TBM rock breaking efficiency intelligent prediction method based on rock size

Abstract

The application discloses a kind of TBM rock breaking efficiency intelligent prediction methods based on rock slag size, which comprises the following steps: install industrial camera on TBM slag belt, and real-time acquisition rock slag image information in TBM tunneling process;The rock slag image information obtained is automatically identified, and the shape and geometric characteristics of the rock slag are obtained;Collect TBM's on-site tunneling parameters, rock slag geometric information characteristics and rock breaking efficiency data, and form sample data set;A gradient boosting regression tree model optimized by particle swarm algorithm is established using the sample data set, and the intelligent prediction of TBM rock breaking efficiency is realized.The application uses particle swarm algorithm to optimize the gradient boosting regression tree model, so as to find the best parameters of the gradient boosting regression tree model, improve the prediction accuracy of the model, and the accurate prediction of rock breaking efficiency can effectively provide reference for TBM tunneling parameter setting and efficient rock breaking.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of TBM rock breaking efficiency prediction, and particularly relates to a TBM rock breaking efficiency intelligent prediction method based on rock residue size. BACKGROUND

[0002] With the continuous development of underground space, the excavation difficulty and mechanization level of tunnel engineering are constantly improving, and the full-face tunnel boring machine (TBM) has been rapidly developed and applied. Compared with the traditional drilling and blasting method, the TBM has the following advantages: ① ensuring the safety of construction; ② the tunneling speed is 3-10 times that of the traditional drilling and blasting method; ③ the tunnel forming effect is good, realizing the integration of assembly; ④ the noise disturbance is small, and the influence on the surrounding environment is small. At present, the TBM has been widely applied in the fields of transportation, municipal, hydraulic, and mining engineering. The TBM mainly relies on the interaction between the cutter installed on the cutter head of the tunneling machine and the rock to achieve rock breaking. The cutter is pressed against the tunnel face under the action of the normal thrust and rolling force, and rock breaking is generated. The accurate prediction of the rock breaking efficiency can effectively provide a reference for the setting of the TBM tunneling parameters.

[0003] Most of the existing researches are about predicting the TBM tunneling performance and rolling wear through artificial intelligence methods, and the research content of intelligent prediction of rock breaking efficiency is still very few. Therefore, it is of great significance to propose an intelligent model that can accurately predict the TBM rock breaking efficiency to realize the rapid tunneling and efficient rock breaking of the TBM. SUMMARY

[0004] The present application relates to the technical field of TBM rock breaking efficiency prediction, and particularly relates to a TBM rock breaking efficiency intelligent prediction method based on rock residue size.

[0005] The technical solution for achieving the present application is as follows:

[0006] A TBM rock breaking efficiency intelligent prediction method based on rock residue size, comprising the following steps:

[0007] Step 1: installing a rock residue image shooting device on the TBM residue belt to real-time collect rock residue image information in the TBM tunneling process;

[0008] Step 2: automatically identifying the obtained rock residue image information to obtain the shape and geometric features of the rock residue;

[0009] Step 3: collecting the on-site tunneling parameters of the TBM, the geometric information features of the rock residue, and the rock breaking efficiency data, and forming a sample data set;

[0010] Step 4: Establish a particle swarm algorithm optimized gradient boosting regression tree model using a sample data set to realize intelligent prediction of TBM rock breaking efficiency.

[0011] Further, the slag image shooting device of step S1 includes an industrial camera and an LED fill light. The industrial camera is a Canon N10-W02 model with 1920x1080 effective pixels and a focal length of 4.3mm. It is arranged on both sides and above the TBM slag belt to shoot real-time photo information of the slag. The LED fill light is a nijia200 professional fill light with a maximum illumination of 70400 lux / 0.5m and a maximum light brightness of 5800 lm, which is used to compensate the light of the slag on the slag belt. The data storage center can transmit the slag images obtained by shooting into the database in real time, and then transmit the data of the database into the data storage center through wireless network.

[0012] Further, the slag image processing of step S2 includes converting the collected color image of the slag into a grayscale image, i.e. when the grayscale value is greater than the threshold value set in advance, it is 1, and less than the threshold value, it is 0. The connected region detection algorithm is used to judge the connected region, which is the position of the slag. According to the pixel points of the connected region, the geometric characteristics of the slag can be calculated.

[0013] Further, the tunneling parameter indicators recorded on site by the TBM of step 3 include total thrust, cutterhead torque and cutterhead speed. The geometric information characteristics of the slag include the minor axis, major axis and intermediate axis of the slag. The rock breaking efficiency is represented by specific energy.

[0014] Further, the step 4 of establishing a particle swarm algorithm optimized gradient boosting regression tree model using a sample data set to realize intelligent prediction of TBM rock breaking efficiency specifically includes:

[0015] Step 4.1: Normalize the collected sample data set and divide it into training set and test set.

[0016] Step 4.2: Use the training set to build a GBRT model and initialize the parameters of GBRT, including the number of weak learners, maximum tree depth and learning rate.

[0017] Step 4.3: Initialize the parameters of particle swarm algorithm, use particle swarm optimization algorithm to optimize the parameters of GBRT, and get the optimal values of the number of weak learners, maximum tree depth and learning rate.

[0018] Step 4.4: Substitute the optimal parameters into the GBRT model to get the trained PSO-GBRT prediction model. Put the test set sample data into the model to get the specific energy prediction value. Use R2 and RMSE indicators to test the prediction performance of the model.

[0019] Further, the data normalization calculation formula described in step 4.1 is as follows:

[0020]

[0021] In the formula, X norm is the normalized value of the data, X is the original value of each parameter in the sample, X max and X min are the maximum and minimum values of each parameter in the sample, respectively.

[0022] Further, the specific steps of using the particle swarm optimization algorithm to optimize the GBRT described in step 4.3 are as follows:

[0023] Step 4.3.1 initializes the population size of the particle swarm optimization algorithm to 20, the maximum number of iterations to 200, the learning factors c1 and c2 to 1.5, determines the model hyperparameters weak learner number, maximum tree depth, and initial value of learning rate, the weak learner number is in the range of 1 to 1000, the maximum tree depth is in the range of 1 to 5, and the learning rate is in the range of 0.01 to 1, and the initial values of these parameters are randomly obtained.

[0024] Step 4.3.2 calculates the fitness value of the particle swarm individual, the fitness value is calculated by the mean square error between the sample predicted value and the sample true value, and the best fitness value in all particles is set as the global extreme value.

[0025] Step 4.3.3 updates the speed and position of the particle and compares it with the extreme value of the previous moment, and updates the extreme value.

[0026] Step 4.3.4 judges whether the maximum number of iterations or the minimum error condition is reached. If not, return to step 4.3.2, if the end condition is met, stop iteration, and output the optimal solution of the parameter weak learner number, maximum tree depth, and learning rate.

[0027] Compared with the prior art, the present application has the following beneficial effects:

[0028] The present application proposes a TBM rock breaking efficiency prediction method based on rock size, while the existing research mainly focuses on using machine learning methods to predict TBM excavation rate and rolling wear, and lacks intelligent prediction of rock breaking efficiency. In addition to using the conventional gradient boosting regression tree model, the present application also combines the particle swarm optimization algorithm to find the best parameters of the gradient boosting regression tree model, thereby improving the prediction accuracy of the model, and the accurate prediction of rock breaking efficiency can effectively provide reference for TBM excavation parameter setting and efficient rock breaking. BRIEF DESCRIPTION OF DRAWINGS

[0029] Figure 1 is the working process diagram of the present application.

[0030] Figure 2 Rock image shooting device schematic diagram DETAILED DESCRIPTION

[0031] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, but not all the embodiments.

[0032] Embodiment 1, reference Figures 1-2 A TBM rock breaking efficiency intelligent prediction method based on rock size includes the following steps:

[0033] Step 1: Install a rock shooting device on the TBM slag belt to collect real-time rock image information during TBM tunneling, as shown in Figure 2 (1) is an industrial camera, (2) is an LED fill light, and (3) is a data storage center. The industrial camera model is Canon N10-W02, the effective pixel number is 1920x1080, and the focal length is 4.3mm. It is arranged on both sides and above the TBM slag belt to shoot real-time rock image information. The LED model is nijia200 professional fill light, the maximum illumination is 70400 lux / 0.5m, and the maximum light brightness is 5800lm, which is used for light compensation of the rock on the slag belt. The data storage center can transmit the rock image obtained by shooting into the database in real time, and then transmit the data of the database into the data storage center through wireless network.

[0034] Step 2: Automatically identify the obtained rock image information to obtain the shape and geometric characteristics of the rock. Rock image processing includes converting the collected rock color image into a gray image, that is, when the gray value is greater than the threshold value set in advance, it is 1, and less than the threshold value, it is 0. The connected region detection algorithm is used to judge the connected region, which is the position of the rock. According to the pixel points of the connected region, the geometric characteristics of the rock can be calculated.

[0035] Step 3: Collect TBM on-site tunneling parameters, rock geometric information characteristics and rock breaking efficiency data, and form a sample data set. The tunneling parameter indexes recorded by TBM on site include total thrust, cutter torque and cutter speed. The geometric information characteristics of the rock include the minor axis, the middle axis and the major axis of the rock. The rock breaking efficiency is represented by specific energy.

[0036] Step 4: Use the sample data set to establish a gradient boosting regression tree model optimized by particle swarm algorithm to realize intelligent prediction of TBM rock breaking efficiency, as follows:

[0037] Step 4.1, the collected sample data set is normalized and divided into training set and test set, and the data is normalized, and the calculation formula is as follows:

[0038]

[0039] In the formula, X norm is the normalized value of the data, X is the original value of each parameter in the sample, X max and X min are the maximum and minimum values of each parameter in the sample, respectively.

[0040] Step 4.2, use the training set to build the GBRT model, initialize the parameters of GBRT, including the number of weak learners, the maximum tree depth, and the learning rate.

[0041] Step 4.3, initialize the parameters of particle swarm algorithm, use particle swarm optimization algorithm to optimize the parameters of GBRT, get the optimal values of weak learner number, maximum tree depth and learning rate, as follows:

[0042] Step 4.3.1, initialize the population number of particle swarm algorithm as 20, the maximum iteration number as 200, the learning factors c1 and c2 as 1.5, determine the initial values of the model hyperparameters weak learner number, maximum tree depth and learning rate, the value range of weak learner number is 1 to 1000, the value range of maximum tree depth is 1 to 5, and the value range of learning rate is 0.01 to 1, and the initial values of these parameters are randomly obtained.

[0043] Step 4.3.2, calculate the fitness value of the particle swarm individual, the fitness value is calculated by the mean square error between the sample predicted value and the sample true value, and the best fitness value in all particles is set as the global extreme value.

[0044] Step 4.3.3, update the velocity and position of the particle, and compare with the extreme value of the previous moment, and update the extreme value.

[0045] Step 4.3.4, judge whether the maximum iteration number or the minimum error condition is reached. If not, return to step 4.3.2, if the end condition is met, stop iteration, and output the optimal solution of the parameters weak learner number,

[0046] maximum tree depth and learning rate.

[0047] The above is only the preferred specific embodiment of the present application, but the protection scope of the present application is not limited thereto, any skilled person in the art can make equivalent replacement or change according to the technical solution and the inventive concept of the present application within the technical range disclosed by the present application, which should be covered within the protection scope of the present application.

Claims

1. A TBM rock breaking efficiency intelligent prediction method based on rock size, characterized by, The steps are as follows: S1: install the rock slag image shooting device on the TBM slag belt, and collect the rock slag image information in the TBM tunneling process in real time; S2: automatically identify the obtained rock slag image information to obtain the shape and geometric characteristics of the rock slag; S3: collect the on-site tunneling parameters of TBM, the geometric information characteristics of rock slag and the rock breaking efficiency data, and form a sample data set; S4: use the sample data set to establish a gradient boosting regression tree model optimized by particle swarm algorithm, and realize intelligent prediction of TBM rock breaking efficiency; The rock slag image processing in step S2 includes converting the collected rock slag color image into a gray image, that is, when the gray value is greater than the threshold value set in advance, it is 1, and less than the threshold value, it is 0, the connected region detection algorithm is used to judge the connected region, and the connected region is the position of the rock slag. According to the pixel points of the connected region, the geometric characteristics of the rock slag can be calculated. The tunneling parameter indexes recorded on site in step S3 include total thrust, cutter torque and cutter speed, the geometric information characteristics of rock slag include short axis, medium axis and long axis of rock slag, and the rock breaking efficiency is represented by specific energy.

2. The intelligent rock breaking efficiency prediction method based on rock fragment size of a TBM according to claim 1, characterized in that, The rock slag image shooting device in step S1 includes an industrial camera, an LED fill light and a data storage center. The industrial camera is arranged on both sides and above the TBM slag belt to shoot the photo information of the rock slag in real time. The LED fill light is used for light compensation of the rock slag on the slag belt. The data storage center is connected with the industrial camera through wireless signal, can real-time transmit the rock slag image obtained by shooting into the database, and then transmit the data of the database into the data storage center through wireless network.

3. The intelligent rock breaking efficiency prediction method based on rock fragment size of a TBM according to claim 1, characterized in that, The step S4 specifically includes: S4.1: normalize the collected sample data set, and divide it into training set and test set; S4.2: use the training set to build the GBRT model, initialize the parameters of GBRT, including the number of weak learners, the maximum tree depth and the learning rate; S4.3: initialize the related parameters of particle swarm algorithm, use particle swarm optimization algorithm to optimize the parameters of GBRT, and get the optimal number of weak learners, maximum tree depth and learning rate parameter value; S4.4: put the optimal parameters into the GBRT model to get the trained PSO-GBRT prediction model, put the test set sample data into the model to get the specific energy prediction value, and use R2 and RMSE indexes to test the prediction performance of the model.

4. The intelligent rock breaking efficiency prediction method based on rock fragment size of a TBM according to claim 3, characterized in that, The normalization processing formula of the data in step S4.1 is as follows: In the formula, Xnorm is the normalized value of the data, X is the original value of each parameter in the sample, Xmax and Xmin are the maximum and minimum values of each parameter in the sample respectively.

5. The intelligent rock breaking efficiency prediction method based on rock fragment size of a TBM according to claim 3, characterized in that, In step S4.3, the GBRT is optimized by using particle swarm algorithm as follows: S4.3.1 initialize the population size of particle swarm algorithm 20, the maximum iteration number 200, the learning factor c1 and c2 1.5, determine the value range of model hyperparameters weak learner number, maximum tree depth, learning rate, the value range of weak learner number is 1 to 1000, the value range of maximum tree depth is 1 to 5, the value range of learning rate is 0.01 to 1, and the initial values of these parameters are randomly obtained; S4.3.2 calculate the fitness value of the particle swarm individual, the fitness value is calculated by the mean square error between the sample predicted value and the sample true value, and the best fitness value in all particles is set as the global extreme value; S4.3.3 update the speed and position of the particle, and compare with the extreme value of the previous moment, and update the extreme value; S4.3.4 judge whether the pre-set maximum iteration number or minimum error condition is reached, if not satisfied, return to step S4.3.2, if the end condition is satisfied, stop iteration, and output the optimal solution of parameter weak learner number, maximum tree depth and learning rate.

Images