A method and system for analyzing the influence of TBM operating parameters based on multi-source data

Abstract

The application discloses a kind of based on the influence analysis method and system of TBM operating parameter of multi-source data, it is related to parameter analysis technical field, including, using double-layer graph neural network to carry out depth analysis to multi-source data.The relationship between TBM operating parameter and environmental data is analyzed by the parameter graph of the first graph layer, the relationship between TBM effect data and geological data is analyzed by the observation data graph of the second graph layer.Crossover graph convolution mechanism is associated with the node of two layers in depth, the influence relationship between each parameter node and effect data node is quantified, and the influence of each parameter on TBM effect is generated.The change sensitivity of TBM effect data is evaluated by selecting parameter combination and calculating influence intensity through human-computer interaction interface.This method can improve the operating efficiency of TBM, prolong the service life of equipment, and reduce maintenance cost, and has higher practical application value.

Description

Technical Field

[0001] This invention relates to the field of parameter analysis technology, and in particular to a method and system for analyzing the impact of TBM operating parameters based on multi-source data. Background Technology

[0002] With the widespread application of tunnel boring machines (TBMs) in tunnel construction, improving their operating efficiency and accuracy has become a pressing issue in the engineering field. When TBMs operate under complex geological and environmental conditions, their performance is affected by various factors, such as geological structure, soil type, temperature, and humidity. Traditional TBM parameter adjustment methods typically rely on experience or fixed model settings, making it difficult to adapt to changes in different working conditions and geological conditions in real time. This leads to unstable TBM tunneling efficiency and even problems such as equipment damage and excessive wear.

[0003] In recent years, with the rapid development of artificial intelligence (AI) technology, especially the advantages of graph neural networks (GNNs) in graph data processing, scholars have proposed data-driven methods for optimizing TBM operating parameters. Traditional GNN-based models typically focus on a single data source, making it difficult to comprehensively consider the complex relationships between TBM operating parameters, geological data, environmental data, and TBM performance data. Therefore, how to integrate multi-source data to comprehensively and accurately analyze the impact of TBM operating parameters on tunneling performance has become an important research direction. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a method for analyzing the impact of TBM operating parameters based on multi-source data to solve the problem that existing TBM operating parameter optimization methods cannot effectively integrate multi-source data and dynamically adjust parameters in real time.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] In a first aspect, the present invention provides a method for analyzing the impact of TBM operating parameters based on multi-source data, which includes collecting multi-source data of TBM, including TBM operating parameters, environmental data, geological data and TBM effect data;

[0008] Deep analysis of real-time multi-source data using a two-layer graph neural network:

[0009] The two-layer graph neural network includes a parameter graph of the first layer and an observation data graph of the second layer;

[0010] In the parameter graph of the first layer, each TBM running parameter is used as a node. The edge relationships are adjusted using environmental data, and the influence relationship of each parameter on other parameters is generated by passing the data.

[0011] In the observation data map of the second layer, each TBM effect data is used as a node, and the edge relationship is adjusted using geological and environmental data. The influence relationship of each effect data on other effects is generated by passing the data.

[0012] By using the cross-graph convolution mechanism in the two-layer graph neural network, the nodes in the parameter graph of the first layer and the observation data graph of the second layer are deeply correlated, the combined influence between each parameter node and the observation data node is quantified, and the influence of each parameter on the TBM effect is generated.

[0013] As a preferred embodiment of the TBM operating parameter impact analysis method based on multi-source data described in this invention, the TBM operating parameters include: power, thrust, torque, cutterhead rotation speed, and propulsion speed.

[0014] The environmental data includes temperature, humidity, air pressure, and groundwater content;

[0015] The geological data includes rock layer hardness, lithological classification, soil particle size, and integrity index;

[0016] The TBM performance data includes tool wear, tunneling distance, rock breaking efficiency, energy consumption, vibration level, and tool life.

[0017] As a preferred embodiment of the TBM operating parameter impact analysis method based on multi-source data described in this invention, the parameter graph of the first layer includes: taking each TBM operating parameter as a node, taking the physical relationship between the parameters as an edge, and the weight of the edge representing the coupling strength between the parameters. Through two-step training, a parameter impact relationship graph adapted to different environmental conditions is obtained.

[0018] The first step of the training is to train the physical correlation between nodes in a standard environment to generate a basic parametric relationship graph. Using the basic parametric relationship graph, the second step is to train the edge weights using data from different environments, so that the graph structure can dynamically adjust the edge relationships in different environments.

[0019] As a preferred embodiment of the TBM operation parameter impact analysis method based on multi-source data described in this invention, the second layer of the observation data map includes: taking each TBM effect data as a node, taking the relationship between different effects as an edge, and the weight of the edge representing the correlation strength between different effect data. Through two-step training, an impact relationship map of the observation data is obtained under working conditions that are adapted to different geological and environmental data.

[0020] The first step of the training is to train the physical correlation between nodes in a standard environment to generate a basic observation data relationship graph. Based on the training in the first step, the edge weights are trained using different geological and environmental data so that the graph structure can dynamically adjust the edge relationships under different geological and environmental data.

[0021] As a preferred embodiment of the TBM operation parameter impact analysis method based on multi-source data described in this invention, the multi-source data collected in real time is filtered using the two-layer graph structure in the two-layer graph neural network, so that each graph structure only inputs trained data types, thereby obtaining the influence relationship of each parameter on other parameters at the current time, and the influence relationship of each effect data on other effects at the current time.

[0022] The influence relationship is reflected by adjusting the weight of the edges to show the degree of mutual influence between nodes;

[0023] The influence relationship refers to the relationship between changes in other nodes in the same layer when a node changes, calculated by a graph convolutional neural network. This means that a change in one node propagates within the same layer, thereby causing changes in the data of other nodes.

[0024] As a preferred embodiment of the TBM operation parameter impact analysis method based on multi-source data described in this invention, the cross-graph convolution mechanism includes: transferring information between the parameter graph of the first layer and the observation data graph of the second layer, and using a cross-convolution kernel to fuse the parameter node features of the first layer and the effect node features of the second layer;

[0025] The cross-convolution kernel couples the changes of each parameter node in the first layer with the influence features of each effect data node in the second layer through node feature alignment and weighted fusion, generating the influence relationship between the nodes of the two layers.

[0026] As a preferred embodiment of the TBM operating parameter impact analysis method based on multi-source data described in this invention, the impact manifestation includes: selecting the TBM operating parameters or parameter combinations to be analyzed through a human-computer interaction interface and selecting parameter values; obtaining parameter combinations after relationship coupling through the analysis of the parameter graph in the two-layer graph neural network; and analyzing the parameter combinations after relationship coupling using the cross-graph convolution mechanism to obtain the change in the effect data of each TBM.

[0027] Under the parameter values ​​selected in the human-computer interaction interface, the influence intensity is obtained by dividing the change in TBM effect data by the maximum value of TBM effect data, which represents the sensitivity of the change in TBM effect data.

[0028] When analyzing the changes in the TBM effect data, self-verification is performed using the two-layer graph neural network:

[0029] Step 1: Assume that the parameter value selection is valid during human-computer interaction, and obtain the change in each TBM effect data;

[0030] Step 2: Analyze the relationship between each effect data using the observation data graph of the second layer, whether it conforms to the edge coupling relationship, and output the probability of conformity;

[0031] Step 3: Optimize the model parameters in the cross-graph convolution mechanism: Based on the difference between each sample in each training set and the current working condition, gradually reduce the attention from the sample with the smallest difference to the sample with the largest difference according to the degree of difference, and optimize the model; use the optimized cross-graph convolution mechanism to recalculate the change in each TBM effect data; repeat the optimization until the probability of the output of step 2 no longer increases.

[0032] The optimized model parameters are applied to the cross-graph convolution mechanism to output the corresponding impact performance.

[0033] Secondly, the present invention provides a TBM operating parameter impact analysis system based on multi-source data, including a data acquisition unit for acquiring multi-source data of the TBM, including TBM operating parameters, environmental data, geological data and TBM effect data;

[0034] The analysis unit performs in-depth analysis of real-time multi-source data through a two-layer graph neural network: the two-layer graph neural network includes a parameter graph of the first layer and an observation data graph of the second layer;

[0035] The output unit, through the cross-graph convolution mechanism in the two-layer graph neural network, deeply correlates the nodes in the parameter graph of the first layer with the observation data graph of the second layer, quantifies the combined influence between each parameter node and the observation data node, and generates the influence of each parameter on the TBM effect.

[0036] Thirdly, the present invention provides a computer device including a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, it implements any step of the method for analyzing the impact of TBM operating parameters based on multi-source data as described in the first aspect of the present invention.

[0037] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the method for analyzing the impact of TBM operating parameters based on multi-source data as described in the first aspect of the present invention.

[0038] The beneficial effects of this invention are as follows: By using a two-layer graph neural network to perform in-depth analysis of multi-source TBM data, the interaction between TBM operating parameters and geological, environmental, and performance data can be accurately quantified. This method uses a cross-graph convolution mechanism to deeply correlate nodes in the parameter graph of the first layer and the observation data graph of the second layer, thereby quantifying the impact of each parameter on the TBM's performance. Through two-step training, the model can adapt to dynamic changes under different environments and working conditions, ensuring real-time adjustment of TBM operating parameters under various geological and environmental conditions, optimizing key parameters such as thrust, torque, and cutterhead speed, and improving tunneling efficiency and equipment lifespan. This invention, through a human-machine interface, allows operators to flexibly select the parameter combinations to be analyzed and perform self-verification and optimization, ensuring the accuracy and reliability of each analysis result. It not only provides a more precise and intelligent control strategy but also significantly reduces maintenance costs and improves work efficiency, showing broad engineering application prospects. Attached Figure Description

[0039] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0040] Figure 1 This is a flowchart of a method for analyzing the impact of TBM operating parameters based on multi-source data.

[0041] Figure 2 This is a two-layer graph neural network diagram for an impact analysis method of TBM operating parameters based on multi-source data.

[0042] Figure 3 This is a computer equipment diagram illustrating a method for analyzing the impact of TBM operating parameters based on multi-source data. Detailed Implementation

[0043] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0044] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0045] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0046] Example 1, referring to Figure 1 and Figure 2 This is one embodiment of the present invention, which provides a method for analyzing the impact of TBM operating parameters based on multi-source data, including the following steps:

[0047] S1: Collect multi-source data of TBM, including TBM operating parameters, environmental data, geological data and TBM effect data;

[0048] Specifically, the TBM operating parameters include power, thrust, torque, cutterhead rotation speed, and propulsion speed; the environmental data include temperature, humidity, air pressure, and groundwater content; the geological data include rock hardness, lithology classification, soil particle size, and integrity index; and the TBM performance data include cutter wear, tunneling distance, rock breaking efficiency, energy consumption, vibration level, and cutter life.

[0049] S2: Deep analysis of real-time multi-source data using a two-layer graph neural network:

[0050] like Figure 2 As shown, the two-layer graph neural network includes a parameter graph of the first layer and an observation data graph of the second layer.

[0051] In the parameter graph of the first layer, each TBM running parameter is used as a node. The edge relationships are adjusted using environmental data, and the influence relationship of each parameter on other parameters is generated by passing the parameters.

[0052] In the observation data map of the second layer, each TBM effect data is used as a node. The edge relationships are adjusted using geological and environmental data, and the influence relationship of each effect data on other effects is generated by passing it.

[0053] Furthermore, the parameter graph of the first layer includes each TBM operating parameter as a node, the physical relationships between parameters as edges, and the edge weights representing the coupling strength between parameters. Through two-step training, a parameter influence relationship graph adapted to different environmental conditions is obtained. The first step of training trains the physical correlation between nodes in a standard environment to generate a basic parameter relationship graph. Using this basic parameter relationship graph, the second step trains the edge weights using data from different environments, enabling the graph structure to dynamically adjust edge relationships under different conditions.

[0054] Specifically, the two-step training refers to a two-stage training process when constructing the first layer of the parametric graph. This ensures that the graph neural network accurately reflects the physical relationships between parameters and adapts to different environments. The goal of the first training step is to learn and construct the basic physical correlations between parameters under standard conditions (i.e., ideal geological and environmental conditions). In this stage, the nodes in the first layer of the parametric graph represent TBM operating parameters (such as thrust, torque, cutterhead speed, and feed rate), and the edges between nodes represent the physical relationships between parameters, such as the coupling relationship between thrust and torque, or the influence between cutterhead speed and feed rate. Physical models (such as finite element analysis models) and historical data are used to define the weights of the edges. These edge weights are set to fixed values ​​under standard conditions to reflect the physical coupling between parameters. For example, the relationship between thrust and torque can be predicted using a mechanical model under standard conditions, thus generating a preliminary graph structure. A graph convolutional neural network (GCN) is used to train the graph structure under standard environmental conditions. Through the GCN, information between nodes is propagated, and the influence of each parameter node under standard conditions is calculated. After the first training step, a basic parameter relationship graph is obtained. At this point, the graph structure can represent the physical coupling relationships between various TBM operating parameters under standard environmental conditions. The goal of the second training step is to build upon the first step by using environmental data (such as temperature, humidity, rock hardness, etc.) and geological data to train the dynamic adjustment of edge weights, enabling the graph structure to adapt to changes under different environmental conditions and reflect more realistic physical coupling relationships. In the second training step, environmental data (such as temperature, humidity, air pressure, etc.) and geological data (such as rock hardness, soil type, etc.) under different environmental conditions are added. Based on this environmental data, the edge weights are adjusted according to environmental changes. For example, rock hardness may increase the impact of thrust on tool wear, while humidity may affect tool friction and propulsion efficiency. In the second training step, GCN uses this additional data to dynamically adjust edge relationships. For example, the impact of thrust on tunneling efficiency differs in soft soil environments compared to hard rock environments; the network adjusts the influence relationships of relevant parameters based on this environmental data. Reinforcement learning or other adaptive algorithms can be used to adjust the weights of each edge to ensure that the graph network can dynamically adapt to different working conditions and geological environments. After the second training step is completed, the graph structure can adaptively adjust under different environments, so that the relationship between each node (parameter) and other nodes can more accurately reflect the degree of influence under actual working conditions, thereby enabling the graph neural network to better simulate the performance of TBM under complex and changing geological and environmental conditions.

[0055] It's also important to know that the second layer of the observation data graph includes each TBM effect data point as a node, the relationships between different effects as edges, and the edge weights representing the correlation strength between different effect data. Through two-step training, an influence relationship graph of the observation data is obtained that adapts to different geological and environmental data conditions. The first step of training trains the physical correlation between nodes under a standard environment, generating a basic observation data relationship graph. Based on the first step, the edge weights are trained using different geological and environmental data, enabling the graph structure to dynamically adjust edge relationships under different geological and environmental data conditions.

[0056] Through a two-step training process, combining standard environmental data and actual working condition data, a Graph Neural Network (GNN) is enabled to accurately reflect the interrelationships between TBM performance data and adapt to different geological and environmental conditions. The specific process is as follows: The first training step aims to train the physical correlations between nodes under standard environmental conditions and generate a basic observation data relationship graph. Using standard environmental data (such as standard temperature, humidity, rock hardness, etc.), nodes are established in the graph for each TBM performance data point (such as tool wear, tunneling distance, rock breaking efficiency, energy consumption, etc.), and edge relationships between each performance data node are defined. These edges represent the basic physical associations between performance data (e.g., the relationship between tool wear and thrust). Through training with a Graph Convolutional Neural Network (GCN), the network learns the influence relationships between these nodes and edges, generating a preliminary observation data relationship graph reflecting the basic influence relationships under standard environmental conditions. The second training step aims to build upon the first step by training the edge weights using different environmental and geological data, enabling the graph structure to adapt to changes under different environmental conditions. Specifically, environmental data (such as temperature, humidity, and groundwater content) and geological data (such as rock hardness, soil particle size, and integrity index) are introduced into the model to dynamically adjust the base graph model generated in the first step. Through training, the edge weights are fine-tuned based on changes in different environmental and geological data. For example, increased rock hardness may lead to accelerated tool wear, thus affecting the edge relationships between tool wear nodes and other effect data nodes. The graph convolutional neural network further trains the adjusted graph structure, propagating the updated edge relationships to make the influence features of each node more accurate and consistent with changes under actual working conditions.

[0057] Through these two training steps, the system can establish a basic observation data relationship diagram under standard conditions and adjust it using environmental and geological data under actual working conditions to ensure that the model can adapt to different working environments. This process improves the accuracy and adaptability of the TBM operating parameter impact analysis method based on multi-source data, enabling the method to provide accurate impact analysis under various environmental and geological conditions.

[0058] Furthermore, by utilizing the two-layer graph structure in the aforementioned two-layer graph neural network, the multi-source data collected in real time is filtered, ensuring that each graph structure only receives trained data types. This yields the influence relationship between each parameter and other parameters at the current moment, as well as the influence relationship between each effect data point and other effects at the current moment. The influence relationship reflects the degree of mutual influence between nodes through edge weight adjustments. This influence relationship refers to the relationship between changes in other nodes within the same layer calculated by the graph convolutional neural network when a node changes; it indicates that a change in one node propagates within the same layer, thereby causing changes in the data of other nodes.

[0059] S3: Through the cross-graph convolution mechanism in the dual-layer graph neural network, the nodes in the parameter graph of the first layer and the observation data graph of the second layer are deeply correlated, the combined influence between each parameter node and the observation data node is quantified, and the influence of each parameter on the TBM effect is generated.

[0060] The cross-graph convolution mechanism involves transferring information between the parameter graph of the first layer and the observation data graph of the second layer, and fusing the parameter node features of the first layer with the effect node features of the second layer using a cross-convolution kernel. The cross-convolution kernel couples the changes of each parameter node in the first layer with the influence features of each effect data node in the second layer through node feature alignment and weighted fusion, generating the influence relationship between the nodes of the two layers.

[0061] The goal of the cross-graph convolution mechanism is to deeply fuse the parameter nodes of the first layer (such as thrust, torque, etc.) and the effect data nodes of the second layer (such as tool wear, tunneling efficiency, etc.) to capture their mutual influence. Through convolution, node features can be fused across layers, enhancing the network's expressive power. In this mechanism, the parameter node features of the first layer and the effect data node features of the second layer must first be aligned. This process ensures that node features from different levels can be represented uniformly, facilitating subsequent convolution operations. Node feature alignment can be accomplished using a mapping matrix or feature transformation network, ensuring that the node features of the two layers operate in the same feature space. Next, weighted fusion is performed, which is the core of cross-convolution. The features of each node are fused, where the fusion weights are determined by the physical relationship and similarity between nodes. Through weighted averaging or weighted summation, the parameter node features of the first layer and the effect data node features of the second layer are fused, generating the influence relationships in the two-layer graph structure.

[0062] Convolutional operations are used to propagate information within graph structures. The convolutional kernel aggregates information based on the features of neighboring nodes and updates the node features. In the cross-graph convolution mechanism, the convolutional kernel operates on two layers, passing and fusing the features of each parameter node and the effect data node to obtain updated influence features. After multiple rounds of convolution, the node features gradually stabilize, and the learning performance of the graph neural network gradually improves. To further enhance the expressive power of features, a fully connected layer can be added after the convolutional operation to map the convolutional node features to a higher-dimensional feature space. After processing by the fully connected layer, the influence relationships between nodes are expressed more accurately.

[0063] Ultimately, the output influence relationships represent the mutual influence between each parameter node (such as thrust, torque, etc.) and the effect data nodes (such as tool wear, tunneling distance, etc.). This influence relationship can be represented by the node's influence vector, reflecting the strength of each parameter node's influence on the effect data node, or by an correlation matrix, where each element represents the strength of the influence between the parameter node and the effect data node. This mechanism enables graph neural networks to effectively quantify the interaction between TBM parameters and effect data, providing a basis for subsequent TBM parameter optimization.

[0064] The workflow of cross-graph convolution can be summarized as follows: input data, the node features of the first layer and the second layer represent the TBM running parameters and effect data respectively; the nodes of the two layers are aligned and weighted and fused through cross-convolution kernels; the node features are propagated and updated through graph convolution operations; the fully connected layer further maps the node features; finally, the influence relationship between the parameter nodes and the effect data nodes is output, and these relationships are quantified to support subsequent TBM parameter optimization.

[0065] The impact is manifested by selecting the TBM operating parameters or parameter combinations to be analyzed through a human-computer interaction interface and selecting parameter values; obtaining the parameter combinations after relationship coupling through the analysis of the parameter graph in the two-layer graph neural network; and analyzing the parameter combinations after relationship coupling using the cross-graph convolution mechanism to obtain the change in the effect data of each TBM.

[0066] Under the parameter values ​​selected in the human-computer interaction interface, the influence intensity is obtained by dividing the change in TBM effect data by the maximum value of TBM effect data, which represents the sensitivity of the change in TBM effect data.

[0067] When analyzing the changes in the TBM effect data, self-verification is performed using the two-layer graph neural network:

[0068] Step 1: Assume that the parameter value selection is valid during human-computer interaction, and obtain the change in each TBM effect data.

[0069] Step 2: Analyze the relationships between each effect data point using the observation data map of the second layer, whether they conform to the edge coupling relationship, and output the probability of conformity. By verifying in real time whether the changes in TBM effect data after each parameter adjustment meet expectations, the system ensures that the impact of each parameter adjustment is effective and reasonable. Through this verification mechanism, the system can detect the correctness of parameter adjustments in real time and automatically identify any erroneous adjustments or effects that do not conform to actual working conditions. If the effect data fails to respond as expected, the system can immediately provide feedback and make corrections. This verification mechanism ensures the system's real-time response capability, enabling parameter optimization to adapt in real time to changes in the influence of geological, environmental, and other variables, thereby improving the system's intelligence level and dynamic adaptability. The design purpose of using probability as a basis is to quantify the influence intensity of each parameter on the TBM effect by evaluating the probability of changes in effect data after parameter adjustments. By calculating the conformity probability, the system can determine whether the adjusted parameter changes conform to the actual working condition change pattern, thereby optimizing the calculation of influence intensity and making the influence of each parameter on the effect data more accurate. Using probability as a basis not only reflects the interrelationship between each parameter and the effect data but also reflects the stability and reliability of adjustments under different working conditions. When the probability of matching reaches a certain threshold, the system will confirm that the parameter combination is the optimal solution; otherwise, it will make adaptive adjustments.

[0070] Step 3: Optimize the model parameters in the cross-graph convolution mechanism: Based on the difference between each sample in each training set and the current working condition, gradually reduce the attention from the sample with the smallest difference to the sample with the largest difference according to the degree of difference, and optimize the model; use the optimized cross-graph convolution mechanism to recalculate the change in each TBM effect data; repeat the optimization until the probability of the output of step 2 no longer increases.

[0071] The optimized model parameters are applied to the cross-graph convolution mechanism to output the corresponding impact performance.

[0072] It's important to understand that in traditional methods, TBM operating parameters are typically adjusted based on experience or optimized using fixed models. This approach often fails to accurately adapt to complex geological and environmental conditions, leading to low TBM operating efficiency or equipment failure. This invention, however, through in-depth analysis and dynamic adjustment, can accurately identify the impact of various parameters on TBM performance under different environmental conditions, thereby achieving real-time parameter optimization.

[0073] The integration of a human-computer interface allows operators to intuitively select and adjust the TBM operating parameters or combinations thereof to be analyzed. Operators can not only adjust parameters according to on-site needs but also quickly understand and participate in the optimization process through the interface, greatly improving the system's operability and flexibility. Simultaneously, the use of a two-layer graph neural network effectively integrates data from different sources. The first layer processes TBM operating parameters and environmental data, while the second layer focuses on performance data. Through a cross-graph convolution mechanism, the two layers are deeply correlated within the same network framework, reflecting the complex coupling relationships between various parameters and performance data.

[0074] Furthermore, the adopted Graph Convolutional Network (GCN) and cross-convolution mechanism ensure that the network can adaptively adjust in complex environments. It can not only learn the basic parameter influence relationships from the standard environment, but also adjust the edge weights under changing working conditions to achieve dynamic optimization. Through this design, the network can accurately calculate the impact of each parameter on TBM performance data (such as tool wear, tunneling efficiency, etc.) and provide real-time and accurate optimization results.

[0075] Furthermore, in other optional embodiments, the impact analysis results generated by the two-layer graph neural network can be combined with multiple optimization objectives (such as tunneling efficiency, tool wear, energy consumption, etc.) to generate optimized thrust and torque analysis results. The thrust and torque adjustment scheme can then be optimized using multi-objective optimization algorithms (such as genetic algorithms, particle swarm optimization, etc.) to ensure the optimal operating state of the system. Moreover, through an optimization control framework combining adaptive PID control and deep reinforcement learning (DRL), the thrust and torque are dynamically adjusted based on real-time feedback to adapt to TBM operation under different geological and environmental conditions, ensuring optimal tunneling efficiency and equipment maintenance.

[0076] This embodiment also provides a TBM operating parameter impact analysis system based on multi-source data, including: a data acquisition unit for acquiring multi-source data of the TBM, including TBM operating parameters, environmental data, geological data and TBM effect data.

[0077] The analysis unit performs in-depth analysis of real-time multi-source data through a two-layer graph neural network: the two-layer graph neural network includes a parameter graph of the first layer and an observation data graph of the second layer.

[0078] The output unit, through the cross-graph convolution mechanism in the two-layer graph neural network, deeply correlates the nodes in the parameter graph of the first layer with the observation data graph of the second layer, quantifies the combined influence between each parameter node and the observation data node, and generates the influence of each parameter on the TBM effect.

[0079] This embodiment also provides a computer device applicable to the TBM operating parameter impact analysis method based on multi-source data, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the TBM operating parameter impact analysis method based on multi-source data as proposed in the above embodiment.

[0080] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0081] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements the method for analyzing the impact of TBM operating parameters based on multi-source data as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0082] In summary, this invention addresses the problem of analyzing the influence relationships between TBM operating parameters and between these parameters and TBM performance data by combining a two-layer graph neural network and a cross-graph convolution mechanism. Through the parameter graph of the first layer and the observation data graph of the second layer, the system can deeply analyze the mutual influence relationships between TBM operating parameters (such as thrust, torque, cutterhead speed, etc.) and TBM performance data (such as cutter wear, tunneling distance, rock breaking efficiency, etc.). Utilizing the cross-graph convolution mechanism, the node features in the two-layer graph structure are fused, thereby accurately quantifying the impact of parameter changes on performance data and providing a representation of the impact of each parameter on the TBM performance.

[0083] Example 2 is an embodiment of the present invention, which provides a method for analyzing the impact of TBM operating parameters based on multi-source data. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments.

[0084] In this embodiment, an open tunnel boring machine (TBM) at a subway tunnel construction site in a certain city was selected as the experimental object. The tunnel section was about 1.6 km long and the burial depth ranged from 18 m to 32 m. The strata were mainly moderately weathered sandstone and local mudstone interlayers. The geological conditions had obvious non-uniformity and staged variation characteristics, which was suitable for verifying the effectiveness of multi-source data coupling modeling.

[0085] During TBM construction, a multi-source data acquisition system is built to simultaneously collect the following data:

[0086] TBM operating parameter data: including propulsion thrust, cutterhead torque, cutterhead rotation speed, propulsion speed, and main drive power, with a sampling period of 1 second. Environmental data: including temperature, humidity, air pressure, and groundwater content changes in the tunnel chamber, with a sampling period of 10 seconds. Geological data: including rock hardness index, lithology classification code, soil particle size distribution index, and integrity index, derived from advanced geological forecasts and historical borehole data, mapped by tunneling mileage. TBM performance data: including cutter wear per unit mileage, tunneling distance per unit time, rock breaking efficiency, energy consumption per cubic meter, and vibration level, with a sampling period of 1 second.

[0087] Over 30 consecutive construction days, approximately 2.6 million valid data points were collected. After removing data from abnormal shutdowns and non-tunneling conditions, a dataset was formed for model training and validation.

[0088] After data preparation, the TBM operating parameters were first analyzed independently according to this method. Using propulsion thrust, cutterhead torque, cutterhead rotation speed, propulsion speed, and main drive power as parameter nodes, initial correlations between parameters were constructed based on historical construction data. Before introducing environmental data, training on standard operating condition data revealed that when the propulsion thrust varied between 8000kN and 8600kN, the synchronous change in cutterhead torque was approximately 6% to 9%, while the change in propulsion speed did not exceed 3%. This result was used to generate the basic correlation structure between parameters. Based on this, real-time environmental data was introduced into the analysis process. When the groundwater content increased by approximately 12% compared to the baseline condition and the humidity in the tunneling chamber increased by approximately 18%, the system detected a significantly enhanced increase in the response of thrust changes to torque changes in the parameter relationships. Its influence weight increased by approximately 15% compared to the standard condition, while the influence weight of cutterhead rotation speed on propulsion speed decreased by approximately 9%.

[0089] It should be noted that the above-mentioned change process was not manually set, but was automatically formed under the drive of continuous tunneling data, indicating that the mutual influence between parameters will be dynamically adjusted according to environmental conditions.

[0090] After modeling the relationships between operating parameters, the TBM performance data was further analyzed independently. Tool wear, tunneling efficiency, energy consumption, vibration level, and tool life were used as the analysis objects, and in the initial stage, training was conducted solely based on data from standard geological sections.

[0091] The results show that in sections with a rock hardness index of approximately 65, there is a strong synchronous variation between cutter wear per unit mileage and energy consumption per unit volume, with the correlation strength remaining stable at around 0.7. As tunneling progresses into sections with a hardness index exceeding 80, this correlation strength gradually increases to approximately 0.84, while the impact of vibration intensity on cutter wear begins to significantly increase. Through this process, the system has formed an observational data structure reflecting the changes in the internal relationships of the performance data under different geological conditions.

[0092] After the two types of relationships mentioned above have stabilized, the operational parameter relationships and effect data relationships are jointly analyzed according to this method. A representative working condition period during the tunneling process is selected, and the operator sets a parameter variation scheme in the system interface: increasing the propulsion thrust by 8%, decreasing the cutterhead rotation speed by 5%, and keeping the propulsion speed unchanged. The system does not directly output the predicted tunneling results, but first calculates the propagation result of this parameter change in the parameter relationship structure, and then maps the propagation result to the effect data relationship. The analysis results show that under this parameter combination, the cutter wear per unit mileage is expected to increase by about 6.2%, the tunneling efficiency is expected to increase by about 3.5%, the energy consumption per unit volume is expected to decrease by about 2.1%, and the vibration magnitude variation is controlled within ±1.3%.

[0093] To verify the reliability of the above analysis results, actual tunneling was conducted using the same parameter combination in a section with the same geological conditions, and the actual construction data were compared and statistically analyzed. The results show that the actual measured tool wear increased by approximately 6.8%, tunneling efficiency improved by approximately 3.2%, and energy consumption decreased by approximately 2.4%, with deviations from the analysis results all controlled within 0.6%. During this process, the system simultaneously performed consistency checks on the relationships between the effect data. Initially, the probability of agreement between the effect relationships was 0.81, lower than the preset threshold. The system then automatically reduced the weight of samples with significant differences based on the differences between the current working conditions and historical samples. After three rounds of data updates, the probability of agreement between the effect relationships increased to 0.89, and subsequent analysis results tended to stabilize.

[0094] Comparing the construction effects of this method with traditional experience-based parameter adjustment over 15 consecutive tunneling cycles, statistical results show that, without increasing construction risks, the method reduces the number of cutter replacements per unit mileage by approximately 12%, increases average tunneling efficiency by approximately 4.1%, reduces unit energy consumption by approximately 3.6%, and decreases the frequency of abnormal vibration events by approximately 18%. These results demonstrate that this method, through joint analysis of multi-source data in real-world construction scenarios, can objectively reflect the impact of TBM operating parameter changes on tunneling performance, providing verifiable and traceable decision-making basis for parameter adjustment.

[0095] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for analyzing the impact of TBM operating parameters based on multi-source data, characterized in that: This includes collecting multi-source data on the TBM, including TBM operating parameters, environmental data, geological data, and TBM performance data; Deep analysis of real-time multi-source data using a two-layer graph neural network: The two-layer graph neural network includes a parameter graph of the first layer and an observation data graph of the second layer; In the parameter graph of the first layer, each TBM running parameter is used as a node. The edge relationships are adjusted using environmental data, and the influence relationship of each parameter on other parameters is generated by passing the data. In the observation data map of the second layer, each TBM effect data is used as a node, and the edge relationship is adjusted using geological and environmental data. The influence relationship of each effect data on other effects is generated by passing the data. By using the cross-graph convolution mechanism in the two-layer graph neural network, the nodes in the parameter graph of the first layer and the observation data graph of the second layer are deeply correlated, the combined influence between each parameter node and the observation data node is quantified, and the influence of each parameter on the TBM effect is generated.

2. The method for analyzing the impact of TBM operating parameters based on multi-source data as described in claim 1, characterized in that: The TBM operating parameters include power, thrust, torque, cutterhead speed, and feed rate. The environmental data includes temperature, humidity, air pressure, and groundwater content; The geological data includes rock layer hardness, lithological classification, soil particle size, and integrity index; The TBM performance data includes tool wear, tunneling distance, rock breaking efficiency, energy consumption, vibration level, and tool life.

3. The method for analyzing the impact of TBM operating parameters based on multi-source data as described in claim 2, characterized in that: The parameter graph of the first layer includes each TBM operating parameter as a node, the physical relationship between the parameters as an edge, and the weight of the edge representing the coupling strength between the parameters. Through two-step training, a parameter influence relationship graph adapted to different environmental conditions is obtained. The first step of the training is to train the physical correlation between nodes in a standard environment to generate a basic parametric relationship graph. Using the basic parametric relationship graph, the second step is to train the edge weights using data from different environments, so that the graph structure can dynamically adjust the edge relationships in different environments.

4. The method for analyzing the impact of TBM operating parameters based on multi-source data as described in claim 3, characterized in that: The second layer of observation data map includes each TBM effect data as a node, the relationship between different effects as edges, and the weight of the edges representing the correlation strength between different effect data. Through two-step training, an influence relationship map of observation data under different geological and environmental data conditions is obtained. The first step of the training is to train the physical correlation between nodes in a standard environment to generate a basic observation data relationship graph. Based on the training in the first step, the edge weights are trained using different geological and environmental data so that the graph structure can dynamically adjust the edge relationships under different geological and environmental data.

5. The method for analyzing the impact of TBM operating parameters based on multi-source data as described in claim 4, characterized in that: By utilizing the two-layer graph structure in the dual-layer graph neural network, the multi-source data collected in real time is filtered so that each graph structure only receives the data type that has been trained, thereby obtaining the influence relationship of each parameter on other parameters at the current time, as well as the influence relationship of each effect data on other effects at the current time. The influence relationship is reflected by adjusting the weights of the edges to show the degree of mutual influence between nodes; The influence relationship refers to the relationship between changes in other nodes in the same layer when a node changes, calculated by a graph convolutional neural network. This means that a change in one node propagates within the same layer, thereby causing changes in the data of other nodes.

6. The method for analyzing the impact of TBM operating parameters based on multi-source data as described in claim 5, characterized in that: The cross-graph convolution mechanism includes fusing the parameter node features of the first layer with the effect node features of the second layer by transmitting information between the parameter graph of the first layer and the observation data graph of the second layer using a cross-convolution kernel. The cross-convolution kernel couples the changes of each parameter node in the first layer with the influence features of each effect data node in the second layer through node feature alignment and weighted fusion, generating the influence relationship between the nodes of the two layers.

7. The method for analyzing the impact of TBM operating parameters based on multi-source data as described in claim 6, characterized in that: The impact manifestations include: selecting the TBM operating parameters or parameter combinations to be analyzed through a human-computer interaction interface, and selecting parameter values; obtaining parameter combinations after relationship coupling through the analysis of the parameter graph in the two-layer graph neural network; and analyzing the parameter combinations after relationship coupling using the cross-graph convolution mechanism to obtain the change in the effect data of each TBM. Under the parameter values ​​selected in the human-computer interaction interface, the influence intensity is obtained by calculating the change in TBM effect data and dividing it by the maximum value of TBM effect data, which represents the sensitivity of the change in TBM effect data. When analyzing the changes in the TBM effect data, self-verification is performed using the two-layer graph neural network: Step 1: Assume that the parameter value selection is valid during human-computer interaction, and obtain the change in each TBM effect data; Step 2: Analyze the relationship between each effect data using the observation data graph of the second layer, whether it conforms to the edge coupling relationship, and output the probability of conformity; Step 3: Optimize the model parameters in the cross-graph convolution mechanism: Based on the difference between each sample in each training set and the current working condition, gradually reduce the attention from the sample with the smallest difference to the sample with the largest difference according to the degree of difference, and optimize the model; use the optimized cross-graph convolution mechanism to recalculate the change in each TBM effect data; repeat the optimization until the probability of the output in Step 2 no longer increases. The optimized model parameters are applied to the cross-graph convolution mechanism to output the corresponding impact performance.

8. A system for analyzing the impact of TBM operating parameters based on multi-source data, based on the method for analyzing the impact of TBM operating parameters based on multi-source data according to any one of claims 1 to 7, characterized in that: This includes a data acquisition unit that collects multi-source data from the TBM, including TBM operating parameters, environmental data, geological data, and TBM performance data. The analysis unit performs in-depth analysis of real-time multi-source data through a two-layer graph neural network: the two-layer graph neural network includes a parameter graph of the first layer and an observation data graph of the second layer; The output unit, through the cross-graph convolution mechanism in the two-layer graph neural network, deeply correlates the nodes in the parameter graph of the first layer with the observation data graph of the second layer, quantifies the combined influence between each parameter node and the observation data node, and generates the influence of each parameter on the TBM effect.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the TBM operating parameter impact analysis method based on multi-source data as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the TBM operating parameter impact analysis method based on multi-source data as described in any one of claims 1 to 7.

Images