Geological tunnability classification method based on tunneling energy efficiency ratio

Abstract

The invention relates to the technical field of tunnel boring machine construction, in particular to a geological tunnability classification method based on a tunneling energy efficiency ratio, which utilizes geological data extraction and conversion software to realize conversion from a geological exploration report to a geological detail table, displays a geological longitudinal section diagram in a data form, completes conversion from a geological exploration diagram to digital geology, and realizes classification of the tunnability of the geological exploration diagram. According to a three-dimensional visualization result of distribution of four characteristic parameters of an upper quartile, a median, a lower quartile and a quartile distance of the shield tunneling energy efficiency ratio, four statistical characteristics of tunneling energy efficiency ratios of six inter-cluster regions have obvious boundaries, and it is indicated that shield geologic characteristics have distinguishability.

Description

Technical Field

[0001] The invention relates to the technical field of tunnel boring machine construction, and in particular to a geological excavability classification method based on excavation energy efficiency ratio. Background Art

[0002] The classification of surrounding rock is mainly based on the geological exploration data before design. Through the design exploration data, it is found that there are many complex geological conditions in the tunnel. However, it is difficult for tunnel boring machines to adapt to complex and changeable geological conditions. Inaccurate judgment of rock mass will lead to disasters such as machine jams and rock bursts, thus affecting the excavation efficiency. In order to scientifically select construction methods, reasonably arrange construction periods, save construction costs and avoid construction disasters, the classification of geological information on the excavability of tunnel boring machines has important research significance and application value.

[0003] The classification methods of tunnel surrounding rock at home and abroad still use traditional classification methods, such as BQ method, HC method, RMR method, etc., and most of the research is a small amount of data from a single project. The relationship between the tunnel boring machine excavation performance parameters and rock mass parameters is ignored, resulting in the traditional surrounding rock classification not being well matched with the excavation performance, reducing the excavation efficiency of the tunnel boring machine. Therefore, the rock mass parameters and excavation parameters should be comprehensively considered, and a geological excavability classification method based on the excavation energy efficiency ratio is designed for the excavation and adaptability of the tunnel boring machine. Summary of the invention

[0004] The purpose of the present invention is to provide a geological excavability classification method based on excavation energy efficiency ratio in order to solve the above problems.

[0005] The present invention achieves the above-mentioned purpose through the following technical solutions:

[0006] A geological excavability classification method based on excavation energy efficiency ratio is designed, which includes the following steps:

[0007] (1) Geological digitization

[0008] The geological data extraction and conversion software is used to realize geological digital transformation, which includes five steps: 1) geological longitudinal section vector graphics processing, retaining the stratigraphic boundary line, removing redundant graphics, and closing the stratigraphic area; 2) stratigraphic area filling, filling each stratigraphic area with a fill pattern of a specified color; 3) drawing reference lines, drawing positioning reference lines for stratigraphic identification based on the actual working conditions of the full-section tunnel boring machine; 4) stratigraphic identification, using the reference lines as the positioning basis, using VBA vector graphics recognition technology to identify the various stratigraphic layers that the tunnel passes through in the geological longitudinal section; 5) multi-source geological information fusion, fusing the multi-source geological information with the corresponding stratigraphic layers.

[0009] (2) Data selection and preprocessing

[0010] More than 350,000 rings of excavation data from more than 166 lines were selected. By analyzing the data characteristics of tunnel boring machine excavation parameters, effective excavation sections were extracted based on the change point detection method. Outliers were screened out using box plots and moving median filtering, and the data was smoothed to reduce the interference caused by random fluctuations and improve data consistency and analysis accuracy. In addition, in order to improve data processing efficiency and simplify the data analysis process, the original samples were aggregated from "excavation sections" to "excavation rings".

[0011] (3) Selection of digging index

[0012] Based on engineering experience, penetration rate, excavation speed, cutter head speed and excavation energy efficiency ratio are preliminarily selected as indicators for geological excavability classification. In order to avoid the redundancy of the selected excavability indicators, the Pearson correlation coefficient is used to measure the linear correlation between excavation parameters. The Pearson correlation coefficient of two variables is defined as the quotient of the product of the covariance and standard deviation between the two variables. The formula is:

[0013]

[0014] Among them, ρ xy That is, it is the Pearson correlation covariance of the population of variables X and Y; COV(x, y) is the population covariance of variables X and Y; μ and σ are the mean and variance of the variables respectively.

[0015] The Pearson correlation coefficient has a value range between 1 and -1. Variables close to 0 are considered to have no correlation, and those close to 1 or -1 are considered to have strong correlation. Through analysis, it was found that the Pearson coefficients of the tunneling energy efficiency ratio and the tunneling speed, cutterhead speed, penetration, TPI and FPI all exceeded 0.9. At the same time, the tunneling energy efficiency ratio includes parameters such as tunneling speed, cutterhead speed and cutterhead thrust. Therefore, the comprehensive parameter tunneling energy efficiency ratio is selected as an indicator of excavability classification. Its specific definition is: From the perspective of energy conservation, the work consumed by the tunnel boring machine for excavating and drilling into the rock and soil body should be equivalent to the energy absorbed by the corresponding rock and soil body to overcome its own strength or stability to form a broken body, and the tunnel boring machine's excavation of the rock and soil body is mainly completed by the thrust system jacking and the cutterhead rotation cutting and crushing. Therefore, the energy consumption of excavating a unit of rock and soil is as follows, which is called the tunneling energy efficiency ratio:

[0016]

[0017] Where: T represents the cutter head torque (MN·m); D is the diameter of the tunnel boring machine (m); other parameters are the same as above. The shield tunneling specific energy characterizes the energy consumed by the shield tunneling per unit volume of rock and soil, which is essentially an energy index that indicates the excavability of the stratum.

[0018] (4) Geological excavability classification

[0019] The excavation energy efficiency ratio and its statistical characteristic indicators were integrated, and the C-means clustering algorithm was used to classify the excavability, which was finally divided into 11 categories. Among them, there are 6 categories of shield machines, namely easy to excavate, excavable I, excavable II, difficult to excavate I, difficult to excavate II and unexcavable; there are 5 categories of TBMs, namely easy to excavate, excavable, difficult to excavate I, difficult to excavate II and unexcavable. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 is a table of geological details in the embodiment;

[0021] Figure 2 is a graph of the original excavation section in the embodiment;

[0022] Figure 3 is a graph of effective excavation section in the embodiment;

[0023] Figure 4 It is a schematic diagram of a box diagram structure in the embodiment;

[0024] Figure 5 is a data denoising graph based on median filtering in an embodiment;

[0025] Figure 6 is a correlation coefficient matrix diagram of 8 groups of excavation parameters in the embodiment;

[0026] Figure 7 : It is a visualization diagram of the six-category clustering results of shield machines in the embodiment; (a) distribution of the upper quartile, median and lower quartile of tunneling energy efficiency ratio under each category; (b) distribution of the upper quartile, median and interquartile range of tunneling energy efficiency ratio under each category; (c) distribution of the upper quartile, lower quartile and interquartile range of tunneling energy efficiency ratio under each category; (d) distribution of the median, lower quartile and interquartile range of tunneling energy efficiency ratio under each category.

[0027] Figure 8 : is a clustering result diagram of six types of shield tunneling energy efficiency ratio characteristics in the embodiment;

[0028] Fig. 9 : It is a visualization diagram of the five-category clustering results of TBM in the embodiment; (a) distribution of the upper quartile, median and lower quartile of the tunneling energy efficiency ratio under each category; (b) distribution of the upper quartile, median and interquartile range of the tunneling energy efficiency ratio under each category; (c) distribution of the upper quartile, lower quartile and interquartile range of the tunneling energy efficiency ratio under each category; (d) distribution of the median, lower quartile and interquartile range of the tunneling energy efficiency ratio under each category.

[0029] Fig.10 It is a clustering result diagram of the five types of TBM excavation energy efficiency ratio characteristics in the embodiment.

[0030] Beneficial technical effects

[0031] The geological data extraction and conversion software is used to realize the conversion of geological survey reports into geological detail tables, and the geological longitudinal section map is displayed in the form of data, completing the conversion of geological exploration maps into digital geology. The correct selection of excavation parameters is conducive to improving the accuracy and stability of classification; before data mining, erroneous data and abnormal data in the collected data should be eliminated to ensure the rationality of the results of subsequent data processing and data analysis; extraction of excavation sections and screening of outliers cannot completely solve the problem. In order to increase robustness and adaptability, the data is filtered and denoised on the basis of the current division to reduce the fluctuation trend of the data; the three-dimensional visualization results of the distribution of the four characteristic parameters of the upper quartile, median, lower quartile and interquartile distance of the shield tunneling energy efficiency ratio, the four statistical characteristics of the excavation energy efficiency ratio in the six clusters have obvious boundaries, indicating that the shield geological characteristics are distinguishable. DETAILED DESCRIPTION

[0032] The specific implementation modes of the present invention are described below in conjunction with examples. However, the following examples are only used to illustrate the present invention in detail and are not intended to limit the scope of the present invention in any way.

[0033] Example 1

[0034] Taking a large number of tunneling parameters in 166 lines as an example, the data comes from the shield / TBM engineering big data platform. The geological digging feasibility classification method based on tunneling parameters in the present invention is used to classify the geology, and the specific steps are as follows:

[0035] (1) Geological digitization

[0036] Use geological data extraction and conversion software to transform geological survey reports into geological detail tables, display geological longitudinal sections in the form of data, and complete the transformation of geological survey maps into digital geology, such as Figure 1 shown.

[0037] (2) Data selection and preprocessing

[0038] 1) Data selection

[0039] The tunneling process of a tunnel boring machine is a complex rock-machine interaction process. Correct selection of tunneling parameters is conducive to improving the accuracy and stability of classification. In order to avoid a large amount of invalid data in the extracted lines, the lines with a ratio of 0≤cutter head torque≤100 greater than 45% or a ratio of cutter head torque≤0 greater than 30% were eliminated, and data from 166 lines of 95 projects were obtained. Detailed data are shown in Tables 1 and 2:

[0040] Table 1 Details of route data extracted in this project

[0041]

[0042] Table 2 Route data statistics by TBM type

[0043]

[0044] 2) Data preprocessing

[0045] Extracting effective excavation sections: During the actual excavation process, the tunnel boring machine data acquisition system records the operating status of the tunnel boring machine, with values ​​of 0 (off state), 1 (standby state), and 2 (operating state). Figure 2 As shown in the figure, during the excavation of one ring, the tunnel boring machine will have multiple standby or stop states in the middle (such as tool replacement, abnormal shutdown, shoe step change, etc.). When the tunnel boring machine is in standby or shutdown state, the tunnel boring machine does not excavate and the total thrust is 0MN. When pre-processing the tunnel boring machine operating parameters, it is necessary to remove the sample points where the tunnel boring machine operating state is 0 or 1. The cyclic excavation process after removing the abnormal state data, that is, the excavation section, can be divided into empty inference, rising section, stable section and descending section, as shown in Figure 3 shown.

[0046] Outlier processing: Due to the complexity and variability of geological conditions, sensor anomalies, and electromagnetic interference during tunnel excavation, many outliers may appear in a tunneling cycle, usually manifested as abnormally large or abnormally small excavation parameters, and these data lack rationality. In order to ensure the accuracy and rationality of the results of data processing and analysis, before data mining, the erroneous data and abnormal data in the collected data should be eliminated to ensure the rationality of the results of subsequent data processing and data analysis. The box plot method is used to identify outliers. The drawing of the box plot relies on actual data. There is no need to assume in advance that the data obeys a specific distribution form. There is no restrictive requirement for the data. It only truly and intuitively shows the original appearance of the data shape; on the other hand, the standard for judging outliers in the box plot is based on quartiles and interquartile ranges. Quartiles have a certain resistance. Up to 25% of the data can become arbitrarily far without greatly disturbing the quartiles, so outliers cannot exert an impact on this standard. The results of identifying outliers by the box plot are relatively objective. The distribution of the box plot is described as follows: Figure 4As shown. The first quartile (Q1), also known as the "smaller quartile", is equal to the 25th percentile of all the values ​​in the sample arranged from small to large. The second quartile (Q2), also known as the "median", is equal to the 50th percentile of all the values ​​in the sample arranged from small to large. The third quartile (Q3), also known as the "larger quartile", is equal to the 75th percentile of all the values ​​in the sample arranged from small to large. The difference between the third quartile and the first quartile is also called the interquartile range (InterQuartile Range, IQR), as shown Figure 4 shown.

[0047] Data denoising: Data fluctuations are also affected by geological conditions and the driver's operating habits. Therefore, only extracting the excavation section and filtering out outliers cannot completely solve the problem. In order to increase robustness and adaptability, the data is filtered and denoised on the basis of the current division to reduce the fluctuation trend of the data. The median filter algorithm is used for data denoising. The median filter is a nonlinear smoothing filter technology that replaces the value of a point in the time series with the median of the values ​​of each point in a neighborhood of the point. Since it only needs to sort the odd number of data from a sampling window in the sequence in actual operation, and does not need to count the points in the sequence, the idea is simple and the calculation is convenient, which is suitable for the massive data processing of this project. Figure 5 The data denoising results of four operating parameters of the 325th to 329th rings of the second phase of Guangzhou Metro Line 7 are shown.

[0048] Data aggregation: After outlier and noise reduction processing, short periods with a stable segment duration of less than 60 seconds were screened out, and a database containing 382,799 normal excavation stable segments was established. Considering the characteristics of large volume, high diversity, high real-time and multiple outliers of TBM big data, statistical features such as upper quartile (Q3), lower quartile (Q1), median (Median) and interquartile range (IQR) were used to aggregate the "excavation segments" in the original data into "excavation rings", that is, the statistical information of the stable segments in all excavation cycles within a certain ring was used as the input feature of clustering. Such processing can greatly reduce the amount of data while ensuring data quality, avoid the distortion of analysis results by outliers, accurately identify patterns and structures in the data, and thus simplify the analysis process, improve analysis efficiency, and enhance the stability, reliability and interpretability of analysis results.

[0049] Based on quartile aggregation, a total of 113,436 groups of samples (tunneling rings) were obtained, among which the number of samples of earth pressure balance shield machines accounted for the largest proportion, 76.6%, while single soil, single rock and composite rock and soil accounted for 75.7%, 16.7% and 7.6% respectively.

[0050] Table 3 Aggregate sample data statistics by TBM type

[0051]

[0052] Table 4 Statistics of aggregated sample data classified by formation type

[0053]

[0054] (3) Selection of digging index

[0055] The Pearson correlation coefficient matrix between tunneling parameters is as follows: Figure 6 As shown in the figure, the Pearson coefficients of tunneling energy efficiency ratio and tunneling speed, cutter head speed, penetration, TPI and FPI are all over 0.9, indicating that the parameters contain repeated redundant information, that is, they are highly correlated.

[0056] (4) Geological excavability classification

[0057] 1) C-means clustering algorithm process

[0058] Preparation: In the actual shield tunneling process, the operating parameters of the tunnel boring machine are continuously collected by the data acquisition system. The operating parameter data is streaming data, and it is difficult to obtain the maximum and minimum values ​​of the parameters of all sample points. Therefore, the traditional MinMax normalization method cannot be used to process the shield sensor data. In order to avoid the influence of the different ranges of different types of shield tunneling parameters on data mining and machine learning models, the following formula is used to normalize the operating parameters of the tunnel boring machine:

[0059]

[0060] Where: P o is the original value of the operating parameter; P s is the rated value of the operating parameter; P is the normalized operating parameter.

[0061] Taking into account the characteristics of shield big data, in order to fully explore the essential characteristics and potential information in the original data, the tunneling energy efficiency ratio is selected as the input feature of the clustering model, and its quartiles and median are used to describe the central trend and distribution of the data to avoid the distortion of the analysis results by outliers. The interquartile range is used to standardize the data so that data from different strata and different types of tunnel boring machines can be compared on the same scale.

[0062] Hyperparameter confirmation and algorithm initialization: The existing tunnel surrounding rock classification takes rock mass as the research object and divides the surrounding rock categories according to the physical and mechanical properties of rock mass. The stratum state encountered by the tunnel boring machine when excavating in the tunnel is complex and changeable. If the tunnel boring machine excavation strategy is adjusted only based on the traditional surrounding rock classification, it will cause a large error; if the stratum is divided only according to typical unfavorable geology, it will also affect the excavation efficiency of the tunnel boring machine. Therefore, combined with the needs, expert knowledge and engineering experience, it is confirmed that the surrounding rock classification is 11 levels, including 6 types of shield and 5 types of TBM, that is, the number of clusters of shield K = 6, and the number of clusters of TBM K = 5. Initialization stop threshold ε = 1e-4, fuzzy factor m = 2, maximum number of iterations T = 300.

[0063] 2) Clustering results

[0064] The shield tunneling energy efficiency ratio is divided into six categories. The three-dimensional visualization results of the distribution of the four characteristic parameters of the upper quartile, median, lower quartile and interquartile distance are shown in Figure 2. Figure 7 As shown, the four statistical characteristics of the tunneling energy efficiency ratio in the six clusters have obvious boundaries, indicating that the geological characteristics of the shield tunnel are distinguishable. Support.

[0065] The classification method of the present invention comprehensively considers the rock mass parameters and tunneling parameters, and selects multiple indicators such as penetration, tunneling speed, cutter head speed and tunneling energy efficiency ratio, and uses the Pearson correlation coefficient for in-depth analysis, and finally determines the tunneling energy efficiency ratio that comprehensively reflects multiple factors as the key excavability classification indicator. This comprehensive consideration method enables the classification results to more accurately reflect the actual tunneling situation and achieve a better match with the working performance of the tunneling machine, which is expected to greatly improve the tunneling efficiency. In the data processing link, the existing technology lacks targeted and effective processing means. This method shows obvious advantages. First, in terms of geological digitization, professional software is used to realize a series of fine operations such as geological longitudinal section vector graphics processing, stratigraphic area filling, reference line drawing, stratigraphic identification, and multi-source geological information fusion, so that the geological information is comprehensively and accurately converted into analyzable data form, laying a solid foundation for subsequent classification. Secondly, the data selection and preprocessing process is rigorous and scientific. After selecting a large number of rich excavation data of lines, the effective excavation section is extracted based on the change point detection method, and the outliers are screened out and the data is smoothed using the box plot and moving median filter. The data aggregation processing from "excavation section" to "excavation ring" is also performed. These steps effectively reduce the interference of random fluctuations, abnormal data, etc. on the analysis results, improve the consistency of data and the accuracy of analysis, and make the final classification results more reliable. The present invention adopts the C-means clustering algorithm to classify excavability, integrates the excavation energy efficiency ratio and its statistical characteristic indicators, and finally subdivides the geological excavability into 11 categories, including 6 categories of shield and 5 categories of TBM, covering different degrees of categories from easy to excavate to unexcavable. This fine classification can more accurately reflect the differences in excavability under different geological conditions, and provide a very accurate basis for the construction party to reasonably select construction methods, arrange construction periods, save construction costs and avoid construction disasters according to specific geological conditions. In the data processing process, through reasonable data aggregation and other methods, the data quality is guaranteed while the data volume is greatly reduced, the distortion of the analysis results by outliers is avoided, and the patterns and structures in the data are accurately identified. This makes the method not only suitable for current engineering cases, but also has the potential to be promoted and applied in a wider range of tunnel excavation projects, and can better cope with the increasingly complex tunnel construction needs in the future.

[0066] Figure 7 Visualization of the clustering results of shield tunneling in six categories. (a) The distribution of the upper quartile, median and lower quartile of tunneling energy efficiency ratio in each category; (b) The distribution of the upper quartile, median and interquartile range of tunneling energy efficiency ratio in each category; (c) The distribution of the upper quartile, lower quartile and interquartile range of tunneling energy efficiency ratio in each category; (d) The distribution of the median, lower quartile and interquartile range of tunneling energy efficiency ratio in each category.

[0067] like Figure 8As shown in the box plot, the median of the tunneling energy efficiency ratio shows good regularity and distinguishability in the order of cluster 0, cluster 5, cluster 3, cluster 2, cluster 1, and cluster 4, which can better reflect the excavability of the shield rock mass, showing a gradual increase from easy to unexcavable, indicating that the energy consumed by the shield tunneling per unit volume of rock and soil increases successively, and the excavability becomes worse and worse. The increasing interquartile range of the tunneling energy efficiency ratio also means the increase of control difficulty and uncertainty of operation parameters. Therefore, combined with engineering experience, cluster 0, cluster 5, cluster 3, cluster 2, cluster 1 and cluster 4, 6 CM categories are defined as easy to excavate, excavable I, excavable II, difficult to excavate I, difficult to excavate II and unexcavable, and the excavability of the shield stratum can be divided into 6 levels.

[0068] TBM5 clustering results: The three-dimensional visualization results of the distribution of the four characteristic parameters of the upper quartile, median, lower quartile and interquartile distance of the TBM excavation energy efficiency ratio are shown in Figure 2. Fig. 9 As shown, the four statistical characteristics of the tunneling energy efficiency ratio in the five cluster areas have obvious boundaries, indicating that the TBM geological characteristics are distinguishable. Fig. 9 Visualization of clustering results of TBM5. (a) Upper quartile, median and lower quartile of tunneling energy efficiency ratio

[0069] (a) The distribution of quartiles in each category; (b) The distribution of upper quartile, median and interquartile range of tunneling energy efficiency ratio in each category; (c) The distribution of upper quartile, lower quartile and interquartile range of tunneling energy efficiency ratio in each category; (d) The distribution of median, lower quartile and interquartile range of tunneling energy efficiency ratio in each category.

[0070] like Fig.10 As shown in the box plot, the median of the excavation energy efficiency ratio shows good regularity and distinguishability in the order of cluster 3, cluster 1, cluster 0, cluster 0, cluster 4, and cluster 2, which can better reflect the excavability of TBM rock mass, showing a gradual increase from easy to unexcavable, indicating that the energy consumed by TBM excavation per unit volume of rock mass increases successively, and the excavability becomes worse and worse. The increasing interquartile range of the excavation energy efficiency ratio also means the increase of control difficulty and uncertainty of operation parameters. Therefore, combined with engineering experience, cluster 3, cluster 1, cluster 0, cluster 4 and cluster 2, the five CM categories are defined as easy to excavate, excavable, difficult to excavate I, difficult to excavate II and unexcavable, and the excavability of TBM formations can be divided into five levels.

[0071] Example 2

[0072] A geological excavability classification method based on deep learning neural network is proposed. This model has not been reported in the relevant field before.

[0073] Comprehensive data collection: Collect various types of tunneling data from multiple tunnel projects, including but not limited to tunneling parameters (such as penetration, tunneling speed, cutterhead speed, cutterhead torque, etc.) of different types of tunnel boring machines (such as earth pressure balance shield machine, slurry balance shield machine, TBM, etc.) under different geological conditions, as well as detailed geological exploration data (such as stratum lithology, geological structure, etc.), to form a large-scale and diverse dataset.

[0074] Data cleaning and normalization: The collected data is cleaned strictly to remove data points with obvious errors or anomalies. At the same time, a standardized normalization method is used to map tunneling parameters and geological data of different magnitudes to a unified numerical range to facilitate the training and processing of the neural network. For example, for tunneling speed, it can be normalized to the [0,1] interval so that it can be operated on the same scale as other parameters.

[0075] Key feature extraction: Through the analysis of a large amount of data and engineering experience, a series of key features are extracted as the input of the neural network. In addition to the common excavation parameters, some derivative features that can comprehensively reflect the geological excavability are also introduced, such as the excavation energy consumption rate (calculated by the cutter head torque, excavation speed, etc., indicating the energy consumed by excavating a unit volume of rock and soil in a unit time), the formation hardness coefficient (estimated based on geological exploration data, reflecting the resistance of the formation to excavation), etc.

[0076] Feature combination and encoding: The extracted multiple features are reasonably combined and encoded to form a vector form suitable for neural network input. For example, the features such as penetration, tunneling speed, cutter head speed and tunneling energy consumption rate are spliced ​​into a high-dimensional vector in a certain order as the input layer node of the neural network.

[0077] Multilayer Perceptron (MLP) architecture: Multilayer Perceptron is used as the basic architecture, which includes an input layer, multiple hidden layers, and an output layer. The input layer receives vector data processed by feature engineering. The hidden layer sets multiple neuron nodes and uses nonlinear activation functions (such as ReLU function) to achieve complex mapping and feature extraction of input data. The output layer outputs the classification results of geological excavability according to the requirements of the classification task. For example, geological excavability can be divided into multiple categories such as easy to excavate, excavable, difficult to excavate, and non-excavable. The output layer converts the output of neurons into probability distributions of each category through the Softmax function.

[0078] Adaptive adjustment mechanism: In order to make the neural network better adapt to the data characteristics of different projects, an adaptive adjustment mechanism is introduced. During the training process, the number of neurons in the hidden layer, the learning rate and other hyperparameters are automatically adjusted according to the distribution of the training data and the training effect of the model to ensure that the model can converge quickly and achieve better classification performance.

[0079] Large-scale data set division: The collected large-scale data set is divided into training set, validation set and test set according to a certain ratio. Usually, 70% of the data can be used as training set for model training, 20% as validation set for adjusting model hyperparameters and evaluating model performance on unseen data, and 10% as test set for final evaluation of model generalization ability.

[0080] Loss function selection and optimization algorithm: Select a suitable loss function to measure the gap between the model prediction result and the true label, such as the cross entropy loss function, which performs well in multi-category classification tasks. At the same time, use advanced optimization algorithms (such as the Adam optimization algorithm) to update the weight parameters of the neural network, speed up the training speed of the model and improve the convergence performance of the model. During the training process, through continuous iteration, the performance of the model on the validation set is optimized, while ensuring that good generalization ability is maintained on the test set.

[0081] Real-time classification prediction: During the actual tunnel excavation process, the excavation data and geological data collected in real time are preprocessed and feature-engineered, and then input into the trained deep learning neural network model. The model can quickly output the current geological excavation classification results, providing construction personnel with timely decision-making basis.

[0082] Construction decision support: Based on the geological excavability classification results output by the model, the construction team can adjust the excavation strategy in a targeted manner. For example, for difficult or unexcavable strata, the excavation speed can be appropriately reduced, the cutter torque can be increased, etc. At the same time, the construction period can be reasonably arranged and construction resources can be deployed to improve construction efficiency, reduce construction costs and ensure construction safety.

[0083] Comprehensive consideration of multiple factors to accurately match the actual situation: By comprehensively collecting and integrating excavation parameters and geological data, and utilizing the powerful feature learning capabilities of deep learning neural networks, it is possible to deeply explore the complex relationship between the two, so that the classification results can more accurately reflect the actual geological excavability and achieve a high degree of match with the actual working performance of the tunnel boring machine, thereby effectively improving excavation efficiency.

[0084] Adapt to the characteristics of big data and have strong generalization ability: This method is trained and optimized based on large-scale data sets, and can fully adapt to the characteristics of tunnel engineering big data such as large volume, high diversity, high real-time performance, and multiple outliers. Through reasonable data partitioning and optimization algorithms, the model is ensured to have good generalization ability, which can not only achieve good results on training data, but also accurately predict geological excavability on new data that has never been seen, and has broad application prospects.

[0085] Fine classification to effectively deal with complex geological conditions: Subdividing geological excavability into multiple categories can more accurately reflect the differences in excavability under different geological conditions and provide more accurate construction guidance for the construction party. Construction personnel can take targeted measures based on the specific classification results to deal with the challenges brought by complex geological conditions in advance and ensure the smooth progress of construction.

[0086] In summary, the tunnel excavation geological excavation feasibility classification solution based on deep learning neural network overcomes many defects of the existing technology and provides a more advanced, accurate and practical geological excavation feasibility classification method for tunnel excavation projects.

[0087] Example 3

[0088] A tunneling geological excavability classification scheme based on the fusion of hybrid kernel support vector machine and fuzzy logic

[0089] A new algorithmic mathematical model based on the fusion of hybrid kernel function support vector machine and fuzzy logic is proposed for the classification of geological excavation feasibility of tunnel excavation. This model has not been reported in the relevant field before.

[0090] Data collection: Collect rich excavation data from different tunnel projects, covering the operating parameters of various types of tunnel boring machines (such as earth pressure balance shield machines, slurry balance shield machines, TBMs, etc.) under different geological conditions (including but not limited to penetration, excavation speed, cutterhead speed, cutterhead torque, etc.) and detailed geological exploration data (such as stratum lithology, geological structure, etc.), to form a comprehensive large data set.

[0091] Data preprocessing: preprocess the collected data. First, clean the data to remove obvious errors or abnormal data points to ensure data quality. Then, use normalization methods to unify the tunneling parameters and geological data of different magnitudes into a suitable numerical range for subsequent model processing. For example, the values ​​of each parameter can be mapped to the [0,1] interval through linear transformation.

[0092] Determination of key features: Based on engineering experience and analysis of a large amount of data, a series of key features are determined as inputs to the model. In addition to conventional excavation parameters, some features that can comprehensively reflect geological excavability are also considered, such as excavation energy consumption rate (calculated by cutter head torque, excavation speed, etc., indicating the energy consumed by excavating a unit volume of rock and soil in unit time), stratum compressive strength coefficient (estimated based on geological exploration data, reflecting the resistance of the stratum to excavation), etc.

[0093] Feature screening: Use statistical methods such as correlation analysis to screen the many extracted features, remove redundant features, and retain features that have a significant impact on geological excavability classification to improve the efficiency and accuracy of the model.

[0094] Kernel function selection and mixing: Traditional support vector machines usually use a single kernel function, such as linear kernel function, polynomial kernel function or Gaussian kernel function. In this model, we innovatively use a mixed kernel function to mix the linear kernel function and the Gaussian kernel function. The linear kernel function is suitable for processing linearly separable data, while the Gaussian kernel function performs well in processing nonlinear data. By reasonably adjusting the mixing ratio, the model can better adapt to the complex distribution of tunneling data under different geological conditions.

[0095] Model parameter optimization: After determining the hybrid kernel function, it is necessary to optimize other parameters of the support vector machine, such as the penalty parameter C and the kernel function parameter γ. Through methods such as cross-validation, these parameters are continuously adjusted on the training data set to find the parameter combination that optimizes the model performance.

[0096] Fuzzy processing: The output results (such as classification categories or classification probabilities, etc.) obtained by support vector machine processing and some key input features (such as excavation energy consumption rate, stratum compressive strength coefficient, etc.) are fuzzified. Fuzzification is to convert precise numerical values ​​into the form of fuzzy sets. For example, the excavation energy consumption rate is divided into fuzzy sets such as "low consumption", "medium consumption" and "high consumption" according to certain fuzzy rules.

[0097] Fuzzy reasoning rule formulation: Based on engineering experience and in-depth understanding of geological excavability, a series of fuzzy reasoning rules are formulated. These rules describe the relationship between different fuzzy input conditions (such as different fuzzy set combinations such as excavation energy consumption rate and formation compressive strength coefficient) and geological excavability classification results. For example, if the excavation energy consumption rate is "high consumption" and the formation compressive strength coefficient is "high", the geological excavability may be "difficult to excavate".

[0098] Defuzzification output: Fuzzy reasoning operation is performed according to the established fuzzy reasoning rules to obtain a fuzzy output result. Finally, the fuzzy output result is converted into an accurate geological excavability classification result through defuzzification methods (such as centroid method, maximum membership method, etc.), such as "easy to excavate", "excavable", "difficult to excavate", "unexcavable" and other specific categories.

[0099] Dataset division: The collected comprehensive large data set is divided into training set, validation set and test set according to a certain ratio. Usually, 70% of the data can be used as training set for model training, 20% as validation set for adjusting model hyperparameters and evaluating model performance on unseen data, and 10% as test set for final evaluation of model generalization ability.

[0100] Model training: The hybrid kernel support vector machine and fuzzy logic fusion model are trained using the training set. First, the hybrid kernel support vector machine part is trained, and the model parameters are continuously adjusted to achieve better classification performance on the training set. Then, the output of the support vector machine is sent to the fuzzy logic fusion module together with the key input features for training, so that the fuzzy logic part can accurately perform fuzzy reasoning and defuzzification output based on the output and input features of the support vector machine.

[0101] Model evaluation: Use validation sets and test sets to evaluate the trained model. By calculating evaluation indicators such as accuracy, recall, and F1 value, the performance of the model on different data sets is evaluated to ensure that the model has good generalization ability and accurate classification performance.

[0102] Real-time classification prediction: During the actual tunnel excavation process, the excavation data and geological data collected in real time are input into the trained hybrid kernel function support vector machine and fuzzy logic fusion model after preprocessing, feature extraction and selection. The model can quickly output the current geological excavation classification results and provide timely decision-making basis for construction personnel.

[0103] Construction decision support: Based on the geological excavability classification results output by the model, the construction team can adjust the excavation strategy in a targeted manner. For example, for difficult or unexcavable strata, the excavation speed can be appropriately reduced, the cutter torque can be increased, etc. At the same time, the construction period can be reasonably arranged and construction resources can be deployed to improve construction efficiency, reduce construction costs and ensure construction safety.

[0104] Comprehensive consideration of multiple factors to accurately match the actual situation: By integrating excavation parameters and geological data, and utilizing the powerful classification capabilities of the hybrid kernel support vector machine and the ability of fuzzy logic to handle uncertainty, it is possible to deeply explore the complex relationship between the two, so that the classification results can more accurately reflect the actual geological excavability and achieve a high degree of match with the actual working performance of the tunnel boring machine, thereby effectively improving excavation efficiency.

[0105] Adapt to the characteristics of big data and have strong generalization ability: This method is trained and optimized based on large-scale data sets, and can fully adapt to the characteristics of tunnel engineering big data such as large volume, high diversity, high real-time performance, and multiple outliers. Through reasonable data partitioning and optimization algorithms, the model is ensured to have good generalization ability, which can not only achieve good results on training data, but also accurately predict geological excavability on new data that has never been seen, and has broad application prospects.

[0106] Fine classification to effectively deal with complex geological conditions: Subdividing geological excavability into multiple categories can more accurately reflect the differences in excavability under different geological conditions and provide more accurate construction guidance for the construction party. Construction personnel can take targeted measures based on the specific classification results to deal with the challenges brought by complex geological conditions in advance and ensure the smooth progress of construction.

[0107] In summary, the tunnel excavation geological excavability classification scheme based on the fusion of hybrid kernel function support vector machine and fuzzy logic overcomes many defects of the existing technology and provides a more advanced, accurate and practical geological excavability classification method for tunnel excavation projects.

[0108] The present invention has been described in detail above in conjunction with the embodiments. However, those skilled in the art will appreciate that, without departing from the spirit of the present invention, the specific parameters in the above embodiments may be changed to form a plurality of specific embodiments, which are all within the common variation range of the present invention and will not be described in detail herein.

Claims

1. A method for classifying geological excavability based on excavation energy efficiency ratio, characterized in that: The process includes the steps of digitally processing various geological data for subsequent analysis and classification, including but not limited to using geological data extraction and conversion software to achieve geological digital conversion, and specifically includes the following steps: S1 processes the geological longitudinal section vector graphics, selectively retains the stratigraphic boundary, removes the redundant graphics elements that interfere with subsequent analysis, and closes the stratigraphic area; S2 performs a filling operation on the stratum area, filling each stratum area with a specified color or pattern according to a predetermined rule or actual needs; S3 can draw positioning reference lines for stratum identification based on relevant information of full-face tunnel boring machines under different actual working conditions; S4 uses appropriate identification means including but not limited to VBA vector graphics recognition technology to identify the various strata that the tunnel passes through in the geological longitudinal section map based on the reference line; S5 realizes the fusion of multi-source geological information and corresponding strata, where the multi-source geological information includes but is not limited to data obtained from various sources such as geological exploration and on-site monitoring.

2. The method for classifying geological excavability based on excavation energy efficiency ratio according to claim 1, characterized in that: It also includes the steps of acquiring, screening and preprocessing the excavation data. The following operations are included: S1 extracts effective excavation sections by analyzing data characteristics of tunnel boring machine excavation parameters, using methods including but not limited to change point detection methods; S2 uses one or more methods including box plots and moving median filtering to filter out outliers. The data is smoothed to reduce the impact of data fluctuations on subsequent analysis, and the criteria for determining outliers can be flexibly set according to actual conditions; S3 performs data aggregation processing on the original samples and can aggregate the "excavation section" into the "excavation ring" in a manner including but not limited to those based on statistical characteristics and actual engineering demand factors to meet data analysis needs of different scales and complexities.

3. The method for classifying geological excavability based on excavation energy efficiency ratio according to claim 2, characterized in that: In the data selection and preprocessing steps, when extracting the effective excavation section, according to different tunnel boring machine operating state judgment rules, the sample points corresponding to the operating state that is determined to interfere with the subsequent analysis are removed. The cyclic excavation process after removing the abnormal state data, that is, the excavation section, can be divided into different stages according to actual conditions, including but not limited to empty inference, rising section, stable section and descending section and other similar stage divisions.

4. The method for classifying geological excavability based on excavation energy efficiency ratio according to claim 2, characterized in that: In the data selection and preprocessing steps, outliers are identified using methods including but not limited to the box plot method, and the method for identifying outliers sets judgment criteria based on different statistical principles or engineering experience.

5. The method for classifying geological excavability based on excavation energy efficiency ratio according to claim 2, characterized in that: In the data selection and preprocessing steps, data noise reduction is performed using algorithms including but not limited to median filtering.

6. The method for classifying geological excavability based on excavation energy efficiency ratio according to claim 2, characterized in that: In the data selection and preprocessing steps, when performing data aggregation processing, statistical features including but not limited to the upper quartile (Q3), lower quartile (Q3), median (Median) and interquartile range (IQR) and other indicators that can reflect data characteristics are used to aggregate the "excavation sections" in the original data into "excavation rings", that is, the statistical information and other relevant information of the stable sections in all excavation cycles within a certain ring can be used as input features for clustering.

7. The method for classifying geological excavability based on excavation energy efficiency ratio according to claim 1, characterized in that: It also includes a step of selecting excavability indicators, including but not limited to engineering experience and theoretical analysis methods, preliminarily selecting but not limited to penetration, excavation speed, cutter head rotation speed and excavation energy efficiency ratio indicators and other related indicators that may affect geological excavability as indicators for geological excavability classification, using but not limited to engineering experience and theoretical analysis methods, measuring the degree of linear correlation between excavation parameters using appropriate correlation analysis methods including but not limited to Pearson correlation coefficient, and selecting appropriate parameters including but not limited to comprehensive parameter excavation energy efficiency ratio as indicators for excavability classification based on the analysis results, the specific definition of the excavation energy efficiency ratio is considered from the perspective of energy conservation or other relevant physical principles, the power consumed by the tunnel boring machine for excavating and excavating the rock and soil body should be equal to the energy absorbed by the corresponding rock and soil body to overcome its own strength or stability to form a broken body, and its calculation formula can be adjusted according to different engineering settings and physical principles.

8. The method for classifying geological excavability based on excavation energy efficiency ratio according to claim 1, characterized in that: It also includes a geological excavability classification step, which integrates but is not limited to the excavation energy efficiency ratio and its statistical characteristic indicators and other relevant indicators, and uses but is not limited to the C-means clustering algorithm to classify the excavability. The final classification result may include but is not limited to 11 categories, including 6 categories of shield machines, namely easy to excavate, excavable I, excavable II, difficult to excavate I, difficult to excavate II and unexcavable; and 5 categories of TBMs, namely easy to excavate, excavable, difficult to excavate I, difficult to excavate II and unexcavable.

9. The method for geological excavability classification based on excavation energy efficiency ratio according to claim 8, characterized in that: In the geological excavability classification step, the tunnel boring machine operating parameters are normalized using a suitable normalization method including but not limited to the following formula.

10. The method for geological excavability classification based on excavation energy efficiency ratio according to claim 8, characterized in that: In the geological excavability classification step, the surrounding rock classification is determined in combination with factors including but not limited to demand, expert knowledge, engineering experience and other relevant factors, which may include but not limited to 11 levels, including 6 types of shield machines and 5 types of TBMs, that is, the cluster number of shield machines K=6, the cluster number of TBMs K=5, the initialization stop threshold ε=1e-4, the fuzzy factor m=2, the maximum number of iterations T=300, and each parameter and classification standard can be flexibly adjusted according to actual conditions.

Images