Method for predicting tunneling parameters of large-diameter slurry shield in volcanic ash stratum

Abstract

The invention relates to a volcanic ash stratum large-diameter slurry shield tunneling parameter prediction method, which comprises the following steps of collecting data of shield tunneling parameters of a slurry shield tunneling machine, and performing data processing on the collected data; performing data processing to obtain a reference variable, standardizing the data of the reference variable, and inputting the standardized data into a prediction model for training to obtain a corresponding optimal solution; the optimal solution comprises two hyper-parameters which respectively correspond to the number of decision trees in the random forest model and the layer number of each decision number; constructing a decision tree through the two hyper-parameters, training a random forest regression model according to the samples in the training set and the values of the target variables, and predicting the samples in the test set by using the trained random forest regression model; the random forest regression model averages the prediction result of each decision tree to obtain a final regression prediction result; and analyzing and evaluating the output prediction result.

Description

Technical Field

[0001] The present application relates to the field of shield tunneling technology, and in particular to a method for predicting parameters of large-diameter slurry shield tunneling in volcanic ash strata. Background Art

[0002] Volcanic ash strata refer to soil or rock layers formed by volcanic ash and other volcanic debris after certain physical, chemical and hydrological effects in geological history. They are quite different from other strata because of their cementing materials, shear strength and stability.

[0003] Normally, when the mud tank is full and the appropriate mud pressure is maintained, the surface settlement can be effectively controlled. However, in volcanic ash formations, the control effect of surface settlement is sometimes not ideal. This is because volcanic ash formations tend to turn into mud after contact with water, thus losing shear strength. Under the influence of the vibration generated by the shield machine during excavation, the original soil is easily destroyed, resulting in a sudden drop in soil strength and possible thixotropic settlement.

[0004] Under normal shield construction conditions, the cutting tool cuts the soil in front and mixes it with mud, which is then discharged through a mud pump. However, during shield excavation in volcanic ash formations, when the cutter removes the slag in front, the uncut soil behind may lose stability, causing collapse and blocking the slag discharge channel, which in turn causes the mud bin to become blocked, the slag discharge efficiency to decrease, and the cutter head torque to increase.

[0005] How to predict the parameters of large-diameter slurry shield tunneling in volcanic ash strata can avoid the risk of instability during shield tunneling. Summary of the invention

[0006] The purpose of the present application is to provide a method for predicting the excavation parameters of a large-diameter slurry shield in volcanic ash strata. The method for predicting the excavation parameters of a large-diameter slurry shield in volcanic ash strata can determine the target excavation parameters, adjust the parameters of the shield machine by the target excavation parameters, and determine the reference variables according to the determined target excavation parameters to simulate and predict the excavation speed. It can avoid the risk of instability in the shield excavation process to the greatest extent, and at the same time, through the random forest model based on lion group optimization, it ensures a high degree of prediction accuracy and fitting ability.

[0007] In order to achieve the above object, the present invention provides a method for predicting the excavation parameters of a large-diameter slurry shield in a volcanic ash stratum, comprising the following steps: Collect data on the tunneling parameters of the slurry shield machine and process the collected data; After data processing, reference variables are obtained, and the data of the reference variables are standardized and input into the prediction model for training, which is trained by the lion group algorithm; During the training process, as iterations continue, when the maximum number of iterations is reached, the global optimal fitness value and the global optimal position are obtained, that is, the corresponding optimal solution is obtained; The optimal solution includes two hyperparameters, corresponding to the number of decision trees and the number of layers per decision tree in the random forest model; Build a decision tree through two hyperparameters, train a random forest regression model based on the samples in the training set and the values ​​of the target variable, and use the trained random forest regression model to predict the samples in the test set; The random forest regression model will average the prediction results of each decision tree to obtain the final regression prediction result; Analyze and evaluate the output prediction results.

[0008] In an optional embodiment, data processing includes performing a correlation analysis on shield tunneling parameters, determining the tunneling parameter having the smallest linear relationship with other tunneling parameters through the correlation analysis, using the determined tunneling parameter as the target tunneling parameter, calculating the maximum information coefficient between the target tunneling parameter and other tunneling parameters, and determining the reference variable according to the shield tunneling process.

[0009] In an optional embodiment, the correlation analysis includes determining the data type, determining whether the data conforms to the normal distribution, using the Pearson correlation coefficient to analyze the correlation coefficient for parameters that conform to the normal distribution, and using the Spearman correlation coefficient to analyze the correlation coefficient for parameters that do not conform to the normal distribution.

[0010] In an optional implementation, based on the normal distribution test results of shield tunneling parameters, a corresponding correlation analysis method is selected to perform correlation analysis between the tunneling parameters of the corresponding formation section and the total shield section, and the tunneling speed is determined as the target tunneling parameter.

[0011] In an optional embodiment, the tunneling strategy is optimized based on the obtained correlation analysis results, including adjusting the parameters controlling the shield tunneling according to the obtained tunneling parameter variation patterns during the shield tunneling through the volcanic ash strata.

[0012] In an optional implementation, after using the box plot method to filter out some abnormal data of the shield tunneling parameters collected in real time, a certain number of tunneling parameters are selected as variable data. After data processing, the maximum information coefficient between the tunneling speed and other tunneling parameters is calculated, and the tunneling parameters with a maximum information coefficient greater than 0.4 are selected as reference variables.

[0013] In an optional implementation, the reference variable data after filtering out some abnormal data is standardized using Z-Score, the original data is linearly transformed to obtain the reference variable data for model input, a portion of the reference variable data is selected as a training set, and the remaining reference variable data is selected as a test set, the training set data is input into the prediction model, and training is performed using the lion group algorithm.

[0014] In an optional implementation, during the training process using the lion group algorithm, an initialization variable is first generated, an initial fitness value is calculated, the initial fitness value is a root mean square error, and it is assumed that the initial fitness value is a global optimal fitness value, and the initial position is a global optimal position; After each iteration, compare whether the fitness value of the current iteration is smaller than the previous global optimal fitness value. If it is smaller, update the global optimal fitness value to the current fitness value, and update the global optimal position to the current position. As the iterations proceed, if a smaller fitness value is found, the global optimal fitness value and position will be updated. Finally, when the maximum number of iterations of 30 is reached, the global optimal fitness value and the global optimal position are obtained.

[0015] In an optional implementation, in the prediction model, based on two hyperparameters, a group of 20 lions is set in a 2-dimensional space, and the positions of the lion king, lionesses and cubs are continuously updated iteratively.

[0016] In an optional embodiment, the model is evaluated by calculating the mean absolute error, mean square error, root mean square error and determination coefficient between the predicted value and the true value.

[0017] Through the method for predicting the excavation parameters of large-diameter slurry shield in volcanic ash formations in the present invention, the reference variables in the training model can be obtained through data processing, and then the reference variables are input into the random forest model based on lion group optimization to perform prediction analysis on the excavation speed with the minimum linear relationship between the excavation parameters, thereby obtaining the final regression prediction result, and by analyzing and evaluating the prediction model, the performance of the random forest model based on lion group optimization is understood, thereby providing a reliable guarantee for the prediction of the excavation speed of large-diameter slurry shield in volcanic ash formations.

[0018] Other features and advantages of the present application will be described in detail in the subsequent detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0020] Figure 1 This is a schematic diagram of the algorithm flow of the random forest model based on lion group optimization in this application; Figure 2 This is a schematic diagram of the results of the shield tunneling parameter correlation analysis in this application; Figure 3 This is the error analysis diagram of the random forest model based on lion group optimization in this application; Figure 4 Schematic diagram of the comparison between the predicted value and the true value of the random forest model based on lion group optimization in this application. DETAILED DESCRIPTION

[0021] In order to make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the technical scheme in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all of the embodiments. The components of the embodiments of the present application described and shown in the drawings here can be arranged and designed in various different configurations.

[0022] In the description of this application, it should be noted that the terms "inside", "outside", etc. indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, or the orientation or positional relationship in which the product of the application is usually placed when in use, which is only for the convenience of describing this application and simplifying the description, and does not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on this application. In addition, the terms "first", "second", etc. are only used to distinguish the description, and cannot be understood as indicating or implying relative importance.

[0023] In the description of this application, it should also be noted that, unless otherwise clearly specified and limited, the terms "disposed" and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a direct connection, or an indirect connection through an intermediate medium, or it can be the internal communication of two elements. For ordinary technicians in this field, the specific meanings of the above terms in this application can be understood according to specific circumstances.

[0024] The method for predicting the excavation parameters of a large-diameter slurry shield in volcanic ash formations in the present application mainly involves predicting the excavation parameters of a large-diameter slurry shield in volcanic ash formations, especially predicting the excavation speed. By processing the data before the prediction process, the excavation speed is determined as the target excavation parameter, and then the maximum information coefficient between the excavation speed and other excavation parameters is calculated, and the reference variable is determined in combination with the shield excavation process.

[0025] By inputting the determined reference variable data into the Lion Group Optimized Random Forest model, the excavation speed can be accurately predicted, thereby ensuring the stability of shield excavation in specific volcanic ash strata.

[0026] See also Figure 1 The method for predicting the excavation parameters of a large-diameter slurry shield in a volcanic ash stratum of the present invention comprises the following steps: Collect data on the tunneling parameters of the slurry shield machine and process the collected data; After data processing, reference variables are obtained, and the data of the reference variables are standardized and input into the prediction model for training, which is trained by the lion group algorithm; During the training process, as iterations continue, when the maximum number of iterations is reached, the global optimal fitness value and the global optimal position are obtained, that is, the corresponding optimal solution is obtained; The optimal solution includes two hyperparameters, corresponding to the number of decision trees and the number of layers per decision tree in the random forest model; Build a decision tree through two hyperparameters, train a random forest regression model based on the samples in the training set and the values ​​of the target variable, and use the trained random forest regression model to predict the samples in the test set; The random forest regression model will average the prediction results of each decision tree to obtain the final regression prediction result; Analyze and evaluate the output prediction results.

[0027] In the specific steps, it is first necessary to process the data of the shield tunneling parameters of the slurry shield machine collected in real time to determine the reference variables input into the prediction model, and at the same time obtain the target tunneling parameters with the smallest linear relationship between the tunneling parameters and other tunneling parameters. Taking this target tunneling parameter as a reference, the reference variable for simulating the target tunneling parameter is obtained through calculation.

[0028] That is to say, in the process of data processing, two concepts need to be clarified. One is the excavation speed with the smallest linear relationship between excavation parameters and other excavation parameters, and the excavation speed is used as the target excavation parameter to predict the excavation speed. Then, the excavation speed is used as a reference to obtain the reference variable used to simulate the excavation speed and input it into the excavation speed prediction model through calculation.

[0029] The above data processing process first includes correlation analysis of shield tunneling parameters. Correlation analysis can be used to quantify the degree of influence between research objects. The characteristic quantity is the correlation coefficient, represented by r, with a range of -1≤r≤1, where negative numbers represent negative correlation, positive numbers represent positive correlation, and 0 represents no correlation. The absolute value of the correlation coefficient |r| is used to evaluate the degree of correlation between variables, where |r|∈(0,0.2] is extremely weakly correlated, |r|∈(0.2,0.4] is weakly correlated, |r|∈(0.4,0.6] is moderately correlated, |r|∈(0.6,0.8] is strongly correlated, and |r|∈(0.8,1.0] is extremely strongly correlated.

[0030] The correlation coefficient analysis methods include Spearman correlation coefficient, Pearson correlation coefficient, etc. The Pearson correlation coefficient is generally used in statistics, but the conditions must be met: continuous data, normal distribution, and linear relationship. The Spearman correlation coefficient is a linear correlation analysis based on the rank size between the research objects. It has no requirements for the distribution of the analysis objects and has a wide adaptability, but its statistical efficiency is lower than that of the Pearson correlation. Therefore, in this application, the Pearson correlation coefficient is used for the collected parameters that conform to the normal distribution, and the Spearman correlation coefficient is used for the rest that do not conform to the normal distribution.

[0031] Before conducting a correlation analysis of shield tunneling parameters, it is necessary to determine whether the data obeys a normal distribution in order to select the appropriate correlation analysis method. Commonly used methods for testing whether the data obeys a normal distribution include numerical methods (SW method and KS method) and graphical methods (histogram, PP graph, QQ graph). Among them, the SW method test belongs to a small sample test, and the sample size is generally less than 50. The QQ graph test is mainly used to test the similarity of data distribution. Combined with the shield tunneling data, the KS numerical method and the PP graphical method are selected for normality test: observe whether the points on the PP graph are distributed on the theoretical straight line. When the corresponding p value of the KS test is greater than 0.05, it is considered that the data obeys a normal distribution. The normal distribution test results are shown in the following table:

[0032] Note: √- obeys normal distribution; ( )- normality test p-value.

[0033] According to the normal distribution test results of shield tunneling parameters, the corresponding correlation analysis method is selected to perform the correlation analysis between the tunneling parameters of the corresponding stratum section and the total shield section. The correlation results are shown in Figure 2 .according to Figure 2 The results of the shield tunneling parameter correlation analysis test were analyzed to optimize the tunneling strategy, including adjusting the parameters controlling the shield tunneling according to the obtained tunneling parameter variation law during the shield tunneling through the volcanic ash stratum, mainly including the following different angles: 1) Mud water pressure and grouting pressure In volcanic ash strata, the difference between grouting pressure and mud water pressure is concentrated in the range of 0.2~0.3MPa, and the grouting pressure is about 0.22MPa greater than the mud water pressure. Mud water pressure is positively correlated with tunnel depth. The grouting pressure should be controlled at about 0.22MPa greater than the mud water pressure, and secondly, it should take into account ground settlement and local uplift. For such shallow shield tunnels, it is necessary to monitor ground settlement and the safety of the upper building structure corresponding to the shield excavation position in a timely manner, so as to adjust the grouting pressure in a timely manner.

[0034] 2) Grouting pressure and grouting volume In volcanic ash formations, the grouting pressure is negatively correlated with the grouting volume. The rapid filling of the slurry ensures the relative stability of the soil. If the grouting pressure is found to be positively correlated with the grouting volume, the grouting equipment needs to be checked in time to ensure smooth grouting and safe shield tunneling.

[0035] 3) Total thrust and mud water pressure In volcanic ash formations, the total thrust is strongly negatively correlated with mud water pressure (correlation coefficient is -0.61), indicating that as mud water pressure increases, the total thrust setting must be reduced. This is particularly evident in the "volcanic ash debris + gravel" composite formation (correlation coefficient is -0.64).

[0036] 4) Excavation volume and total thrust In the volcanic ash strata, the amount of excavation is positively correlated with the total thrust. In the composite strata of "clay + volcanic ash debris + round gravel", the amount of excavation is moderately positively correlated with the total thrust (correlation coefficient is 0.54), which is much greater than that of the composite strata of "volcanic ash debris + round gravel" and "clay + volcanic ash debris".

[0037] 5) Cutter torque and total thrust In volcanic ash formations, the cutter head torque is negatively correlated with the total thrust. In the composite formation of "volcanic ash debris + round gravel", it is strongly negatively correlated (correlation coefficient is -0.77), but in the composite formation of "clay + volcanic ash debris + round gravel", it is weakly positively correlated. Clay in the formation is a great obstacle to shield excavation, and attention should be paid to the formation of mud cakes.

[0038] 6) Cutter speed and total thrust In volcanic ash formations, the cutterhead torque is very weakly correlated with the total thrust. The cutterhead speed is more concerned with the cutting effect of the soil, but the total thrust is also affected by factors such as the viscosity, density, and hardness of the soil. The increase in the cutterhead speed will increase the contact area between the cutterhead and the soil, thereby increasing the cutting force. However, in the actual excavation process, the change in the contact area between the cutterhead and the soil may be affected by other factors, such as the posture of the shield machine, the unevenness of the soil, etc., so the relationship between the cutterhead speed and the total thrust is not necessarily linear.

[0039] 7) Excavation speed and cutterhead speed In volcanic ash formations, the excavation speed is very weakly negatively correlated with the cutterhead speed. Within a certain range, the excavation speed is positively correlated with the cutterhead speed, because the increase in cutterhead speed will increase the tool's ability to cut the soil, thereby increasing the excavation speed to a certain extent. However, in volcanic ash formations, tool wear and formation resistance weaken this positive correlation.

[0040] It should be pointed out that during the normal shield tunneling process, the sensors on the shield machine collected a total of 36 types of raw data, the main parameters of which are mud water pressure, grouting pressure, grouting volume, excavation volume, total thrust, cutter head torque, cutter head speed and excavation speed. The shield tunneling parameters are collected in real time. Considering that the excavation parameter values ​​in the same ring do not change much, the average of the excavation parameters in the same ring under normal excavation conditions is taken as the representative of the ring. In order to ensure the reliability of the data, the box plot method is used to filter out some abnormal data, and finally the excavation parameters of a total of 555 rings are selected as variable data. Then, the data type of the filtered variable data is judged to be in accordance with the normal distribution and the correlation analysis is performed.

[0041] At the same time according to Figure 2 From the test results of the shield tunneling parameter correlation analysis, it can be seen that the linear correlation coefficient values ​​of the tunneling speed and other tunneling parameters are small, and the maximum correlation coefficient value is 0.46. It is difficult to determine the extent to which other tunneling parameters affect the tunneling speed through linear correlation analysis.

[0042] Therefore, by taking the excavation speed as the target excavation parameter, that is, by taking the excavation speed as the target excavation parameter and independent variable of the prediction model in this application, it can be shown that the excavation speed is a relatively independent control factor parameter in the excavation process of large-diameter slurry shield in volcanic ash strata. Through the predictive simulation of the excavation speed, its relative linear relationship with other variable parameters can be reduced, thereby improving the true accuracy of the predictive simulation.

[0043] There is a highly nonlinear relationship between the excavation parameter data, and the traditional linear analysis method cannot be used. Therefore, it is difficult to determine the reference variables in the excavation speed prediction simulation process.

[0044] In the present invention, the maximum information coefficient between the excavation speed and other excavation parameters is calculated, and the excavation parameter with a maximum information coefficient greater than 0.4 is selected as a reference variable.

[0045] The maximum information coefficient (MIC) is a measurement method used to detect nonlinear correlations between variables. It aims to overcome the limitations of linear correlation measurement methods such as the Pearson correlation coefficient when dealing with nonlinear relationships. The calculation of the maximum information coefficient is based on the principle of information theory. It measures the degree of association between two variables. The larger the value, the stronger the association. The value range of MIC is 0 to 1. When MIC is close to 1, it indicates that there is a strong nonlinear correlation between the two variables; when MIC is close to 0, it indicates that there is almost no nonlinear correlation between the two variables. The calculation formula of MIC is shown in equations (1) and (2).

[0046] ; Where: X, Y are variable names, N is the number of samples, and p(X,Y) is the joint probability of variables X and Y.

[0047] Based on the data obtained from data preprocessing, Python software was used to calculate the MIC value between the tunneling speed and the other seven tunneling parameters. The calculation results are shown in the table below. Taking the shield tunneling process into consideration, mud water pressure, excavation volume, grouting pressure, and cutter head torque were finally selected as reference variables:

[0048] In machine learning, data standardization or normalization enables comprehensive analysis and comparison of indicators of different units or orders of magnitude, thereby improving the accuracy and stability of the model. Z-Score standardization is used to perform linear transformation on the original data so that the result is mapped between [-1, 1]. The formula is shown in formula (3).

[0049] ; Where: x' is the standardized value, x is the original data value, μ is the mean of the sample data, and σ is the standard deviation of the sample data.

[0050] After linear transformation of the original data, preferably the 555 rings of tunneling parameter variable data after filtering abnormal data, the reference variable data for model input is obtained. 70% of the reference variable data are randomly selected from the 555 groups of data as training sets, and the remaining 30% of the reference variable data are used as test sets. The training set data are input into the prediction model and trained using the lion group algorithm.

[0051] The Lion Group Algorithm (LSO) is a global optimization strategy that simulates the social behavior of lions in nature. In this algorithm, lion groups are divided into three categories: lion kings, lionesses, and lion cubs. The lion king is the most powerful individual in the group, and the lionesses are responsible for reproduction and hunting. The behavior patterns of lion cubs vary according to their hunger state and growth stage: when they feel hungry, they tend to look for food near the lion king; after eating, they follow the lionesses to learn hunting skills; as they grow up, they may be expelled from the territory by the lion king and become stray lions. These stray lions may become new lion kings after a series of challenges.

[0052] Mathematically, the LSO algorithm starts searching from the initial solution space, and the solution with the best fitness value is regarded as the lion king. In the algorithm, a certain proportion of solutions are selected as hunting lions, and these solutions cooperate with each other during the search process. If a better solution is found, the position of the lion king is updated. The cubs imitate the hunting behavior of the lioness or look for food near the lion king. When the cubs grow to a certain stage, they will be "expelled" by the algorithm, and these expelled solutions will try to move closer to the best position in memory in order to survive. In short, the LSO algorithm realizes the search and optimization of the solution space by simulating the social structure and behavior of the lion group. Through this simulation, the algorithm can effectively explore and utilize the solution space to achieve the goal of global optimization.

[0053] Based on the subsequent combination with the random forest regression model, from the composition of the two hyperparameters, the space setting in the model is taken as 2D, that is, two hyperparameters. In the 2D space, there is a group of 20 lions, the number of adult lions is βN, β in the model is the proportional factor, which is 0.5. The remaining adult lions are lionesses, and the rest are cubs. The position updates of the three types of lions during the hunting process are shown in equations (4) to (9).

[0054] Lion King Position: ; Where: is the optimal position of the kth generation population, is the historical optimal position of the i-th lion in the k-th generation, and γ is a random number that follows the normal distribution N[0,1].

[0055] Lioness Position: ; Where: is the historical optimal position of a randomly selected partner from the kth generation of lioness, is the disturbance factor of the lioness, H and h represent the minimum mean and maximum mean of each dimension of the lion activity space, T is the maximum number of iterations, and t is the current number of iterations.

[0056] Young Lions Position: ; Where: is the position where the i-th lion cub is driven away within the hunting range, It is the best historical position of the kth generation for the lion cub to follow the lioness. is the lion cub perturbation factor, and q is a random number generated by uniform distribution U[0,1].

[0057] First, the built-in Rand function of Matlab software is used to generate initialization variables and calculate the initial fitness value. The fitness value in the model is taken as the root mean square error RMSE, and it is assumed that the initial fitness value is the global optimal fitness value and the initial position is the global optimal position. After each iteration, compare whether the fitness value of the current iteration is smaller than the previous global optimal fitness value. If it is smaller, update the global optimal fitness value to the current fitness value, and update the global optimal position to the current position. As the iteration proceeds, if a smaller fitness value is found, the global optimal fitness value and position will be updated. Finally, when the maximum number of iterations is reached, after 30 iterations in the model, the global optimal fitness value and the global optimal position are obtained, and the corresponding optimal solution is obtained, that is, two hyperparameters, corresponding to the number of decision trees in the random forest model and the number of layers of each decision tree.

[0058] The decision tree is constructed by using the two hyperparameters obtained above, specifically, the number of decision trees n-trees = 4.482, and the number of decision tree layers n-layers = 4.494. The random forest regression model is trained based on the samples in the training set and the values ​​of the target variable. The trained random forest regression model is used to predict the samples in the test set. The model averages the prediction results of each decision tree to obtain the final regression prediction result. The random forest regression is shown in formula (10), and the flowchart is shown in Figure 1 .

[0059] ; Where: is the corresponding output generated in the decision tree; k is the number of decisions.

[0060] By using the lion group optimization-based random forest model (LSO-RF) in the present invention to predict and analyze the excavation speed, the excavation speed of large-diameter slurry shield in volcanic ash strata can be truly predicted, thereby avoiding the risk of instability during shield excavation.

[0061] In the present invention, the excavation speed is determined as the target excavation parameter of the prediction model because the excavation speed plays a role in the excavation of a large-diameter slurry shield in volcanic ash strata and the linear relationship is minimal.

[0062] The prediction of the tunneling speed of large-diameter slurry shield in volcanic ash formations solves many problems, including geological risk management, engineering planning and progress control, resource optimization, equipment performance maximization, real-time risk management, and cost control. By accurately predicting the tunneling speed, the uncertainty caused by geological conditions can be effectively reduced, ensuring the geological robustness of the project.

[0063] At the same time, the prediction of the tunneling speed of large-diameter slurry shield in volcanic ash strata covers multiple technical fields such as geological engineering, data science and statistics, underground engineering and tunnel engineering technology, mechanical engineering, real-time monitoring technology, and computer simulation and emulation. By understanding geological characteristics, using large amounts of data and applying statistical methods, combining underground engineering principles, optimizing mechanical equipment performance and real-time monitoring technology, the prediction model can improve the accuracy of the tunneling speed of shield in volcanic ash strata, thereby helping to formulate reasonable engineering plans and schedules in engineering practice, optimize resource allocation, improve mechanical equipment performance, respond to stratum changes in real time, reduce risks, and ultimately ensure the efficient, safe and economical progress of the project.

[0064] The main body of the present invention solves the problem of predicting the excavation speed of large-diameter slurry shields in volcanic ash strata. First, by deeply studying the shield excavation process in volcanic ash strata, the innovation and adaptability of shield technology are promoted, making it more flexible and diverse. Secondly, predicting the shield excavation speed helps to identify potential risks in advance, so as to better manage risks and ensure the safe progress of shield projects. In addition, the prediction results of the excavation speed help to optimize project planning and resource allocation, reduce costs, improve efficiency, and promote the rational development and utilization of local resources.

[0065] The shield tunneling parameter prediction method in this application is mainly used in volcanic ash strata, where the geological conditions are complex and their characteristics may change rapidly within a small range. By predicting the tunneling speed, we can better understand the characteristics of the strata, thereby reducing geological risks and ensuring the safety of the project.

[0066] The heterogeneity of volcanic ash formations can have an impact on the rate of advancement, so accurate predictions of advancement rates help establish a feasible engineering plan and schedule to ensure that the project is completed on time.

[0067] Different areas of the volcanic ash formation may require different engineering parameters and resource allocation. By predicting the excavation speed, the allocation of manpower, materials and mechanical equipment can be optimized to improve resource utilization efficiency.

[0068] The particularity of volcanic ash strata may have an impact on the performance of shield machines. By predicting the excavation speed, the working parameters of the mechanical equipment can be adjusted to maximize its performance and reduce the risk of equipment failure.

[0069] Predicting the tunneling speed can help the engineering team monitor the changes in the strata in real time and make timely adjustments to deal with possible problems, thereby reducing engineering risks. Accurate tunneling speed prediction can better control engineering costs, avoid unnecessary expenses, and improve overall economic benefits.

[0070] In general, the prediction of the excavation speed of large-diameter slurry shields in volcanic ash formations can help reduce geological and engineering risks, optimize project planning and resource allocation, and improve equipment performance, thereby ensuring the efficient, safe and economical progress of the project.

[0071] After predicting the tunneling speed using the lion group optimization-based random forest model (LSO-RF), the model was evaluated by calculating the mean absolute error, mean square error, root mean square error, and determination coefficient between the predicted value and the true value.

[0072] Specifically, in the evaluation of the prediction results, Matlab software was used for calculation, and the error analysis of the LSO-RF prediction model is shown in Figure 3 , the predicted value and true value of the algorithm are shown in Figure 4 , the evaluation indicators are shown in the following table:

[0073] According to the results, the mean absolute error (MAE) is 0.142, which is a small error, indicating that the model has good accuracy in prediction. The mean square error (MSE) is 0.063, which is also a small error, indicating that the gap between the model's predicted value and the true value is small. The root mean square error (RMSE) is 0.251, which is used to measure the standard deviation of the model's prediction error. The error is also relatively small, indicating that the model's prediction has a small discreteness. The coefficient of determination (R2) is 0.895, which is close to 1, indicating that the model can well explain the variance of the target variable, indicating that the model fits the data very well. Overall, the LSO-RF model performs very well on the test set, with a high degree of prediction accuracy and fitting ability.

[0074] The method for predicting the parameters of large-diameter slurry shield tunneling in volcanic ash strata in the present invention is specifically applied to the shield tunneling excavation of the No. 1 tunnel project of the Jakarta-Bandung High-speed Railway in Indonesia. The entrance mileage of the No. 1 tunnel of the Jakarta-Bandung High-speed Railway in Indonesia is DK2+540.0, the exit mileage is DK4+410.0, the total length is 1885m, and the maximum buried depth of the tunnel is about 36m. The tunnel is a single-hole double-track tunnel, DK2+771.5~DK4+238.5 is the shield range, about 1467m, and the buried depth of the top surface of the inner rail is about 5.0~24.0m.

[0075] The geology of the tunnel is mainly composed of clay, volcanic ash cemented gravel, sandy soil and fine gravel soil. At the same time, the groundwater level is high, with a groundwater depth of 0.3 to 7.8 meters and a water level fluctuation of 2 to 3 meters. The surface water resources are abundant. The tunnel also needs to pass through volcanic accumulation layers, with broken rock structures and poor stability. The strength of mudstone is low and severely weathered. It also needs to pass under densely built (structured) areas, highways and their interchange ramps, etc.

[0076] In view of the extremely complex geological environment and risks of Tunnel No. 1, the excavation strategy can be adjusted by determining the target excavation parameters, and the excavation speed can be predicted by the random forest model based on lion group optimization, which has a high degree of prediction accuracy and fitting ability.

[0077] It should be noted that, in the absence of conflict, the features in the embodiments of this application may be combined with each other.

[0078] The above description is only the preferred embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method for predicting parameters of large-diameter slurry shield tunneling in volcanic ash strata, characterized in that: The following steps are involved: Collect data on the tunneling parameters of the slurry shield machine and process the collected data; After data processing, reference variables are obtained, and the data of the reference variables are standardized and input into the prediction model for training, which is trained by the lion group algorithm; During the training process, as iterations continue, when the maximum number of iterations is reached, the global optimal fitness value and the global optimal position are obtained, that is, the corresponding optimal solution is obtained; The optimal solution includes two hyperparameters, corresponding to the number of decision trees and the number of layers per decision tree in the random forest model; Build a decision tree through two hyperparameters, train a random forest regression model based on the samples in the training set and the values ​​of the target variable, and use the trained random forest regression model to predict the samples in the test set; The random forest regression model will average the prediction results of each decision tree to obtain the final regression prediction result; Analyze and evaluate the output prediction results.

2. The method for predicting parameters of large-diameter slurry shield tunneling in volcanic ash strata according to claim 1 is characterized in that: Data processing includes correlation analysis of shield tunneling parameters. The tunneling parameter with the smallest linear relationship with other tunneling parameters is determined through correlation analysis. The determined tunneling parameter is used as the target tunneling parameter, and the maximum information coefficient between the target tunneling parameter and other tunneling parameters is calculated. According to the shield tunneling process, the reference variable is determined.

3. The method for predicting parameters of large-diameter slurry shield tunneling in volcanic ash strata according to claim 2 is characterized in that: Correlation analysis includes determining the data type and whether the data conforms to the normal distribution. The Pearson correlation coefficient is used to analyze the correlation coefficient for parameters that conform to the normal distribution, and the Spearman correlation coefficient is used to analyze the correlation coefficient for parameters that do not conform to the normal distribution.

4. The method for predicting parameters of large-diameter slurry shield tunneling in volcanic ash strata according to claim 3 is characterized in that: According to the normal distribution test results of shield tunneling parameters, the corresponding correlation analysis method is selected to carry out the correlation analysis between the tunneling parameters of the corresponding stratum section and the total shield section, and the tunneling speed is determined as the target tunneling parameter.

5. The method for predicting parameters of large-diameter slurry shield tunneling in volcanic ash strata according to claim 4 is characterized in that: Based on the correlation analysis results, the tunneling strategy is optimized, including adjusting the parameters controlling the shield tunneling according to the obtained law of tunneling parameter changes during the shield tunneling through the volcanic ash strata.

6. The method for predicting parameters of large-diameter slurry shield tunneling in volcanic ash strata according to claim 4, characterized in that: After using the box plot method to filter out some abnormal data of the shield tunneling parameters collected in real time, a certain number of tunneling parameters are selected as variable data. After data processing, the maximum information coefficient between the tunneling speed and other tunneling parameters is calculated, and the tunneling parameters with a maximum information coefficient greater than 0.4 are selected as reference variables.

7. The method for predicting parameters of large-diameter slurry shield tunneling in volcanic ash strata according to claim 1 is characterized in that: The reference variable data after filtering out some abnormal data are standardized by Z-Score, and the reference variable data for model input are obtained after linear transformation of the original data. A certain proportion of the reference variable data are selected as the training set, and the remaining proportion of the reference variable data are selected as the test set. The training set data are input into the prediction model and trained by the lion group algorithm.

8. The method for predicting parameters of large-diameter slurry shield tunneling in volcanic ash strata according to claim 7, characterized in that: During the training process of the lion group algorithm, the initialization variables are first generated, and the initial fitness value is calculated. The initial fitness value is the root mean square error, and it is assumed that the initial fitness value is the global optimal fitness value and the initial position is the global optimal position; After each iteration, compare whether the fitness value of the current iteration is smaller than the previous global optimal fitness value. If it is smaller, update the global optimal fitness value to the current fitness value, and update the global optimal position to the current position. As the iterations proceed, if a smaller fitness value is found, the global optimal fitness value and position will be updated. Finally, when the maximum number of iterations of 30 is reached, the global optimal fitness value and the global optimal position are obtained.

9. The method for predicting parameters of large-diameter slurry shield tunneling in volcanic ash strata according to claim 7, characterized in that: In the prediction model, based on two hyperparameters, a group of 20 lions is set in a 2-dimensional space, and the positions of the lion king, lionesses and cubs are continuously updated iteratively.

10. The method for predicting parameters of large-diameter slurry shield tunneling in volcanic ash strata according to claim 1, characterized in that: The model was evaluated by calculating the mean absolute error, mean square error, root mean square error and determination coefficient between the predicted value and the true value.

Images