Shield parameter geological recognition calculation method and system based on unsupervised clustering algorithm

Abstract

The invention belongs to the technical field of stratum recognition, and particularly relates to a shield parameter geological recognition calculation method and system based on an unsupervised clustering algorithm, and the method comprises the steps: selecting a plurality of groups of data containing factors affecting stratum category recognition; pCA converts a group of complex related variables into fewer and irrelevant variables, and the new variables are called as main components and can capture most of information in original data; determining the clustering number of K-means + + by adopting an elbow method; determining the clustering number of a hierarchical clustering method by adopting a hierarchical tree; k-means + + and hierarchical clustering are respectively used for clustering the data, and evaluation indexes such as identification precision or contour coefficients of the data and the data are calculated, so that potential modes in the data are analyzed and mined more comprehensively. The invention aims to deeply explore the internal relevance between shield tunnel parameters and geological types, provides an innovative and robust solution for real-time detection of geological features in shield construction by overcoming the limitation of the prior art, and has wide application potential.

Description

Technical Field

[0001] The present invention belongs to but is not limited to the technical field of stratum identification, and in particular relates to a shield parameter geological identification calculation method and system based on an unsupervised clustering algorithm. Background Art

[0002] As a large-scale equipment, tunnel boring machine is widely used in subway, railway and other tunnel projects. Its advantages are high excavation efficiency and high level of automation, which can effectively guarantee the construction progress. However, there are many factors that affect the performance, cost and project progress of the shield during construction. How to optimize the working parameters to improve the performance of the shield while ensuring construction safety has become a core issue in the field of tunneling. Shield parameters are closely related to geological characteristics and can be used to reflect changes in geological conditions. Traditionally, the BP neural network model based on supervised learning has been widely used to evaluate the relationship between shield parameters and stratum characteristics. However, this method relies heavily on the accuracy of stratum information, and under complex and changeable geological conditions, it is difficult for the neural network model to accurately capture the distribution characteristics of tunneling parameters. Therefore, in view of the real-time detection needs of different geological types, it is of great significance to use unsupervised learning models to identify geological characteristics.

[0003] In this context, the use of shield tunnel and borehole data combined with unsupervised clustering techniques such as K-means can achieve the prediction and identification of single or combined strata. Although these methods provide valuable insights in stratum identification, they still have certain limitations, especially when dealing with non-spherical or different density clusters, the K-means algorithm will show certain limitations. In view of this, it is particularly necessary to study other clustering strategies to address the above problems and thus improve the predictability of the model.

[0004] In the field of shield tunneling, although a large number of technologies and algorithms have been used to optimize the relationship between shield parameters and geological characteristics, there are still many technical problems in industrial application, which are mainly reflected in the following aspects:

[0005] 1. Insufficient recognition accuracy under complex geological conditions

[0006] The geological conditions encountered during shield construction are complex and changeable, such as alternating soft and hard strata, sand layers, clay layers, rock layers, etc. Traditional supervised learning methods, such as BP neural networks, require a large amount of labeled data to train the model. However, in actual engineering, the labeled data of the strata may not be accurate or sufficient, especially when faced with complex geological conditions, it is difficult to accurately evaluate the correspondence between shield parameters and stratum characteristics. This will cause the model to be unable to adapt to the diverse geological environment, which in turn affects the safety and efficiency of construction.

[0007] 2. High cost of data annotation and reliance

[0008] Supervised learning methods such as BP neural networks rely on a large amount of labeled data to train models, and the acquisition of these labeled data usually requires professional geological survey personnel to conduct detailed sampling and analysis through drilling or other methods. This process is not only time-consuming, but also increases construction costs. In actual tunneling projects, it is difficult and costly to obtain labeled data under different geological conditions, which limits the widespread application of supervised learning models.

[0009] 3. Clustering limitations of unsupervised models

[0010] Unsupervised learning models, such as K-means, can cluster and identify shield parameters and geological features without labeled data, but their algorithms have certain limitations. K-means assumes that each cluster is spherical and evenly distributed, which performs poorly when dealing with non-spherical or clusters of different densities and can easily lead to recognition errors. In addition, K-means is sensitive to the selection of initial cluster centers and can easily fall into local optimal solutions, resulting in unstable clustering effects, which in turn affects the identification and optimization of shield parameters.

[0011] 4. Real-time response and efficient processing requirements

[0012] During shield construction, the construction progress and the working parameters of the shield machine need to be adjusted in real time to adapt to the changing geological conditions. Traditional supervised or unsupervised learning models often have low computational efficiency when processing large-scale real-time data and cannot meet the requirements of real-time response. This lag leads to the failure to adjust the working parameters in time during the construction process, thus affecting the construction progress and even bringing safety hazards.

[0013] 5. Nonlinear and complex relationship between geological characteristics and shield parameters

[0014] The relationship between geological characteristics and shield parameters is often complex and nonlinear. Traditional linear models and simple clustering algorithms cannot accurately capture this complexity, especially in complex working conditions such as alternating soft and hard strata and fault zones. It is difficult for the model to make accurate predictions on the changes in shield parameters. This will lead to delayed or inaccurate adjustment of construction parameters, affecting the performance and construction quality of the shield machine.

[0015] 6. Limited processing capabilities for multi-source heterogeneous data

[0016] The parameter data and geological data generated during shield construction usually come from multiple different sensors and exploration equipment, including pressure, torque, propulsion speed, geological drilling data, etc. These data are heterogeneous and complex. Traditional clustering algorithms are difficult to process multiple types of data sources at the same time, resulting in the inability of the model to effectively integrate and analyze all relevant information, which in turn limits the application effect of the model.

[0017] Technical problems that need to be solved urgently in industrial applications:

[0018] 1. How to improve the accuracy and stability of geological identification under complex geological conditions: There is an urgent need to develop a more intelligent and robust model that can accurately identify the correspondence between shield parameters and geological characteristics under environments with high geological complexity and can adapt to various stratigraphic changes.

[0019] 2. Reduce dependence on labeled data: Explore unsupervised or semi-supervised learning methods to reduce the need for large amounts of labeled data, reduce data acquisition costs, and improve the universality and scalability of the model.

[0020] 3. Improve unsupervised clustering algorithms to cope with complex cluster shapes: Research and introduce unsupervised clustering algorithms that adapt to complex cluster shapes and density distributions to address the limitations of traditional methods such as K-means when dealing with non-spherical and non-uniformly distributed data.

[0021] 4. Improve real-time processing and prediction capabilities: In response to the real-time requirements of shield construction, develop efficient algorithms and systems that can respond quickly in a large-scale data environment and adjust construction parameters in real time to ensure the safety and efficiency of shield construction.

[0022] 5. Dealing with complex nonlinear relationships: More powerful algorithms need to be developed to model the nonlinear relationship between shield parameters and geological characteristics to ensure that shield parameters can still be accurately predicted and optimized under complex geological conditions.

[0023] 6. Ability to process multi-source heterogeneous data: There is an urgent need for a model that can effectively integrate heterogeneous data generated by different sensors and exploration equipment in order to comprehensively analyze and mine key information in shield construction and provide a basis for decision-making in the construction process.

[0024] In summary, the technical problems existing in current industrial applications mainly focus on accuracy, real-time performance, multi-source data processing and algorithm limitations under complex geological conditions. Researching and developing new unsupervised learning models and clustering strategies will help solve these problems and further improve the intelligence level and operating efficiency of shield tunneling machines in tunnel construction. Summary of the invention

[0025] In view of the problems existing in the prior art, the present invention provides a shield parameter geological identification calculation method and system based on an unsupervised clustering algorithm.

[0026] The present invention is achieved by providing a shield parameter geological identification calculation method based on an unsupervised clustering algorithm, the method comprising:

[0027] S1: First, select several groups of data containing factors that affect the identification of stratigraphic categories; PCA transforms a set of complex related variables into fewer and unrelated variables. These new variables are called principal components, which can capture most of the information in the original data;

[0028] S2: The elbow method is used to determine the number of clusters of K-means++;

[0029] S3: Using hierarchical tree to determine the number of clusters in hierarchical clustering method;

[0030] S4: Use K-means++ and hierarchical clustering to cluster the data respectively, and calculate the evaluation indicators such as recognition accuracy or silhouette coefficient of the two to more comprehensively analyze and mine the potential patterns in the data; by comparing the clustering accuracy of the two algorithms, we can highlight the effectiveness of K-means++ when processing large-scale data with high speed requirements.

[0031] Further, the S1 specifically includes:

[0032] (1) Given a p×n-dimensional data matrix X, where each sample p is n-dimensional, the matrix can be decomposed using singular value decomposition as X = U∑V T ; In this decomposition, U is a p×n orthogonal matrix, V is an n×n orthogonal matrix, and ∑ is a diagonal matrix of dimension n×n;

[0033] (2) The orthogonal matrix satisfies UU T =1 and VV T =1;

[0034] (3) For dimensionality reduction, divide ∑ into r columns, ∑ r , the data after dimension reduction is

[0035] (4) Based on the above analysis, the cumulative explained variance ratio of each principal component is plotted to observe the cumulative explained rate of variance. When the cumulative explained variance reaches a certain threshold, such as 80% or 90%, it can be considered that these principal components can explain most of the data changes. Usually, the number of principal components corresponding to the cumulative explained variance reaching the threshold is selected.

[0036] Further, the S2 specifically includes:

[0037] (1) Data preparation: Standardize the original data to eliminate the influence of different dimensions on the clustering results;

[0038] (2) Initialize cluster centers: Use the K-means++ algorithm to randomly select initial cluster centers to ensure a reasonable distribution of initial points and improve clustering efficiency;

[0039] (3) Calculate the sum of squared errors (SSE): For different numbers of clusters, run the K-means++ clustering algorithm in turn and calculate the sum of squared errors corresponding to each number of clusters;

[0040] (4) Draw the SSE curve: Draw the SSE values ​​of different cluster numbers into a curve, with the horizontal axis being the number of clusters and the vertical axis being the SSE value;

[0041] (5) Determine the “elbow point”: According to the change of the decreasing speed of the SSE curve, find the point with the largest decreasing rate, the so-called “elbow point”. At this time, the number of clusters is the optimal value;

[0042] (6) Apply the optimal number of clusters: Use the optimal number of clusters found to run the K-means++ algorithm again to obtain the final clustering result.

[0043] Further, the S3 specifically includes:

[0044] (1) Calculate the distance or similarity matrix between all data points. Commonly used distance measurements include Euclidean distance;

[0045] (2) Initialize the clustering process by treating each data point as an independent cluster;

[0046] (3) Merge the nearest clusters according to Ward and update the distance matrix until all data points are merged into one cluster;

[0047] (4) Construct a hierarchical tree based on the aggregation process. Each branch represents a merge, and the height of the tree represents the distance or similarity measure at the time of the merge;

[0048] (5) When constructing the data tree, the Euclidean distance is used to calculate the distance between samples, and the average linkage method is used to assemble the hierarchical tree structure.

[0049] Another object of the present invention is to provide a shield parameter geological identification calculation system based on an unsupervised clustering algorithm based on the shield parameter geological identification calculation method based on an unsupervised clustering algorithm, the system specifically comprising:

[0050] In the dimensionality reduction analysis module, first, several groups of data containing factors that affect the identification of stratigraphic categories are selected; PCA transforms a set of complex related variables into fewer and unrelated variables. These new variables are called principal components, which can capture most of the information in the original data;

[0051] A cluster number determination module, connected to the dimension reduction analysis module, uses the elbow method to determine the number of clusters of K-means++ and uses the hierarchical tree to determine the number of clusters of the hierarchical clustering method;

[0052] The clustering module is connected to the cluster number determination module, which uses K-means++ and hierarchical clustering to cluster the data respectively, and calculates the evaluation indicators such as recognition accuracy or silhouette coefficient of the two algorithms to more comprehensively analyze and mine the potential patterns in the data; by comparing the clustering accuracy of the two algorithms, it can highlight the effectiveness of K-means++ in processing large-scale data with high speed requirements.

[0053] Another object of the present invention is to provide a computer device, which includes a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the shield parameter geological identification calculation method based on an unsupervised clustering algorithm.

[0054] Another object of the present invention is to provide a computer-readable storage medium storing a computer program, which, when executed by a processor, enables the processor to execute the steps of the shield parameter geological identification calculation method based on an unsupervised clustering algorithm.

[0055] Another object of the present invention is to provide an information data processing terminal, which is used to implement the shield parameter geological identification calculation system based on the unsupervised clustering algorithm as described above.

[0056] In combination with the above technical solutions and the technical problems solved, the advantages and positive effects of the technical solutions to be protected by the present invention are as follows:

[0057] First, construct a comprehensive analysis framework: By integrating unsupervised clustering algorithms such as K-means++ and hierarchical clustering, a comprehensive framework suitable for shield tunnel stratum identification is constructed, which can handle complex geological features.

[0058] Revealing the relationship between shield parameters and geological characteristics: Through cluster analysis of original data and dimensionality reduction data, the potential relationship between shield parameters and geological characteristics is revealed, providing an effective basis for shield construction.

[0059] Improve recognition accuracy: The K-means++ algorithm showed a high recognition rate of 89.54% in stratum recognition, which is more effective than other methods and improves recognition accuracy.

[0060] Optimizing the selection of construction parameters: The invention can provide real-time feedback for the shield tunneling process, optimize the selection of construction parameters, and ensure construction efficiency and safety.

[0061] Broad engineering application prospects: The present invention has good robustness and can be widely used in tunnel projects such as subways and railways, thereby improving the speed of decision-making processes in similar engineering projects.

[0062] Second, as auxiliary evidence of the inventiveness of the claims of the present invention, it is also reflected in the following important aspects:

[0063] (1) The expected benefits and commercial value of the technical solution of the present invention after transformation are:

[0064] By optimizing the accuracy and efficiency of shield construction parameters, multiple benefits can be achieved. First, optimization can reduce construction costs, avoid waste of resources, and improve efficiency, thereby being economical and efficient. Second, accurate parameter optimization helps to plan and manage project progress, improve the overall project management level, and prevent delays. At the same time, optimized parameters can make construction more efficient and smooth, reduce the number of errors and adjustments, improve the utilization rate of workers and equipment, reduce the occurrence of accidents, and enhance safety. In addition, optimization can also help to cope with complex geological conditions and construction environment risks to ensure the smooth progress of the project. Technological innovation will attract more companies to adopt it, enhance market competitiveness, bring economic benefits and development opportunities to the field of tunneling, promote the industry to develop in a more efficient and safer direction, and stimulate more innovation and progress. (4) The technical solution of the present invention overcomes technical bias:.

[0065] In this technical solution, the traditional BP neural network model based on supervised learning usually requires a large amount of labeled data for training, and has high requirements for the accuracy and completeness of the data, which is limited under complex and changeable geological conditions. In contrast, the introduction of unsupervised learning models and new clustering techniques makes the model more flexible and applicable to geological conditions in actual scenarios. Through unsupervised learning, the model can learn from unlabeled data and discover hidden patterns and structures, thereby reducing the need for a large amount of labeled data, while also improving the model's adaptability to unknown geological conditions. The introduction of new clustering technology further enhances the model's ability to identify and classify geological features, making it more robust and practical. This technological innovation not only improves the predictability and applicability of the model, but also brings a new solution to the field of tunneling. By overcoming the high dependence of traditional methods on stratum information and the difficulty of dealing with complex geological conditions, this technology provides engineers with more effective tools and methods to cope with challenges and provides more security for the smooth progress of engineering projects. In the future, this new idea and technical path will bring more opportunities and innovations to the development of related fields, and promote the entire industry to develop in a more efficient and safer direction.

[0066] Third, the technical solution of the present invention solves several key technical problems in the prior art in industrial applications and brings about significant technical progress, as follows:

[0067] 1. Problems to be solved

[0068] (1) Insufficient geological identification accuracy

[0069] In the prior art, the geological identification method commonly used in shield construction is based on traditional supervised learning and relies on a large amount of labeled data. However, in practical applications, obtaining accurate labeled data is costly and time-consuming, which easily leads to insufficient accuracy in stratum classification identification. The present invention introduces an unsupervised clustering algorithm, which does not rely on a large amount of labeled data, and directly performs automatic clustering analysis from the shield parameter data collected during the construction process, significantly improving the automation level and accuracy of geological identification.

[0070] (2) Problems of high data complexity and information redundancy

[0071] During the shield construction process, the parameter data collected are usually high-dimensional and complex. Direct analysis of these data will lead to low algorithm efficiency and easily generate noise and redundant information. To solve this problem, the present invention uses principal component analysis (PCA) to reduce the dimension of the data, eliminate redundant information, and retain the key features in the data, thereby improving the accuracy and processing efficiency of subsequent clustering.

[0072] (3) The problem of difficulty in determining the number of clusters

[0073] In cluster analysis, choosing the right number of clusters is an important factor affecting the clustering effect, but the prior art lacks effective tools and methods to determine the optimal number of clusters. The present invention introduces the elbow method and the hierarchical tree method to determine the number of clusters for K-means++ and hierarchical clustering, respectively, thereby avoiding the subjectivity and error of manual experience selection, having a high degree of automation, and providing more reliable results.

[0074] (4) Efficiency in processing large-scale data

[0075] Existing geological identification methods are inefficient and prone to computational bottlenecks when processing large-scale data. The present invention uses the K-means++ clustering algorithm, which optimizes the initial centroid selection and greatly improves the processing speed of large-scale data sets, especially in the large amount of real-time data generated during shield construction, and can quickly process and identify stratum categories.

[0076] (5) Lack of comprehensive clustering effect evaluation methods

[0077] In the prior art, the evaluation of clustering results is usually relatively simple, and it is impossible to comprehensively analyze the clustering effect and determine the best model. The present invention comprehensively evaluates the clustering effects of different algorithms by calculating the recognition accuracy, silhouette coefficient and other evaluation indicators of K-means++ and hierarchical clustering, and provides a scientific basis for selecting the optimal algorithm for actual geological identification.

[0078] 2. Technological advancement

[0079] (1) Improvement of automation and intelligence

[0080] Through the unsupervised clustering algorithm of the present invention, the geological identification process is highly automated, and the stratum classification can be automatically completed according to the shield parameter data without a lot of manual intervention. Compared with the traditional supervised learning method, it reduces the dependence on label data and improves the intelligence level of the system.

[0081] (2) Significant improvement in data processing efficiency

[0082] Through the dimensionality reduction analysis module, the original data dimension can be significantly reduced, reducing redundant information and noise, thereby improving data processing efficiency. At the same time, the optimized initial centroid selection mechanism of the K-means++ algorithm significantly improves the processing speed of large-scale data, which is especially suitable for the processing of big data generated in real time during shield construction.

[0083] (3) Comprehensiveness of clustering effect evaluation

[0084] The present invention uses K-means++ and hierarchical clustering algorithms, and combines multi-dimensional evaluation indicators (such as recognition accuracy, silhouette coefficient, etc.) to perform effect analysis, so that the advantages and disadvantages of different algorithms can be fully understood. Such multi-angle evaluation provides a more scientific basis for model selection in practical applications and improves the accuracy and reliability of geological identification results.

[0085] (4) Automatic determination of the number of clusters

[0086] In the cluster number determination module of the present invention, the number of clusters is automatically determined through the elbow method and the hierarchical tree method, avoiding the subjective problem of manually selecting the number of clusters in the traditional clustering method. This innovation makes the algorithm more adaptable to different data sets and improves the automation level of the clustering process.

[0087] (5) Practicality of large-scale data processing

[0088] Since shield construction involves massive amounts of real-time data, the present invention can quickly process and analyze large-scale data by combining unsupervised clustering algorithms, especially the fast clustering capability of K-means++. The system is particularly suitable for real-time geological identification, which helps to improve the efficiency and safety of shield construction and reduce construction risks caused by unknown geology.

[0089] (6) Strong scalability

[0090] The present invention can not only be used for geological identification of shield parameters, but can also be extended to other types of label-free data analysis scenarios. The system architecture has good versatility and can adapt to different industrial application requirements, such as geological exploration, engineering analysis and other fields, providing a powerful tool for stratum identification, classification and analysis.

[0091] In summary, the present invention solves multiple bottleneck problems in the prior art by combining innovative unsupervised clustering algorithms with multiple dimensionality reduction and evaluation methods, significantly improves the accuracy and efficiency of geological identification of shield parameters, and provides strong support for large-scale data processing and automated geological identification. This technical solution has significant technological progress and broad application prospects in industrial applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0092] Figure 1 It is a flow chart of a shield parameter geological identification calculation method based on an unsupervised clustering algorithm provided by an embodiment of the present invention;

[0093] Figure 2 It is a structural diagram of a shield parameter geological identification calculation system based on an unsupervised clustering algorithm provided by an embodiment of the present invention;

[0094] Figure 3 is a principal component explained variance and cumulative variance diagram provided by an embodiment of the present invention;

[0095] Figure 4 The clustering algorithm results provided by the embodiments of the present invention are: (a) the original data clustering result based on the K-mean++ clustering algorithm;

[0096] Figure 5 1. The stratum identification based on K-Mean++ clustering algorithm provided by the embodiment of the present invention is compared with the actual stratum classification: (a) original data and (b) dimension-reduced data. DETAILED DESCRIPTION

[0097] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0098] like Figure 1 As shown, an embodiment of the present invention provides a shield parameter geological identification calculation method based on an unsupervised clustering algorithm, the method comprising:

[0099] S1: First, select several groups of data containing factors that affect the identification of stratigraphic categories; Principal Component Analysis (PCA) transforms a set of complex related variables into fewer and unrelated variables. These new variables are called principal components, which can capture most of the information in the original data;

[0100] S2: The elbow method is used to determine the number of clusters of K-means++;

[0101] S3: Using hierarchical tree to determine the number of clusters in hierarchical clustering method;

[0102] S4: Use K-means++ and hierarchical clustering to cluster the data respectively, and calculate the evaluation indicators such as recognition accuracy or silhouette coefficient of the two to more comprehensively analyze and mine the potential patterns in the data; by comparing the clustering accuracy of the two algorithms, we can highlight the effectiveness of K-means++ when processing large-scale data with high speed requirements.

[0103] Step 1: Data preprocessing and principal component analysis (PCA)

[0104] In this step, several groups of important parameters and data that affect the identification of stratum categories are first selected. These data include multiple dimensions such as stratum pressure, soil density, and water content. Due to the correlation and complexity of these original data, PCA is introduced to convert complex high-dimensional data into fewer and unrelated principal components. Through PCA technology, most of the information in the data can be retained while reducing the variable dimension, helping subsequent clustering algorithms to operate in a simpler data space. This helps to improve the computational efficiency and accuracy of the algorithm.

[0105] Step 2: Combination of K-means++ clustering and elbow method

[0106] In order to determine the number of clusters of the K-means++ algorithm, the embodiment of the present invention adopts the elbow method. The elbow method is a commonly used method for evaluating the number of clusters. By plotting the relationship between the number of clusters and the sum of squared errors (SSE) within the cluster, it can be observed that the curve will be significantly "bent" at a certain point, which is the so-called "elbow". The number of clusters corresponding to this point is considered to be a better choice. As an improved version of K-means, K-means++ can be more efficient in processing large-scale data and reduce local optimal situations by optimizing the selection of initial cluster centers, thereby improving clustering effect and speed.

[0107] Step 3: Hierarchical clustering and hierarchical tree to determine the number of clusters

[0108] In addition to the K-means++ algorithm, the present invention also introduces a hierarchical clustering algorithm for data analysis. By constructing a hierarchical tree, the clustering structure of the data at different levels can be observed. Hierarchical clustering does not require the number of clusters to be specified in advance, but generates a hierarchical structure by gradually merging or splitting data points. After generating the hierarchical tree, the appropriate number of clusters can be determined by observing the segmentation of the tree. This method can handle smaller data sets more flexibly, and is particularly suitable for scenarios with complex data structures that require more fine-grained analysis.

[0109] Step 4: Evaluation and comparative analysis of clustering results

[0110] In the last step, the results of K-means++ and hierarchical clustering are evaluated respectively. By calculating the recognition accuracy, silhouette coefficient and other indicators, the performance of the two algorithms in geological identification is compared. K-means++ is suitable for use in scenarios with high real-time requirements because it can process large-scale data at a faster speed and higher efficiency. Hierarchical clustering, due to its more detailed hierarchical structure, can mine more complex patterns in small-scale data. By comparing and analyzing the effects of the two algorithms, a more reasonable selection basis can be provided for engineering applications, and the accuracy and speed of geological identification can be comprehensively improved.

[0111] This shield parameter geological identification method based on an unsupervised clustering algorithm can adapt to data sets of different sizes and complexities, mine potential patterns in geological parameters, and optimize the clustering effect of the algorithm.

[0112] PCA is a commonly used unsupervised machine learning algorithm, mainly used to process and analyze high-dimensional data. Its core idea is to simplify the data structure by reducing the dimensionality while retaining the main information of the data to make subsequent analysis more efficient.

[0113] Elbow Method: SSE is one of the important indicators to measure the clustering effect, which is mainly used to evaluate the looseness and compactness of the clustering results. SSE is an objective function that reflects the degree of deviation between the cluster center and each data point. Generally speaking, the smaller the SSE, the smaller the distance between the data point and its cluster center, the tighter the clustering result and the better the effect. In practical applications, as the number of clusters k increases, SSE will gradually decrease. This is because as the number of clusters increases, the number of data points contained in each cluster decreases, so the distance between the data points inside each cluster and its center will also decrease. However, excessively increasing the number of clusters will lead to overfitting problems, and the cluster center cannot well summarize the overall distribution of the data. In order to determine the optimal number of clusters k, a curve of the relationship between the sum of squared errors and the number of clusters is usually plotted. This curve is called the "Elbow Method Curve". This curve can help find the appropriate number of clusters.

[0114] A hierarchical tree is a way to visualize the results of hierarchical clustering. Each node in the tree corresponds to a cluster, and the leaf nodes represent specific data points. The height of the tree reflects the level of the clusters, and clusters closer to the root represent earlier mergers or splits. The tree diagram is built based on the aggregation process, each branch represents a merger of clusters, and the height of the tree represents the distance or similarity measure at that time. With the hierarchical tree, users can easily choose the optimal number of clusters and understand the hierarchical structure in the data.

[0115] K-Means clustering uses an iterative algorithm to optimize the cluster center. Its working principle is to maximize the distance between samples of different categories and minimize the distance between samples of the same category. Given a data set D = (x1, x2, ..., x n ), where n is the total number of samples contained in the data set, and p is the data dimension of a single sample. The specific iterative algorithm steps are as follows:

[0116] (1) Input data set D and cluster number;

[0117] (2) Randomly select sample k from the data set D as the initial cluster center μ = μ1, μ2, ..., μ k ;

[0118] (3) For each sample x in the dataset i , which to k cluster centers μ j The distance between them is d ij ij=||x i -μ j ||2 2 Calculate. i Divide into the cluster center with the smallest distance, and the cluster is divided into C = {C1, C2, ..., C k};

[0119] (4) For each cluster center, recalculate all samples belonging to this class obtained in step (A2) to obtain the cluster center value

[0120] (5) If the iteration stop condition is not met, repeat steps (A2) and (A3);

[0121] (6) Output cluster partition set C = {C1, C2, ..., C k}.

[0122] It is worth noting that the random selection of sample points as cluster centers to initialize the centroid in step (A1) has a significant impact on the final result and computational efficiency of the algorithm. Random selection leads to suboptimal convergence. Therefore, this patent uses K-Means++ to initialize the cluster centers in the K-Means algorithm to improve its performance and reliability. The specific steps are as follows:

[0123] (1) Randomly select a sample from the input data set D as the first cluster center μ1;

[0124] (2) For each sample point x i , calculate its distance to the cluster center, and get the minimum distance;

[0125] (3) Select D xThe sample point with the largest value is used as the new cluster center;

[0126] (4) Repeat steps (2) and (3) until two cluster centers are selected.

[0127] The advantage of the K-Means++ clustering algorithm is that it significantly improves the stability and accuracy of the clustering results by optimizing the selection of the initial cluster centers, avoiding the local optimal problem caused by random initialization in the traditional K-Means algorithm. In addition, K-Means++ has high computational efficiency and is suitable for large-scale data processing. However, the disadvantage is that it still requires the number of clusters to be specified in advance, and it is not effective for clusters with non-spherical distribution or different densities, and is sensitive to noise and outliers.

[0128] Hierarchical clustering algorithms decompose the given data layer by layer until specific conditions are met. According to different strategies, these algorithms can be divided into two categories: agglomerative (agglomerative) and divisive (divisive) hierarchical clustering. Agglomerative hierarchical clustering adopts a bottom-up approach, while divisive hierarchical clustering adopts a top-down approach. Given a data set D = (d1, d2, ..., d n ), the basic principle is as follows:

[0129] (1) Take each data point d in D i As a class center c i =d i , forming a cluster set D = (c1, c2, ..., c n );

[0130] (2) Calculate every two classes (c i , c j ) between the similarity sim(c i , c j ), the formula is: in

[0131] (3) Select the two largest clusters with the greatest similarity (c i , c j ), that is, maxsim(c i , c j ), combining them into a new cluster c k =c i ∪c j , and at the same time c i and c j The feature vectors of are combined into a new cluster set D = (c1, c2, ..., c n-1 );

[0132] (4) Iterate the above steps until the required number of clusters is reached to obtain the final clustering result.

[0133] The main advantages of hierarchical clustering algorithms include their suitability for discovering the hierarchical structure of data, their ability to handle non-spherical distributions, their strong stability, and their support for multi-scale analysis. However, their main disadvantage is their high computational complexity, especially when dealing with large-scale data, and their low computational efficiency.

[0134] The S1 specifically includes:

[0135] (1) Given a p×n-dimensional data matrix X, where each sample p is n-dimensional, the matrix can be decomposed using singular value decomposition as X = U∑V T ; In this decomposition, U is a p×n orthogonal matrix, V is an n×n orthogonal matrix, and Σ is a diagonal matrix of dimension n×n;

[0136] (2) The orthogonal matrix satisfies UU T =1 and VV T =1;

[0137] (3) For dimensionality reduction, divide Σ into r columns, ∑ r , the data after dimension reduction is

[0138] (4) Based on the above analysis, the cumulative explained variance ratio of each principal component is plotted to observe the cumulative explained rate of variance. When the cumulative explained variance reaches a certain threshold, such as 80% or 90%, it can be considered that these principal components can explain most of the data changes. Usually, the number of principal components corresponding to the cumulative explained variance reaching the threshold is selected.

[0139] The S2 specifically includes:

[0140] (1) Data preparation: Standardize the original data to eliminate the influence of different dimensions on the clustering results;

[0141] (2) Initialize cluster centers: Use the K-means++ algorithm to randomly select initial cluster centers to ensure a reasonable distribution of initial points and improve clustering efficiency;

[0142] (3) Calculate the sum of squared errors (SSE): For different numbers of clusters, run the K-means++ clustering algorithm in turn and calculate the sum of squared errors corresponding to each number of clusters;

[0143] (4) Draw the SSE curve: Draw the SSE values ​​of different cluster numbers into a curve, with the horizontal axis being the number of clusters and the vertical axis being the SSE value;

[0144] (5) Determine the “elbow point”: According to the change of the decreasing speed of the SSE curve, find the point with the largest decreasing rate, the so-called “elbow point”. At this time, the number of clusters is the optimal value;

[0145] (6) Apply the optimal number of clusters: Use the optimal number of clusters found to run the K-means++ algorithm again to obtain the final clustering result.

[0146] The S3 specifically includes:

[0147] (1) Calculate the distance or similarity matrix between all data points. Commonly used distance measurements include Euclidean distance;

[0148] (2) Initialize the clustering process by treating each data point as an independent cluster;

[0149] (3) Merge the nearest clusters according to Ward and update the distance matrix until all data points are merged into one cluster;

[0150] (4) Construct a hierarchical tree based on the aggregation process. Each branch represents a merge, and the height of the tree represents the distance or similarity measure at the time of the merge;

[0151] (5) When constructing the data tree, the Euclidean distance is used to calculate the distance between samples, and the average linkage method is used to assemble the hierarchical tree structure.

[0152] The embodiment of the present invention provides a shield parameter geological identification calculation system based on an unsupervised clustering algorithm based on the shield parameter geological identification calculation method based on an unsupervised clustering algorithm, and the system specifically includes:

[0153] In the dimensionality reduction analysis module, first, several groups of data containing factors that affect the identification of stratigraphic categories are selected; PCA transforms a set of complex related variables into fewer and unrelated variables. These new variables are called principal components, which can capture most of the information in the original data;

[0154] A cluster number determination module, connected to the dimension reduction analysis module, uses the elbow method to determine the number of clusters of K-means++ and uses the hierarchical tree to determine the number of clusters of the hierarchical clustering method;

[0155] The clustering module is connected to the cluster number determination module, which uses K-means++ and hierarchical clustering to cluster the data respectively, and calculates the evaluation indicators such as recognition accuracy or silhouette coefficient of the two algorithms to more comprehensively analyze and mine the potential patterns in the data; by comparing the clustering accuracy of the two algorithms, it can highlight the effectiveness of K-means++ in processing large-scale data with high speed requirements.

[0156] The working principle of the embodiment of the present invention mainly revolves around the geological identification of shield parameters based on an unsupervised clustering algorithm, and its specific process can be divided into six key steps for explanation:

[0157] 1. Dimensionality reduction analysis module: First, the data is processed through the dimensionality reduction analysis module to select several groups of data containing stratigraphic classification identification factors. Since there are multiple dimensions and strong correlations in the original data, principal component analysis (PCA) is introduced to transform these complex related variables into fewer and unrelated variables. These newly generated variables are called principal components, which can retain the information in the original data to the maximum extent, thereby reducing the data dimension and reducing information loss, laying the foundation for the subsequent clustering process.

[0158] 2. Cluster number determination module: Next, the cluster number determination module uses the elbow method to determine the optimal number of clusters in the K-means++ algorithm based on the reduced-dimensional data. The elbow method selects the point with the best clustering effect by evaluating the change in clustering error under different numbers of clusters. At the same time, the hierarchical tree method is used to determine the number of clusters for hierarchical clustering. These two methods are used to reasonably determine the number of clusters to ensure the accuracy and reliability of the clustering results.

[0159] 3. K-means++ clustering in the clustering module: In the clustering module, the K-means++ algorithm is first used to cluster the reduced-dimensional data. K-means++ is an improved K-means algorithm that improves the clustering speed and effect by optimizing the initial center point selection process, especially showing significant advantages when processing large-scale data sets. The clustering process minimizes the distance from the data point to its nearest centroid and ultimately divides the data into several categories.

[0160] 4. Hierarchical clustering in the clustering module: In parallel with K-means++, the hierarchical clustering algorithm is also used in the clustering module. Hierarchical clustering gradually builds a nested cluster hierarchy of the data, and forms a cluster structure from small to large by calculating the similarity between data points. This method does not require the number of clusters to be set in advance, and is suitable for analyzing data sets with potential hierarchical structures, providing a different perspective for data pattern analysis.

[0161] 5. Evaluation and comparison of clustering results: After completing the clustering tasks of K-means++ and hierarchical clustering, the system calculates the clustering accuracy and evaluation indicators (such as silhouette coefficient) of the two algorithms. By comparing the performance of the two clustering algorithms, it can be found that K-means++ is effective in processing large-scale data sets, while hierarchical clustering has advantages in revealing the hierarchical structure of data and analyzing complex patterns. Such a comprehensive comparative analysis provides a more accurate basis for model selection for geological identification.

[0162] 6. Functions of computer equipment and storage medium: The present invention also includes corresponding computer equipment and computer-readable storage medium. The combination of these hardware and software ensures that the algorithm can run efficiently. The processor of the computer equipment implements the geological identification of shield parameters based on the unsupervised clustering algorithm step by step by executing the program stored in the memory. The storage medium provides a persistent storage space for the computer program, ensuring that the entire identification system can be stably executed in practical applications and generate effective identification results.

[0163] Through the collaborative work of the above modules, the embodiment of the present invention provides an efficient and accurate geological identification system that can process large-scale data in shield construction and provide accurate stratigraphic classification results, thereby improving the efficiency and safety of the project. The embodiment of the present invention is selected from a certain shield tunnel construction project. The data set contains 966 groups of data, each of which includes 11 factors that affect stratigraphic classification identification. The specific parameters are as follows:

[0164] Original shield parameters: propulsion speed (AR), thrust (F), cutterhead torque (T), cutterhead speed (CRS) and actual pressure in the soil bin (RP).

[0165] Calculation parameters: including mean thrust (MF), mean cutter torque (MT), penetration rate (PR), field penetration index (FPI), torque penetration index (TPI) and specific energy (SE).

[0166] These parameters are recorded in real time by sensors inside the shield machine, reflecting the relationship between shield machine operation and geological conditions. First, the 11 parameters are statistically described (as shown in Table 1) and the data distribution is displayed. The statistical characteristics of the data reveal the changing laws of shield machine operating parameters under different geological conditions, such as the significant fluctuations of key parameters such as thrust and cutterhead torque, as well as the distribution of propulsion speed and soil bin pressure under different geological conditions ( Figure 2 ).

[0167] In order to better understand the data structure, this case performed principal component analysis on the data to reduce the dimensionality. The results of the analysis showed that the first five principal components cumulatively explained more than 95% of the total variance, and the first principal component explained more than 50% of the variance ( Figure 2 ). This means that these five principal components can capture most of the information in the data, so we use the five principal components after dimensionality reduction for subsequent clustering analysis.

[0168] In addition, the original data and the data after principal component analysis dimensionality reduction were clustered by K-means++ clustering algorithm and hierarchical clustering algorithm to identify the stratum categories. In the clustering process, the elbow method and hierarchical tree were used to determine the optimal number of clusters. The results show that the three clustering models can better balance the compactness and separation of clusters ( Figure 3 ).

[0169] The K-means++ algorithm was used to perform cluster analysis on the original data and principal component analysis data. The results showed that the two had a high degree of consistency in identifying stratigraphic categories ( Figure 4 ). In particular, in the three-category clustering results of the original data and the reduced-dimensional data, the recognition rate reached 89.54%, indicating that the K-means++ algorithm can maintain a high clustering stability in the original data and the reduced-dimensional data. However, the recognition rate of hierarchical clustering is relatively low, and the recognition rate of the original data is only 18.63% ( Figure 5 (a)), the recognition rate of principal component analysis data is slightly improved to 20.60% ( Figure 5 (b))). This shows that the prediction results of hierarchical clustering in some areas deviate greatly from the actual stratum categories, especially in the intervals of 0 to 200 rings and 600 to 1000 rings of the tunnel, where the clustering results fluctuate greatly and are inconsistent with the actual stratum.

[0170] Based on the unsupervised clustering algorithm, the present invention combines K-means++ and hierarchical clustering methods to conduct an in-depth analysis of the stratum identification of shield tunnels. The results show that the K-means++ algorithm shows a high recognition rate and stability in identifying stratum categories, while the hierarchical clustering performs relatively poorly under complex stratum conditions. The data processed by principal component analysis and dimensionality reduction can simplify the computational complexity while retaining the main geological feature information, which helps to improve the efficiency and accuracy of stratum identification.

[0171]

[0172] It should be noted that the embodiments of the present invention can be implemented by hardware, software or a combination of software and hardware. The hardware part can be implemented using dedicated logic; the software part can be stored in a memory and executed by an appropriate instruction execution system, such as a microprocessor or dedicated hardware. A person of ordinary skill in the art will understand that the above-mentioned devices and methods can be implemented using computer executable instructions and / or contained in a processor control code, for example, such codes are provided on a carrier medium such as a disk, CD or DVD-ROM, a programmable memory such as a read-only memory (firmware), or a data carrier such as an optical or electronic signal carrier. The device and its modules of the present invention can be implemented by hardware circuits such as very large-scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, etc., or programmable hardware devices such as field programmable gate arrays, programmable logic devices, etc., can also be implemented by software executed by various types of processors, or can be implemented by a combination of the above-mentioned hardware circuits and software, such as firmware. The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any modification, equivalent substitution and improvement made by any technician familiar with the technical field within the technical scope disclosed by the present invention and within the spirit and principle of the present invention should be covered within the protection scope of the present invention.

Claims

1. A shield parameter geological identification calculation method based on unsupervised clustering algorithm, characterized in that: The method includes: S1: First, select several groups of data containing factors that affect the identification of stratigraphic categories; PCA transforms a set of complex related variables into fewer and unrelated variables. These new variables are called principal components, which can capture most of the information in the original data; S2: The elbow method is used to determine the number of clusters of K-means++; S3: Using hierarchical tree to determine the number of clusters in hierarchical clustering method; S4: Use K-means++ and hierarchical clustering to cluster the data respectively, and calculate the evaluation indicators such as recognition accuracy or silhouette coefficient of the two to more comprehensively analyze and mine the potential patterns in the data; by comparing the clustering accuracy of the two algorithms, we can highlight the effectiveness of K-means++ when processing large-scale data with high speed requirements.

2. The shield parameter geological identification calculation method based on unsupervised clustering algorithm according to claim 1 is characterized in that: The S1 specifically includes: (1) Given a p×n-dimensional data matrix X, where each sample p is n-dimensional, the matrix can be decomposed using singular value decomposition as X = U∑V T ; In this decomposition, U is a p×n orthogonal matrix, V is an n×n orthogonal matrix, and Σ is a diagonal matrix of dimension n×n; (2) The orthogonal matrix satisfies UU T =1 and VV T =1; (3) For dimensionality reduction, divide Σ into r columns, Σ r , the data after dimension reduction is (4) Based on the above analysis, the cumulative explained variance ratio of each principal component is plotted to observe the cumulative explained rate of variance. When the cumulative explained variance reaches a certain threshold, such as 80% or 90%, it can be considered that these principal components can explain most of the data changes. Usually, the number of principal components corresponding to the cumulative explained variance reaching the threshold is selected.

3. The shield parameter geological identification calculation method based on unsupervised clustering algorithm as claimed in claim 1, characterized in that: The S2 specifically includes: (1) Data preparation: Standardize the original data to eliminate the influence of different dimensions on the clustering results; (2) Initialize cluster centers: Use the K-means++ algorithm to randomly select initial cluster centers to ensure a reasonable distribution of initial points and improve clustering efficiency; (3) Calculate the sum of squared errors (SSE): For different numbers of clusters, run the K-means++ clustering algorithm in turn and calculate the sum of squared errors corresponding to each number of clusters; (4) Draw the SSE curve: Draw the SSE values ​​of different cluster numbers into a curve, with the horizontal axis being the number of clusters and the vertical axis being the SSE value; (5) Determine the "elbow point": According to the change in the decreasing speed of the SSE curve, find the point with the largest decreasing rate, the so-called "elbow point", at which the number of clusters is the optimal value; (6) Apply the optimal number of clusters: Use the optimal number of clusters found to run the K-means++ algorithm again to obtain the final clustering result.

4. The shield parameter geological identification calculation method based on unsupervised clustering algorithm as claimed in claim 1, characterized in that: The S3 specifically includes: (1) Calculate the distance or similarity matrix between all data points. Commonly used distance measurements include Euclidean distance; (2) Initialize the clustering process by treating each data point as an independent cluster; (3) Merge the nearest clusters according to Ward and update the distance matrix until all data points are merged into one cluster; (4) Construct a hierarchical tree based on the aggregation process. Each branch represents a merge, and the height of the tree represents the distance or similarity measure at the time of the merge; (5) When constructing the data tree, the Euclidean distance is used to calculate the distance between samples, and the average linkage method is used to assemble the hierarchical tree structure.

5. A shield parameter geological identification calculation system based on an unsupervised clustering algorithm based on the shield parameter geological identification calculation method based on an unsupervised clustering algorithm as described in claims 1-4, characterized in that: The system specifically includes: In the dimensionality reduction analysis module, first, several groups of data containing factors that affect the identification of stratigraphic categories are selected; PCA transforms a set of complex related variables into fewer and unrelated variables. These new variables are called principal components, which can capture most of the information in the original data; A cluster number determination module, connected to the dimension reduction analysis module, uses the elbow method to determine the number of clusters of K-means++ and uses the hierarchical tree to determine the number of clusters of the hierarchical clustering method; The clustering module is connected to the cluster number determination module, which uses K-means++ and hierarchical clustering to cluster the data respectively, and calculates the evaluation indicators such as recognition accuracy or silhouette coefficient of the two algorithms to more comprehensively analyze and mine the potential patterns in the data; by comparing the clustering accuracy of the two algorithms, it can highlight the effectiveness of K-means++ in processing large-scale data with high speed requirements.

6. A computer device, characterized in that: The computer device includes a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the shield parameter geological identification calculation method based on an unsupervised clustering algorithm.

7. A computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, the processor executes the steps of the shield parameter geological identification calculation method based on an unsupervised clustering algorithm as described in any one of claims 1 to 4.

8. An information data processing terminal, characterized in that: The information data processing terminal is used to implement the shield parameter geological identification calculation system based on the unsupervised clustering algorithm as described in claim 5.

Images