); d is the diameter of the equivalent circle of the tunnel cross section (m);

[0041] The correspondence between the hazard level of tunnel water inrush disaster and the predicted maximum water inrush volume is as follows: Level I, q0>1(10 3 m 3 / d), extremely high risk, catastrophic consequences: Level II, 0.3 < q0 < 1 (10 3 m 3 / d), High risk, serious hazard: Level III, 0.05 < q0 < 0.3 (10 3 m 3 / d), medium risk, construction delay: Level IV, q0 < 0.05 (10 3 m 3 / d), low risk;

[0042] The water inflow per unit length of the tunnel is calculated as follows:

[0043]

[0044] In the formula, Q is the water inflow per unit length of the tunnel (m / d); H is the height from the groundwater level to the centerline of the tunnel (m); r is the radius of the tunnel (m); K is the permeability coefficient of the surrounding rock (m / d). This formula assumes that the initial groundwater level is horizontal and constant, the groundwater is in a stable state and satisfies radial flow, and the medium is homogeneous and isotropic.

[0045] The correspondence between the occurrence level of tunnel water inrush disaster and the predicted water inrush volume is as follows: Level I, Q > 5 (10 2 m 3 / d), extremely high probability: Level II, 1.5 < Q < 5 (10 2 m 3 / d), High probability: Level III, 0.25 < Q < 1.5 (10 2 m 3 / d), moderate probability: Level IV, Q < 0.25 (10 2 m 3 / d), low probability.

[0046] Further, in step 3), the data storage library should contain the information in step 1), and the existing information is screened and cleaned, and the data with the relative error of more than 50% fitting results is removed. When the data storage library is used to complete the geological information of the study area, it should be further ensured that the proportion of missing data in the total data should be less than 25% to avoid data distortion.

[0047] Further, in step 4), the fitting coefficient of the corrected GMS three-dimensional data model and the actual situation should be kept above 0.9 to ensure the accuracy of the filled data.

[0048] Further, in step 5), the selected model is any one of SVM, AdaBoost and BP.

Claims

1. A method for calculating water inflow in tunnel construction based on machine learning, characterized by, The method comprises the following specific steps: 1) Collecting geological environment data of the surrounding area of the tunnel project, including drilling data, surface reconnaissance data, underground rock stratum data, hydrological information, and importing into a heterogeneous gushing disaster big data storage library; 2) Preliminarily determining and calculating the tunnel unit length gushing quantity and the maximum gushing quantity of the gushing disaster in the calculation area, and using the unit length gushing quantity to judge the actual occurrence grade of the gushing disaster, and using the maximum unit length gushing quantity to judge the danger grade of the gushing disaster; 3) Importing the multi-source exploration information of the research section into the heterogeneous data storage library built based on the Hadoop system, filling in the data vacancy of the prediction section according to the fitting results of the field data of the section and the past gushing accident area field data in the original database, and eliminating abnormal data; 4) Inputting the data in the database into the simulation software Groundwater Model System (GMS) to perform three-dimensional geological modeling, and based on the difference between the modeling and the actual situation, the filled data is supplemented and corrected to complete the fitting update of the database; 5) Using the negative terrain type, stratum lithology, unfavorable geological type, structure surface type and development degree, karst development degree, groundwater occurrence or gushing type, rainfall, construction disturbance degree, rock stratum inclination and rock stratum dip angle, permeability coefficient, groundwater level depth, horizontal distance and vertical distance of the groundwater body from the tunnel, tunnel length and tunnel depth, and groundwater head information as input information of the prediction model, iteratively updating the initial model, so that the fitting degree of the tunnel prediction unit length gushing quantity output by the model and the actual unit length gushing quantity of the past gushing disaster is more than 85%, and the training of the neural network model is completed; 6) Collecting the above-mentioned data of the research area and inputting into the trained neural network model to output the tunnel unit length gushing quantity, and using the unit length gushing quantity to judge the actual occurrence grade of the gushing disaster.

2. The machine learning based tunnel construction inflow calculation method according to claim 1, characterized in that, In the step 1), when collecting the engineering example data, the construction tunnel with complete pre-exploration data is selected, the missing rate of the exploration information is less than 30%, and the surrounding area data at least contains the information of tunnel water pressure, surrounding rock permeability coefficient, surrounding rock breaking degree, structure surface type and development degree, tunnel length and depth.

3. The machine learning based tunnel construction inflow calculation method according to claim 1, characterized in that, In the step 2), the tunnel unit length maximum gushing quantity is calculated as follows: where q0is the maximum water inflow per unit length of the tunnel (m 3 / d); K is the permeability coefficient of the surrounding rock (m / d); m is a correction factor with a value of 0.86; h is the distance from the initial groundwater level to the tunnel floor (m); r0is the equivalent circle radius of the tunnel cross section (m); and d is the equivalent circle diameter of the tunnel cross section (m). The corresponding relationship between the tunnel water gushing disaster risk level and the predicted maximum water gushing quantity is: I level, q0>1(10 3 m 3 / d), extremely high risk, disastrous consequences: II level, 0.3 3 m 3 / d), high risk, serious harm: III level, 0.05 3 m 3 / d), medium risk, delay construction: IV level, q0<0.05(10 3 m 3 / d), low risk; The tunnel unit length gushing quantity is calculated as follows: In the formula, Q is the tunnel unit length gushing quantity (m / d), H is the height from the groundwater level to the tunnel center line (m), r is the tunnel radius (m), and K is the surrounding rock permeability coefficient (m / d). The formula assumes that the initial groundwater level is horizontal and unchanged, the groundwater is in a stable state and satisfies radial flow, and the medium is homogeneous and isotropic; The corresponding relationship between the tunnel water inrush disaster grade and the tunnel unit length water inrush amount is: I grade, Q>5(10 2 m 3 / d), extremely high possibility: II grade, 1.5 2 m 3 / d), high possibility: III grade, 0.25 2 m 3 / d), medium possibility: IV grade, Q 2 m 3 / d), low possibility.

4. The machine learning based tunnel construction inflow calculation method of claim 1, wherein, In the step 3), the data storage library built contains the information in the step 1), and the existing information is screened and cleaned, and the data with a relative error of more than 50% in the fitting result is eliminated. When the data storage library is used to complete the geological information of the research area, it is ensured that the missing data accounts for less than 25% of the total data. 5.The machine learning based tunnel construction inflow calculation method according to claim 1, wherein, In the step 4), the fitting coefficient of the corrected GMS three-dimensional data model and the actual situation is kept above 0.

9.

6. The machine learning based tunnel construction inflow calculation method of claim 1, wherein, In the step 5), the selected neural model is any one of SVM, AdaBoost and BP.

Images