Tunneling parameter adaptive optimization method and system for pipe jacking construction

Abstract

This invention provides an adaptive optimization method and system for tunneling parameters in pipe jacking construction, belonging to the field of parameter optimization technology. The method includes: real-time multi-dimensional tunneling data acquisition based on pipe jacking construction, standardized processing, and construction of a time-series feature matrix; construction of a geological-parameter coupling model, feature fusion, and generation of a feature fusion dataset; attitude deviation prediction, generating attitude offset prediction values, which are used as constraints for multi-objective optimization of pipe jacking construction to determine the optimal combination of tunneling parameters; simulated tunneling, and control and optimization based on simulated tunneling parameters, generating control commands and sending them to the controller for automatic parameter adjustment. This invention solves the technical problem in existing pipe jacking construction technologies that primarily focus on controlling construction parameters, neglecting the influence of geological factors on the construction process. This leads to the construction process being unable to adapt to complex geological environments, equipment attitude loss, and consequently, decreased construction accuracy and slow construction progress.

Description

Technical Field

[0001] This invention relates to the field of parameter optimization technology, and more specifically to an adaptive optimization method and system for tunneling parameters used in pipe jacking construction. Background Technology

[0002] Pipe jacking is an underground engineering construction method widely used in urban rail transit, underground pipelines and other projects. Pipe jacking technology relies on large pipe jacking machines for soil excavation and pipeline installation. The construction process is complex and involves the adjustment of various parameters. Therefore, how to accurately control these parameters to ensure efficient, safe and precise construction is crucial.

[0003] Traditional pipe jacking construction mainly focuses on controlling construction parameters, while ignoring the impact of geological factors on the construction process, such as soil type, groundwater level, and stratum distribution. It is impossible to combine these changes in geological conditions with construction parameters in a timely manner, which leads to equipment attitude loss or excessive soil reaction during construction, resulting in decreased construction accuracy and slow construction progress. Summary of the Invention

[0004] This application provides an adaptive optimization method and system for tunneling parameters in pipe jacking construction, aiming to solve the technical problem that existing pipe jacking construction mainly focuses on the control of construction parameters, ignoring the influence of geological factors on the construction process, resulting in the construction process being unable to adapt to complex geological environments, equipment attitude loss, and thus causing a decrease in construction accuracy and slow construction progress.

[0005] The first aspect disclosed in this application provides an adaptive optimization method for tunneling parameters in pipe jacking construction. The method includes: real-time multi-dimensional tunneling data acquisition based on pipe jacking construction; standardizing the obtained construction parameter dataset to construct a time-series feature matrix; constructing a geological-parameter coupling model; fusing the time-series feature matrix with geological exploration data to generate a feature fusion dataset; predicting attitude deviations based on the feature fusion dataset to generate attitude offset prediction values; using the attitude offset prediction values ​​as constraints to perform multi-objective optimization of pipe jacking construction to determine the optimal combination of tunneling parameters; simulating pipe jacking construction based on the optimal combination of tunneling parameters; optimizing and controlling the simulated tunneling parameters; and generating control commands to send to a controller for automatic parameter adjustment.

[0006] The second aspect of this application discloses an adaptive optimization system for tunneling parameters in pipe jacking construction. This system is used in the aforementioned adaptive optimization method for tunneling parameters in pipe jacking construction. The system includes: a multi-dimensional tunneling acquisition module for real-time multi-dimensional tunneling acquisition based on pipe jacking construction, obtaining a construction parameter dataset, standardizing it, and constructing a time-series feature matrix; a feature fusion module for constructing a geological-parameter coupling model, fusing the time-series feature matrix with geological exploration data to generate a feature fusion dataset; a multi-objective optimization module for predicting attitude deviations based on the feature fusion dataset, generating attitude offset prediction values, using the attitude offset prediction values ​​as constraints to perform multi-objective optimization of pipe jacking construction, and determining the optimal combination of tunneling parameters; and a control and optimization module for simulating pipe jacking construction based on the optimal combination of tunneling parameters, performing control and optimization based on the simulated tunneling parameters, generating control commands and sending them to the controller for automatic parameter adjustment.

[0007] One or more technical solutions provided in this application have at least the following beneficial effects: By collecting and standardizing construction parameters in real time, the dimensional differences between different data sources and sensors are eliminated. Constructing a time-series feature matrix enables unified processing of construction data from different time points, facilitating comprehensive analysis of dynamic changes during construction and ensuring high-quality data. This provides accurate input for subsequent optimization and prediction. The geological-parameter coupling model combines construction parameters and geological features, reflecting the impact of geological factors on construction. Through feature fusion, a multi-dimensional feature fusion dataset is generated, providing rich information for subsequent attitude deviation prediction, ensuring the accuracy and adaptability of construction optimization. Attitude deviation prediction allows for early prediction of the construction process. To address the attitude deviation problem in tunneling and ensure the accuracy and safety of the construction path, a multi-objective optimization algorithm is used to optimize the combination of tunneling parameters, achieving a comprehensive and integrated optimization effect. The final output of the optimal combination of tunneling parameters can significantly improve construction efficiency and reduce risks during construction. By simulating tunneling, the impact of different tunneling parameters on the construction process is previewed before actual construction, potential problems are identified and optimized in advance. The control and optimization not only make the construction process more flexible and adaptable to different geological conditions, but also ensure the efficiency and stability of the construction process by automatically adjusting the tunneling parameters in real time. Automatic parameter adjustment reduces human intervention, improves construction accuracy, and reduces deviations and errors in the construction process.

[0008] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application. Attached Figure Description

[0009] Figure 1 This is a schematic diagram of the adaptive optimization method for tunneling parameters in pipe jacking construction provided in an embodiment of this application.

[0010] Figure 2 This is a schematic diagram of the adaptive optimization system structure for tunneling parameters in pipe jacking construction provided in an embodiment of this application.

[0011] Figure labeling: 10 for multi-dimensional tunneling acquisition module, 20 for feature fusion module, 30 for multi-objective optimization module, and 40 for control and optimization module. Detailed Implementation

[0012] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0013] Example 1, as Figure 1 As shown in the embodiments of this application, an adaptive optimization method for tunneling parameters used in pipe jacking construction is provided, the method comprising: Based on pipe jacking construction, real-time multi-dimensional tunneling data is collected, and the construction parameter dataset is standardized to construct a time-series feature matrix.

[0014] During pipe jacking construction, sensor arrays are used to collect multi-dimensional construction parameter data in real time. For example, sensors are installed to obtain information such as the equipment's operating status, the pipe jacking machine's operating speed, thrust, vibration, pressure, and soil conditions. The collected construction parameter data have different dimensions or units. Standardization is used to eliminate these differences and ensure that the data can be compared and processed on a uniform scale. Standardization methods include Z-score standardization and min-max normalization. The standardized construction parameter data is organized chronologically to form a time-series feature matrix. This matrix treats data from each time step as a row and all collected parameters as feature columns, used to capture patterns of data change over time in subsequent analysis.

[0015] A geological-parameter coupling model is constructed, and the time-series feature matrix is ​​fused with geological exploration data to generate a feature fusion dataset.

[0016] The goal of the geological-parameter coupling model is to combine construction parameters collected during construction with geological survey data to more comprehensively analyze various changes during construction. This model is built upon convolutional neural networks (CNNs) and an attention mechanism. CNNs excel at processing time-series data and extracting local features, while the attention mechanism weights the importance of different features, further enhancing the model's capabilities. Geological survey data includes soil layer distribution, rock type, and groundwater conditions. From this data, geological feature vectors are extracted. The time-series feature matrix is ​​input into the first branch of the geological-parameter coupling model. This branch extracts local feature sets through multi-layer convolution operations, capturing the time-series characteristics of the construction parameter data. The geological feature vectors are then fused with the features extracted from the first branch through a second branch. The attention mechanism further weights different features, generating a feature fusion dataset, thus providing a multidimensional dataset containing both construction and geological information.

[0017] Based on the feature fusion dataset, attitude deviation is predicted to generate attitude offset prediction values. These attitude offset prediction values ​​are then used as constraints to perform multi-objective optimization for pipe jacking construction, thereby determining the optimal combination of tunneling parameters.

[0018] Attitude deviation refers to equipment displacement caused by unstable soil layers or other factors during pipe jacking construction. For example, the pipe jacking machine may deviate horizontally or vertically, affecting construction accuracy. Based on a feature fusion dataset, attitude deviation is predicted. Temporal convolution and long-range dependency capture are used to analyze the attitude change trend during construction and predict future attitude deviations. The predicted attitude deviation values ​​include horizontal and vertical offsets, indicating the difference between the current construction state and the expected target.

[0019] The predicted attitude deviation is input as a constraint into the multi-objective optimization algorithm to ensure that the attitude error does not exceed the set allowable range during construction. The optimization objectives include multiple aspects such as tunneling speed, surface settlement, and attitude stability. Through multi-objective optimization, a balance is found among these objectives to ensure that all indicators reach their optimal state during construction. The optimal combination of tunneling parameters is determined through multi-objective optimization to guide subsequent construction and ensure the efficiency and safety of the pipe jacking process.

[0020] The pipe jacking construction is simulated based on the optimal combination of tunneling parameters. The simulated tunneling parameters are then adjusted and optimized, and control commands are generated and sent to the controller to perform automatic parameter adjustments.

[0021] Digital twin simulation of pipe jacking construction is performed based on the optimal combination of tunneling parameters. Digital twin technology can create a virtual model of the construction process and simulate the construction by inputting the optimal combination of tunneling parameters. Feedback data obtained during the simulated tunneling process, such as simulated tunneling speed, settlement, and attitude changes, is compared with preset multi-dimensional target values. If the simulated parameters do not meet the expected targets, a control and optimization signal is triggered, automatically adjusting the tunneling parameters, such as thrust and speed, to achieve the expected goals. After control and optimization, control commands are generated, including the adjusted parameters. These commands are sent to the controller via a communication system. Upon receiving the control commands, the controller adjusts the relevant operating parameters of the pipe jacking machine to ensure the smooth progress of the construction process.

[0022] Furthermore, based on pipe jacking construction, real-time multi-dimensional tunneling data acquisition is performed to obtain a construction parameter dataset. Methods include: Based on the pipe jacking machine, the operation of pipe jacking construction is analyzed, and multiple key location data are located according to the construction operation parameters. A multi-source sensor array is deployed according to the multiple key location data, and the multi-source sensor array is started at the synchronous sampling frequency to collect real-time multi-dimensional tunneling data of the pipe jacking machine to obtain the original construction parameter dataset. The original construction parameter dataset is timestamped and aligned to obtain the synchronous construction parameter dataset, which is then cleaned to generate the construction parameter dataset.

[0023] A comprehensive analysis of the operation of the pipe jacking machine is conducted, including real-time monitoring and recording of various operating parameters such as thrust, speed, and pressure. The operation data of the pipe jacking machine is analyzed to locate multiple key locations. These key locations are areas where construction is difficult or where attitude deviations are likely to occur. For example, areas with complex geological conditions or slow construction progress require special attention. The data from these locations will serve as the basis for subsequent construction optimization.

[0024] After locating key positions, a multi-source sensor array is deployed at these positions. This array includes pressure sensors, accelerometers, temperature sensors, and displacement sensors to monitor the real-time operation data of the pipe jacking machine under different locations and working conditions. The specific locations and number of sensors are determined based on the construction environment and geological conditions to ensure comprehensive collection of relevant construction data. The multi-source sensor array follows a synchronous sampling frequency, meaning all sensors maintain a consistent sampling time to ensure data sequence. The synchronous sampling frequency is selected based on the actual construction situation, ensuring a sufficiently high frequency to capture subtle changes during construction. Through the multi-source sensor array, multi-dimensional tunneling data is collected in real-time. This data includes not only the pipe jacking machine's operating parameters, such as thrust, speed, and pressure, but also information on changes in the surrounding soil, ground settlement, and the vibration of the pipe jacking machine.

[0025] The original construction parameter dataset contains temporal asynchrony, meaning that the sampling time points of each sensor are different. To address this, timestamp alignment is performed on all data points in the original dataset to ensure that all sensor data are processed according to the same time step, resulting in a consistent temporal relationship between the data. Data cleaning is then performed on the timestamp-aligned data. This cleaning process involves removing outliers, filling in missing values, and removing noise to ensure data quality. For example, interpolation is used to fill in missing data points, or filtering methods are used to remove noise caused by environmental interference or sensor malfunctions. The cleaned data is then integrated into a construction parameter dataset. This dataset contains all the key parameters after time alignment and cleaning, and reflects the real-time status of the construction process.

[0026] Furthermore, the construction parameter dataset is obtained and standardized to construct a time-series feature matrix. Methods include: Modal decomposition is performed on the construction parameter dataset to generate multiple frequency intrinsic mode components (EMS). High-frequency noise components are identified according to these EMS, and the cleaned construction parameter data is denoised and reconstructed based on the high-frequency noise components to generate a noise-reduced construction parameter dataset. A tunneling cycle is set based on the pipe jacking construction, and the noise-reduced construction parameter dataset is divided into equal-length segments according to the tunneling cycle to generate multiple equal-length time-series data segments, where each equal-length time-series data segment corresponds to a tunneling cycle. The multiple equal-length time-series data segments are traversed and normalized to generate normalized time-series data. Statistical features are extracted from the normalized time-series data, and the statistical features are arranged in chronological order to construct the time-series feature matrix.

[0027] During construction, the collected parameter data contains signal components of various frequencies. To extract useful information from this data, mode decomposition (MDD) is performed. MDD is a signal processing method that aims to decompose a complex signal into multiple simple intrinsic mode components (EMCs), each corresponding to a different frequency component. MDD methods include empirical mode decomposition (EMD) or wavelet transform. Through MDD, the construction parameter dataset is decomposed into multiple frequency EMCs, which represent signals with different frequency components. These EMCs can reveal different physical phenomena during construction, such as the vibration frequency of the pipe jacking machine and the periodic changes in equipment operation.

[0028] High-frequency noise refers to irrelevant signals in the data generated by environmental interference, equipment failure, sensor errors, and other factors. These signals reside in the high-frequency range and are unrelated to the actual dynamics of construction. By analyzing the intrinsic mode components (EMS), we can identify which components are low-frequency signals relevant to construction and which are noise signals, with the high-frequency components representing noise. After identifying the high-frequency noise components, these noises are removed. During the denoising process, filtering methods such as bandpass filters or wavelet denoising techniques are used to remove high-frequency noise, retaining only the useful low-frequency components. In this way, the construction parameter data is denoised and reconstructed, removing the interference of high-frequency noise and obtaining a more accurate denoised construction parameter dataset.

[0029] In pipe jacking construction, the tunneling cycle refers to the time interval from when the pipe jacking machine starts tunneling from a specific position to the next critical position. This cycle is a fixed length and can be adjusted according to the actual construction situation. First, a tunneling cycle is set, which means dividing the construction process into segments, each segment corresponding to a complete construction cycle. Based on the set tunneling cycle, the noise reduction construction parameter dataset is divided into equal-length time-series data segments. Each segment contains data from a complete construction cycle, and each equal-length time-series data segment directly corresponds to the tunneling cycle. This ensures that each segment reflects the dynamic process of a complete construction cycle. This correspondence is used for subsequent analysis of the periodic characteristics of the construction process.

[0030] Normalization transforms data of different ranges and dimensions to a uniform scale. For example, it standardizes data to the [0, 1] interval or converts it to a standard normal distribution with zero mean and unit variance. Normalizing each equal-length time-series data segment ensures data comparability and stability, especially when different parameters (such as pressure, velocity, vibration, etc.) have different dimensions; normalization effectively eliminates the influence between dimensions. Time-series data that has undergone normalization can be used for subsequent analysis and modeling, avoiding computational biases caused by data scaling issues.

[0031] Statistical features refer to representative data characteristics extracted from normalized time-series data, used to describe the overall trend or pattern of change in the data. Statistical features include mean, variance, standard deviation, maximum value, minimum value, skewness, kurtosis, and autocorrelation coefficient. After extracting statistical features, these features are arranged in chronological order, that is, the statistical features of each time-series segment are combined sequentially according to the time of collection. This time-series arrangement helps maintain the temporal relationship of the data, facilitating subsequent model learning and prediction. Each row of the constructed time-series feature matrix represents the statistical feature of a time-series segment, and each column represents a feature, such as mean or variance, providing overall dynamic information about the construction data.

[0032] Furthermore, a geological-parameter coupling model is constructed, and the time-series feature matrix is ​​fused with geological exploration data to generate a feature fusion dataset. The method includes: The geological-parameter coupling model is constructed based on convolutional neural networks and an attention mechanism. The geological-parameter coupling model includes a first branch and a second branch. Geological exploration data is obtained based on the working face location information, and geological feature vectors are extracted from the geological exploration data. The temporal feature matrix is ​​synchronized to the first branch of the geological-parameter coupling model for multi-layer convolution to extract local feature sets. Dynamic change analysis is performed on the local feature sets to construct a temporal depth feature map. The temporal depth feature map is expanded along the time dimension to obtain a feature vector sequence. The geological feature vectors are synchronized to the second branch for feature space mapping to obtain geological feature embedding vectors. Based on the attention mechanism, the temporal depth feature map and the geological feature embedding vectors are fused to construct the feature fusion dataset.

[0033] The geological-parameter coupling model combines construction parameters with geological features for analysis. By coupling these two types of information, it can more comprehensively predict various changes during construction. Convolutional neural networks (CNNs) are used to process time-series data, extracting local features such as behavioral patterns at specific time points during construction. The first branch extracts local features from the time-series data through multiple convolutional layers, while the second branch performs feature space mapping. An attention mechanism is used to assign different weights to different time steps or features, focusing more on features that significantly impact the construction process. This makes the model more accurate in the coupled analysis of geological and construction parameters, especially when facing complex geological conditions. It helps improve the model's learning and attention to geological information, ensuring the model can effectively integrate construction and geological data.

[0034] The tunnel face refers to the foremost working face or excavation point in pipe jacking construction. The location of the tunnel face changes continuously as construction progresses. By using the tunnel face location information, the specific location of the current construction site can be determined, thereby obtaining the corresponding geological survey data. Based on the geological survey data, geological feature vectors are extracted. These vectors represent the geological conditions at the current tunnel face location, such as soil density, elastic modulus, and porosity. This geological data is obtained in real time through geological survey sensors, providing a basis for construction decisions.

[0035] The constructed temporal feature matrix is ​​input into the first branch of the geoparameter coupling model. This first branch uses a multi-layer convolutional neural network to process the temporal feature matrix. Through multiple convolutional layers, the network automatically learns local features in the temporal data. Each convolutional layer captures patterns in the construction data through its local receptive field and transforms them into higher-level features. These local features represent characteristics at different time steps during construction, such as construction vibrations and pressure changes at a certain stage. After the convolution operation, a local feature set is extracted, which describes the data characteristics at different time points during the construction process.

[0036] Local feature sets reflect local changes during construction, such as thrust variations, pressure fluctuations, and equipment vibrations. To further capture the dynamic characteristics of the construction process, dynamic change analysis is performed on these local features. This analysis focuses on features that change over time, revealing trends and patterns in the construction process, such as analyzing periodic fluctuations and sudden abnormal changes. A time-series depth feature map is generated, which combines time-series data with depth features to comprehensively represent both time-series features (such as changes in construction status at different times) and depth features (such as equipment performance and soil conditions) during construction.

[0037] The time-series deep feature map is expanded according to the time dimension. The expansion process is actually expanding the data in the time-series feature map in the order of time sequence, so that the feature at each time step can be expressed in the form of vector. The expanded time-series deep feature map is transformed into a series of feature vectors. Each feature vector corresponds to the data of a time step. The feature vector sequence is a set of vectors arranged in time order, and each vector is an important state of the construction process.

[0038] In the second branch of the geological-parameter coupling model, the extracted geological feature vectors are input synchronously. This second branch maps these vectors to a new feature space through feature space mapping. This process maps the geological data from the original feature space to a feature space more suitable for integration with construction parameters. The purpose of feature space mapping is to transform the geological information into a format more easily processed by the model, enabling comparison and integration of geological information and construction parameters at the same level. After feature space mapping, the geological feature vectors are transformed into geological feature embedding vectors. These vectors represent the geological data in the new feature space, allowing for fusion with construction parameter data within the model.

[0039] Attention mechanisms are used to focus on the most important information in the input features. The core idea is to assign different weights to each input feature based on its importance, thereby enhancing the learning ability for key features. In this step, the attention mechanism is applied to both the temporal deep feature map and the geological feature embedding vector. Based on the impact of construction data and geological data on the construction process, the weights of these two types of information in feature fusion are dynamically adjusted. Through the attention mechanism, the temporal deep feature map and the geological feature embedding vector are weighted and fused into a unified feature representation. The resulting feature fusion dataset contains information combining changes in construction parameters and geological conditions during the construction process.

[0040] Furthermore, based on the attention mechanism, feature fusion is performed on the temporal deep feature map and the geological feature embedding vector to construct the feature fusion dataset. The method includes: Based on the attention mechanism, the geological feature embedding vector is used as the query vector, and the feature vector sequence is used as the key and value to perform a dot product operation to obtain multiple attention scores. The multiple attention scores are normalized to generate multiple response weights. The feature vector sequence is weighted and summed according to the multiple response weights to obtain a geological weighted time-series feature vector. The geological weighted time-series feature vector is concatenated and fused with the geological feature embedding vector to generate a fused feature vector. Multilayer perceptual transformation is performed based on the fused feature vector to generate a deep geological-parameter coupling relationship. Tunneling cycle analysis is performed according to the deep geological-parameter coupling relationship to generate geological-parameter joint features. The geological-parameter joint features are added to the feature fusion dataset.

[0041] The attention mechanism assigns different weights based on the relative importance of different parts of the input data. For time series data, the attention mechanism can focus on time steps that have a significant impact on prediction or the task. Geological feature embedding vectors are used as query vectors, representing the current state of geological data, such as soil type and stratigraphic structure. The query vectors query features at different time steps to find the moment most relevant to the current geological conditions. The feature vector sequence comes from the unfolding of the time series data, representing the parametric features of each time step in the construction process. These feature vector sequences serve as key and value inputs, representing the construction state at each time point. For each query vector and its corresponding key and value, a dot product is performed. The result of the dot product represents the correlation between the geological conditions and the features at each time step. The resulting attention score reflects the degree of relationship between the geological feature embedding vector and the feature vector at each time step; a higher score indicates a greater influence of the current geological conditions on the construction process at that moment.

[0042] The obtained attention scores are normalized to ensure that the sum of all scores is 1. For example, Softmax is used, which converts the attention scores into weight values ​​through exponential operations and normalization. The specific formula is as follows: ,in, For attention scores, j represents all time steps, and the final value serves as the response weight for each time step. The normalized response weights reflect the importance of each time step under the current geological conditions. Time steps with larger weights indicate a greater impact on the construction process under the geological conditions, while time steps with smaller weights indicate a weaker geological impact.

[0043] For each time step feature vector, a weighted sum is performed using response weights. Specifically, for each time step feature vector, the corresponding response weight is multiplied, and then all weighted feature vectors are summed to obtain a new vector, namely the geological weighted time series feature vector. This vector emphasizes the time step features that are closely related to geological conditions. This ensures that when the model performs feature fusion, it pays more attention to those moments that are highly correlated with the current geological conditions, thereby improving the model's prediction accuracy.

[0044] The generated geological weighted time series feature vector is concatenated with the geological feature embedding vector. The concatenation operation means that the two vectors are directly merged into a new vector, namely the fused feature vector. The fused feature vector contains the fused features of both construction data and geological data, providing a comprehensive feature representation.

[0045] A multilayer perceptron (MLP) is a fully connected neural network used for nonlinear transformations. An MLP contains multiple fully connected layers, each undergoing a nonlinear transformation through an activation function such as ReLU or Sigmoid to extract deeper features. MLP transformations can learn more complex patterns from simple input features. Through MLP transformations, the deep coupling relationship between construction parameters and geological features can be extracted. This coupling relationship reveals the complex interactions between construction parameters and geological conditions, such as how they jointly affect construction progress and outcomes over time.

[0046] Based on the deep coupling relationship between geology and parameters, tunneling cycle analysis is performed. This analysis evaluates each stage of the construction process, such as the start of tunneling, the advancement stage, and soil response, to determine the optimal construction strategy and parameter adjustments. The results of the tunneling cycle analysis generate a joint geology-parameter feature, which indicates how construction parameters can be optimized under given geological conditions to achieve the best construction results.

[0047] The obtained geological-parameter joint features are added to the feature fusion dataset. This addition process ensures that the model can make comprehensive use of the coupling relationship between geological and construction data when training and predicting, which helps to improve the prediction accuracy and generalization ability of the subsequent model.

[0048] Furthermore, based on the aforementioned feature fusion dataset, attitude deviation prediction is performed to generate attitude offset prediction values. These attitude offset prediction values ​​are then used as constraints for multi-objective optimization of the pipe jacking construction to determine the optimal combination of tunneling parameters. The method includes: The feature fusion dataset is used as the input sequence for temporal convolution to extract local temporal patterns; long-range dependency capture is performed on the input sequence to obtain the feature deep hidden state; mapping prediction is performed based on the feature deep hidden state to obtain horizontal offset prediction values ​​and vertical offset prediction values; the horizontal offset prediction values ​​and the vertical offset prediction values ​​are integrated with attitude deviation to generate attitude offset prediction values; an allowable range for axis deviation in pipe jacking construction is set, and the attitude offset prediction values ​​are compared with the allowable range; when the attitude offset prediction value is greater than or equal to the allowable range, an attitude over-limit warning signal is generated; based on the attitude over-limit warning signal, the attitude offset prediction value is used as a constraint to perform multi-objective optimization of pipe jacking construction to determine the optimal combination of tunneling parameters.

[0049] The feature fusion dataset is input as a sequence into a convolutional neural network for temporal convolution. Temporal convolution extracts local patterns from time series data to analyze behavioral characteristics at different time steps during construction. Through temporal convolution, local dependencies and periodic features in the data can be identified, such as the regular changes in certain construction parameters over a period of time, or the direct impact of geological conditions on construction. Multiple convolutional layers automatically learn and extract local temporal patterns from the data. These patterns represent the changing trends of specific behaviors or states during construction, such as fluctuations in thrust and velocity.

[0050] In time-series data, the impact of certain key events or states can extend over long periods, making them difficult to capture directly through local convolutional operations. To capture these long-range dependencies, recurrent neural networks or long short-term memory networks can be used. Capturing long-range dependencies helps in understanding the relationships between long-term time steps in the data, especially the long-term impact of changes in geological conditions on construction progress. After capturing long-range dependencies, a deep hidden state is generated, which contains long-term memories of the construction process and reflects the dynamic changes throughout the entire construction cycle.

[0051] Mapping prediction is performed based on the hidden state of features, that is, predicting the future construction state based on the hidden state. Mapping prediction is performed through fully connected layers or regression models. The goal is to predict the offsets that occur during construction, especially the horizontal and vertical offsets. The predicted values ​​of horizontal and vertical offsets reflect the changes in equipment position during construction. If the horizontal and vertical offsets are too large, it will lead to a decrease in construction accuracy, or even path deviation or equipment damage.

[0052] Attitude deviation integration combines the predicted values ​​of horizontal and vertical offsets into a comprehensive attitude deviation prediction value. This can be achieved by weighted summation or by using Euclidean distance to combine the offsets in these two directions. The integrated attitude deviation prediction value can more comprehensively reflect the deviation of the equipment during construction, facilitating subsequent adjustments and optimizations.

[0053] The allowable range of axial deviation in pipe jacking construction refers to the permissible error range of the equipment deviating from the predetermined trajectory during construction. This range is set by the construction plan and project requirements and is adjusted according to factors such as equipment accuracy, soil conditions, terrain, and construction progress. If the equipment deviates beyond this range, it will cause the trajectory to deviate too much, thereby affecting the subsequent construction progress and quality.

[0054] The predicted attitude deviation value is compared with the preset allowable range for axis deviation. If the predicted attitude deviation value is greater than or equal to this range, it indicates a significant deviation during construction, requiring adjustment measures. At this point, an attitude over-limit warning signal is automatically generated to alert construction personnel to any deviations occurring during construction.

[0055] In pipe jacking construction, multi-objective optimization refers to balancing and optimizing multiple construction objectives. These objectives include tunneling speed, attitude stability, and soil response. Multi-objective optimization can improve construction efficiency while ensuring accuracy. Predicted attitude deviations are used as constraints to ensure that the equipment's deviation does not exceed a preset allowable range during optimization. Based on attitude over-limit warning signals, multi-objective optimization algorithms, such as genetic algorithms and particle swarm optimization, are used to calculate and adjust the optimal combination of tunneling parameters, including thrust, rotational speed, propulsion speed, and mud flow rate. Optimizing these parameter combinations improves construction accuracy and efficiency, ensuring that construction proceeds within a controllable deviation range.

[0056] Furthermore, based on the attitude over-limit early warning signal, the predicted attitude deviation value is used as a constraint condition to perform multi-objective optimization of the pipe jacking construction to determine the optimal combination of tunneling parameters. The method includes: Multiple optimization objectives are constructed, and the predicted attitude deviation values ​​are converted into constraints for these objectives. Based on the attitude over-limit warning signal and the constraints, multiple allowable deviation thresholds are set, including horizontal and vertical allowable deviation thresholds. The predicted horizontal deviation values ​​are compared with the horizontal allowable deviation thresholds to determine the target horizontal deviation range. The predicted vertical deviation values ​​are also compared with the vertical allowable deviation thresholds to determine the target vertical deviation range. Multi-objective optimization is performed according to the target horizontal and vertical deviation ranges to construct candidate tunneling parameter combinations. The candidate tunneling parameter combinations are traversed to perform pipe jacking machine-soil interaction analysis, generating tunneling parameters. The process involves considering the tunneling speed response value and the surface settlement response value. Based on these values, a first optimization target value and a second optimization target value are set. Candidate tunneling parameter combinations are then non-dominated and sorted according to these values, dividing the process into multiple non-dominated levels. A screening analysis is performed on the candidate tunneling parameter combinations to determine parent individuals for crossover mutation, generating offspring populations. The parent individuals and offspring populations are merged and iterated according to multiple non-dominated levels to generate optimal solution sets for multiple non-dominated solutions. Membership degrees are calculated based on these optimal solution sets to generate multiple comprehensive satisfaction levels. These optimal solution sets are then screened to determine the optimal tunneling parameter combination.

[0057] In pipe jacking construction, multiple optimization objectives include attitude stability, tunneling speed, soil stability, and surface settlement control. Each objective has different importance and constraints during the optimization process. The optimization algorithm needs to balance these objectives to achieve the best construction results. Using the predicted attitude offset as a constraint on multiple optimization objectives means that during construction, the equipment offset must be controlled within a predetermined range to ensure the accuracy of the construction path.

[0058] Based on the attitude over-limit warning signal, specific allowable offset thresholds are set for horizontal and vertical offsets. These thresholds define the maximum allowable offset during construction. Exceeding this range is considered an over-limit. The horizontal allowable offset threshold refers to the maximum allowable error range in the horizontal direction; the vertical allowable offset threshold refers to the maximum allowable error range in the vertical direction.

[0059] The predicted horizontal offset is compared with the permissible horizontal offset threshold to determine whether the current horizontal offset exceeds the allowable range. If the horizontal offset exceeds the threshold, adjustment is required; otherwise, construction can continue. Similarly, the predicted vertical offset is compared with the permissible vertical offset threshold. Based on the comparison results, the target horizontal and vertical offset intervals are determined. These intervals define the permissible horizontal and vertical offset ranges during construction, providing constraints for multi-objective optimization.

[0060] A multi-objective optimization algorithm is used to optimize the construction process by finding a balance point among multiple objectives. The objectives of multi-objective optimization include not only offset control but also other factors such as advance speed, soil response, and surface settlement. During the multi-objective optimization process, candidate combinations of tunneling parameters are generated. These combinations represent different construction strategies, each of which affects the behavior, speed, thrust, and other parameters of the equipment during construction.

[0061] In candidate tunneling parameter combinations, an interaction analysis of the pipe jacking machine and the soil is performed. This step simulates the interaction between the pipe jacking machine and the soil under different tunneling parameters, assessing factors such as soil response and settlement during construction. The interaction analysis reveals the impact of different tunneling strategies on the soil, particularly predicting surface settlement and the influence on surrounding buildings. Based on the interaction analysis, tunneling speed response values ​​and surface settlement response values ​​are generated. These response values ​​reflect the impact of different tunneling parameter combinations on construction progress and the surrounding environment. The tunneling speed response value indicates the speed of construction progress, directly affecting construction efficiency, while the surface settlement response value indicates the surface settlement during construction; excessive settlement can lead to structural damage or safety hazards.

[0062] The first optimization target value is set based on the tunneling speed response value. This is the optimization target for tunneling speed, which aims to maximize construction efficiency and increase tunneling speed. The second optimization target value is set based on the surface settlement response value. This is the optimization target for surface settlement, which aims to minimize surface settlement and ensure that the construction process does not cause excessive surface settlement, thus avoiding the impact on the surrounding environment and buildings.

[0063] Non-dominated ranking is a method in multi-objective optimization used to compare the merits of different solutions. In this process, a solution is considered to "dominate" the second solution if it is no worse than another solution across all optimization objectives and is better than the second solution at least in one objective. By non-dominated ranking candidate tunneling parameter combinations, the merits of each solution are evaluated, and multiple non-dominated levels are established according to different priorities. These levels represent the degree of merit of the solutions: the first level is the optimal solution set, the second level is the suboptimal solution set, and so on. This ranking helps to prioritize the optimal solution and progressively improve suboptimal solutions during subsequent selection and optimization processes.

[0064] After non-dominated sorting, candidate tunneling parameter combinations are screened to determine which solutions can serve as parent individuals for the next genetic operation, namely crossover and mutation. The screening process is based on the non-dominated level of the solutions and the objective function value. During the screening process, solutions at the first level are preferentially selected as parent individuals, as these solutions represent the current optimal construction strategy.

[0065] Crossover is an operation that generates new offspring individuals based on selected parent individuals. The crossover process involves exchanging some features of the parent individuals to generate new solutions, thus exploring new regions in the solution space. Mutation, on the other hand, makes small, random adjustments to the solutions after crossover, helping to avoid getting trapped in local optima and increasing solution diversity. Through crossover and mutation, new candidate solutions are generated, forming an offspring population, which provides more possible solutions for subsequent optimization.

[0066] The parent and offspring populations are merged into a larger population. The merged population is then iteratively calculated using a non-dominated sorting algorithm to further evaluate the quality of each solution. The population is updated accordingly, and the solution set is continuously optimized through iteration, gradually generating better solutions. During the iteration process, the optimal solution is continuously selected while suboptimal solutions are discarded until convergence to the optimal solution set. Ultimately, an optimal solution set containing multiple non-dominated solutions is generated. These solutions represent different construction strategies or parameter combinations, each with its own advantages. The most suitable solution can be selected based on actual needs.

[0067] Membership degree calculation is used to evaluate the performance of each solution in multi-objective optimization. By calculating the performance of each solution relative to the objectives, membership degree measures the goodness or badness of a solution under given constraints. Membership degree is calculated based on multiple factors such as objective function value, deviation range, and construction stability; a higher value indicates a better solution. Through membership degree calculation, a comprehensive satisfaction level is generated for each solution. Solutions with high comprehensive satisfaction level represent solutions that perform well across various optimization objectives and are more suitable for practical applications. Finally, the optimal solution set is filtered based on the comprehensive satisfaction level to select the optimal combination of tunneling parameters. This optimal solution combines multiple objectives such as tunneling speed and surface settlement, providing optimal operating parameters for the construction process.

[0068] Furthermore, the method involves simulating tunneling based on the optimal combination of tunneling parameters, optimizing and controlling the simulated tunneling parameters, and generating control commands. The method includes: A digital twin is used to simulate the tunneling cycle based on the optimal combination of tunneling parameters, generating simulated tunneling parameters, including simulated tunneling speed parameters, simulated surface settlement parameters, and simulated attitude change trajectory parameters. Based on pipe jacking construction, multi-dimensional expected target values ​​are set, and the simulated tunneling speed parameters, simulated surface settlement parameters, and simulated attitude change trajectory parameters are compared with these multi-dimensional expected target values. When any one of these parameters fails to meet the multi-dimensional expected target value, a control optimization signal is triggered. Control conversion is performed through the control optimization signal to generate an optimized combination of tunneling parameters, thus constructing the control command.

[0069] Digital twin is a virtual modeling technology that provides a virtual simulation environment for construction processes by digitally simulating real-world systems or processes. This helps construction teams predict and optimize various factors in actual operations. Based on the optimized combination of tunneling parameters, digital twin technology is used to virtually simulate the entire tunneling process. The simulation includes various stages of the tunneling cycle, such as the start of tunneling, the advancement stage, and soil response. Through virtual simulation, the impact of different parameter settings on the construction process can be more clearly understood, providing guidance for actual construction. During the digital twin simulation, a series of simulated tunneling parameters are generated. Among these, the simulated tunneling speed parameter represents the tunneling speed of the pipe jacking machine during the simulation, affecting construction efficiency; the simulated surface settlement parameter reflects the impact of the construction process on surface settlement; excessive settlement can damage the ground structure; and the simulated attitude change trajectory parameter represents the changes in equipment attitude during construction; attitude deviations can lead to a decrease in construction accuracy.

[0070] During pipe jacking construction, multi-dimensional expected target values ​​are set. These target values ​​represent the standards and requirements to be achieved during construction, including target values ​​for tunneling speed, surface settlement, and attitude change trajectory. The simulated tunneling parameters, including tunneling speed, surface settlement, and attitude change trajectory, are compared one by one with the set multi-dimensional expected target values. If the simulated parameters meet the expected target values, it indicates that the construction process is within a controllable range; if there are deviations, adjustments are needed.

[0071] If any parameter in the simulation does not meet the set multidimensional expected target value, a control and optimization signal is triggered. For example, if the simulated tunneling speed is too fast or too slow and cannot meet the construction progress requirements, a signal is triggered to adjust the tunneling speed; if the simulated surface settlement is too large and exceeds the safety range, a signal is triggered to adjust the construction parameters to reduce settlement; if the simulated attitude change trajectory deviates too much and affects the construction accuracy, a signal is triggered to adjust the attitude control parameters. The triggering of the control and optimization signal ensures that the construction process always remains within the set safety range, avoiding deviations or instability during construction.

[0072] Control conversion transforms the regulation and optimization signals into specific control commands. Through control conversion, the regulation and optimization signals are translated into actual parameters that can operate the controller, generating optimized tunneling parameter combinations. These optimized parameter combinations include adjusted parameters to ensure that various indicators during construction meet predetermined targets. Based on the generated optimized tunneling parameter combinations, control instructions are constructed. These control instructions are specific operational commands, including a series of parameters such as tunneling speed, thrust, and soil pressure, used to guide the pipe jacking construction equipment to automatically adjust in real-time.

[0073] Furthermore, the method for generating control commands and sending them to the controller to execute automatic parameter adjustments includes: The combination of tunneling optimization parameters is converted into a format according to the controller's data interface to generate a parameter setting data packet; the parameter setting data packet is sent to the controller for parsing to generate parameter control setting values; the controller is then activated to automatically adjust the tunneling parameters for pipe jacking construction according to the parameter control setting values.

[0074] The controller's data interface is used to receive external commands and convert them into a format that can be executed by the controller. The combination of tunneling optimization parameters is represented in a predetermined standard format, such as JSON, XML, or a specific engineering protocol format. In order to ensure that the controller can correctly understand and execute these parameters, these parameters are converted into a format supported by the controller. After the format conversion, a parameter setting data packet is generated, which contains all the parameter information that needs to be sent to the controller.

[0075] Upon receiving the parameter setting data packet, the controller first parses it. This parsing process involves interpreting the parameters within the packet and converting them into specific commands that the controller can execute. During parsing, the controller updates its internal parameter settings based on the instructions in the parameter setting data packet, ensuring that it can operate correctly according to external input. The parsed data packet generates a series of parameter control setpoints. These setpoints represent the specific operating parameters required during pipe jacking construction, such as tunneling speed, thrust intensity, and attitude adjustment values. These control setpoints guide the controller in actual equipment control and parameter adjustments, ensuring that the construction process proceeds according to the optimized objectives.

[0076] The controller is activated based on the parameter control settings, automatically adjusting various operating parameters of the equipment, such as tunneling speed, thrust, and attitude, to achieve optimal construction results. Automatic parameter adjustment means that the controller automatically adjusts construction parameters based on preset targets and real-time feedback without manual intervention. For example, if the tunneling speed is too fast or the surface subsidence is too large, the controller will automatically reduce the tunneling speed or adjust the thrust to ensure construction progress and safety. After automatic adjustment, the pipe jacking equipment will continue to perform the tunneling task according to the new parameters. The controller monitors the construction process in real time and continues to adjust relevant parameters according to the site conditions to ensure smooth construction.

[0077] Example 2, based on the same inventive concept as the adaptive optimization method for tunneling parameters used in the previous examples for pipe jacking construction, such as... Figure 2 As shown in the embodiment of this application, an adaptive optimization system for tunneling parameters used in pipe jacking construction is provided, the system comprising: The multi-dimensional tunneling acquisition module 10 is used to perform real-time multi-dimensional tunneling acquisition based on pipe jacking construction, obtain construction parameter datasets, perform standardized processing, and construct a time-series feature matrix; the feature fusion module 20 is used to construct a geological-parameter coupling model, fuse the time-series feature matrix with geological exploration data, and generate a feature fusion dataset; the multi-objective optimization module 30 is used to predict attitude deviation based on the feature fusion dataset, generate attitude offset prediction values, use the attitude offset prediction values ​​as constraints to perform multi-objective optimization of pipe jacking construction, and determine the optimal combination of tunneling parameters; the control and optimization module 40 is used to simulate tunneling of pipe jacking construction according to the optimal combination of tunneling parameters, control and optimize based on the simulated tunneling parameters, generate control commands and send them to the controller to execute automatic parameter adjustment.

[0078] Furthermore, the multi-dimensional tunneling acquisition module 10 is used to perform the following operation steps: Based on the pipe jacking machine, the operation of pipe jacking construction is analyzed, and multiple key location data are located according to the construction operation parameters. A multi-source sensor array is deployed according to the multiple key location data, and the multi-source sensor array is started at the synchronous sampling frequency to collect real-time multi-dimensional tunneling data of the pipe jacking machine to obtain the original construction parameter dataset. The original construction parameter dataset is timestamped and aligned to obtain the synchronous construction parameter dataset, which is then cleaned to generate the construction parameter dataset.

[0079] Furthermore, the multi-dimensional tunneling acquisition module 10 is used to perform the following operation steps: Modal decomposition is performed on the construction parameter dataset to generate multiple frequency intrinsic mode components (EMS). High-frequency noise components are identified according to these EMS, and the cleaned construction parameter data is denoised and reconstructed based on the high-frequency noise components to generate a noise-reduced construction parameter dataset. A tunneling cycle is set based on the pipe jacking construction, and the noise-reduced construction parameter dataset is divided into equal-length segments according to the tunneling cycle to generate multiple equal-length time-series data segments, where each equal-length time-series data segment corresponds to a tunneling cycle. The multiple equal-length time-series data segments are traversed and normalized to generate normalized time-series data. Statistical features are extracted from the normalized time-series data, and the statistical features are arranged in chronological order to construct the time-series feature matrix.

[0080] Furthermore, the feature fusion module 20 is used to perform the following operation steps: The geological-parameter coupling model is constructed based on convolutional neural networks and an attention mechanism. The geological-parameter coupling model includes a first branch and a second branch. Geological exploration data is obtained based on the working face location information, and geological feature vectors are extracted from the geological exploration data. The temporal feature matrix is ​​synchronized to the first branch of the geological-parameter coupling model for multi-layer convolution to extract local feature sets. Dynamic change analysis is performed on the local feature sets to construct a temporal depth feature map. The temporal depth feature map is expanded along the time dimension to obtain a feature vector sequence. The geological feature vectors are synchronized to the second branch for feature space mapping to obtain geological feature embedding vectors. Based on the attention mechanism, the temporal depth feature map and the geological feature embedding vectors are fused to construct the feature fusion dataset.

[0081] Furthermore, the feature fusion module 20 is used to perform the following operation steps: Based on the attention mechanism, the geological feature embedding vector is used as the query vector, and the feature vector sequence is used as the key and value to perform a dot product operation to obtain multiple attention scores. The multiple attention scores are normalized to generate multiple response weights. The feature vector sequence is weighted and summed according to the multiple response weights to obtain a geological weighted time-series feature vector. The geological weighted time-series feature vector is concatenated and fused with the geological feature embedding vector to generate a fused feature vector. Multilayer perceptual transformation is performed based on the fused feature vector to generate a deep geological-parameter coupling relationship. Tunneling cycle analysis is performed according to the deep geological-parameter coupling relationship to generate geological-parameter joint features. The geological-parameter joint features are added to the feature fusion dataset.

[0082] Furthermore, the multi-objective optimization module 30 is used to perform the following operation steps: The feature fusion dataset is used as the input sequence for temporal convolution to extract local temporal patterns; long-range dependency capture is performed on the input sequence to obtain the feature deep hidden state; mapping prediction is performed based on the feature deep hidden state to obtain horizontal offset prediction values ​​and vertical offset prediction values; the horizontal offset prediction values ​​and the vertical offset prediction values ​​are integrated with attitude deviation to generate attitude offset prediction values; an allowable range for axis deviation in pipe jacking construction is set, and the attitude offset prediction values ​​are compared with the allowable range; when the attitude offset prediction value is greater than or equal to the allowable range, an attitude over-limit warning signal is generated; based on the attitude over-limit warning signal, the attitude offset prediction value is used as a constraint to perform multi-objective optimization of pipe jacking construction to determine the optimal combination of tunneling parameters.

[0083] Furthermore, the multi-objective optimization module 30 is used to perform the following operation steps: Multiple optimization objectives are constructed, and the predicted attitude deviation values ​​are converted into constraints for these objectives. Based on the attitude over-limit warning signal and the constraints, multiple allowable deviation thresholds are set, including horizontal and vertical allowable deviation thresholds. The predicted horizontal deviation values ​​are compared with the horizontal allowable deviation thresholds to determine the target horizontal deviation range. The predicted vertical deviation values ​​are also compared with the vertical allowable deviation thresholds to determine the target vertical deviation range. Multi-objective optimization is performed according to the target horizontal and vertical deviation ranges to construct candidate tunneling parameter combinations. The candidate tunneling parameter combinations are traversed to perform pipe jacking machine-soil interaction analysis, generating tunneling parameters. The process involves considering the tunneling speed response value and the surface settlement response value. Based on these values, a first optimization target value and a second optimization target value are set. Candidate tunneling parameter combinations are then non-dominated and sorted according to these values, dividing the process into multiple non-dominated levels. A screening analysis is performed on the candidate tunneling parameter combinations to determine parent individuals for crossover mutation, generating offspring populations. The parent individuals and offspring populations are merged and iterated according to multiple non-dominated levels to generate optimal solution sets for multiple non-dominated solutions. Membership degrees are calculated based on these optimal solution sets to generate multiple comprehensive satisfaction levels. These optimal solution sets are then screened to determine the optimal tunneling parameter combination.

[0084] Furthermore, the regulation and optimization module 40 is used to perform the following operation steps: A digital twin is used to simulate the tunneling cycle based on the optimal combination of tunneling parameters, generating simulated tunneling parameters, including simulated tunneling speed parameters, simulated surface settlement parameters, and simulated attitude change trajectory parameters. Based on pipe jacking construction, multi-dimensional expected target values ​​are set, and the simulated tunneling speed parameters, simulated surface settlement parameters, and simulated attitude change trajectory parameters are compared with these multi-dimensional expected target values. When any one of these parameters fails to meet the multi-dimensional expected target value, a control optimization signal is triggered. Control conversion is performed through the control optimization signal to generate an optimized combination of tunneling parameters, thus constructing the control command.

[0085] Furthermore, the regulation and optimization module 40 is used to perform the following operation steps: The combination of tunneling optimization parameters is converted into a format according to the controller's data interface to generate a parameter setting data packet; the parameter setting data packet is sent to the controller for parsing to generate parameter control setting values; the controller is then activated to automatically adjust the tunneling parameters for pipe jacking construction according to the parameter control setting values.

[0086] Through the foregoing detailed description of the adaptive optimization method for tunneling parameters used in pipe jacking construction, those skilled in the art can clearly understand the adaptive optimization system for tunneling parameters used in pipe jacking construction in this embodiment. Since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and relevant parts can be referred to in the method section.

[0087] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. An adaptive optimization method for tunneling parameters used in pipe jacking construction, characterized in that, The method includes: Real-time multi-dimensional tunneling data acquisition based on pipe jacking construction is performed to obtain construction parameter datasets, which are then standardized to construct a time-series feature matrix. A geological-parameter coupling model is constructed, and the time-series feature matrix is ​​fused with geological exploration data to generate a feature fusion dataset. Based on the feature fusion dataset, attitude deviation is predicted to generate attitude offset prediction values. The attitude offset prediction values ​​are used as constraints to perform multi-objective optimization of pipe jacking construction to determine the optimal combination of tunneling parameters. The pipe jacking construction is simulated based on the optimal combination of tunneling parameters. The simulated tunneling parameters are then adjusted and optimized, and control commands are generated and sent to the controller to perform automatic parameter adjustments.

2. The adaptive optimization method for tunneling parameters in pipe jacking construction as described in claim 1, characterized in that, Real-time multi-dimensional tunneling data acquisition based on pipe jacking construction to obtain construction parameter datasets includes the following methods: Based on the pipe jacking machine, we conducted an operational analysis of pipe jacking construction and located multiple key location data based on the operational parameters. A multi-source sensor array is deployed according to the data of the multiple key locations, and the multi-source sensor array is started according to the synchronous sampling frequency to collect real-time multi-dimensional tunneling data of the pipe jacking machine, so as to obtain the original construction parameter dataset. The original construction parameter dataset is timestamped to obtain a synchronized construction parameter dataset, which is then cleaned to generate the construction parameter dataset.

3. The adaptive optimization method for tunneling parameters in pipe jacking construction as described in claim 1, characterized in that, The construction parameter dataset is obtained, standardized, and a time-series feature matrix is ​​constructed. Methods include: Modal decomposition is performed based on the construction parameter dataset to generate multiple frequency intrinsic mode components. High-frequency noise components are identified according to the multiple intrinsic mode components of frequency, and the construction parameter cleaning data is de-noised and reconstructed based on the high-frequency noise components to generate a noise-reduced construction parameter dataset. Based on the tunneling cycle set for pipe jacking construction, the noise reduction construction parameter dataset is divided into equal-length segments according to the tunneling cycle to generate multiple equal-length time series data segments, wherein the multiple equal-length time series data segments correspond to the tunneling cycle. The multiple time-series data segments of equal length are traversed and normalized to generate normalized time-series data; Statistical features are extracted based on the normalized time series data, and the statistical features are arranged in chronological order to construct the time series feature matrix.

4. The adaptive optimization method for tunneling parameters in pipe jacking construction as described in claim 1, characterized in that, A geological-parameter coupling model is constructed, and the time-series feature matrix is ​​fused with geological exploration data to generate a feature fusion dataset. The method includes: The geological-parameter coupling model is constructed based on convolutional neural networks and attention mechanisms, and the geological-parameter coupling model includes a first branch and a second branch; Geological exploration data is obtained based on the working face location information, and geological feature vectors are extracted based on the geological exploration data; The temporal feature matrix is ​​synchronized to the first branch of the geological-parameter coupling model for multi-layer convolution to extract local feature sets; Dynamic change analysis is performed based on the local feature set to construct a temporal deep feature map; The temporal deep feature map is expanded along the time dimension to obtain a sequence of feature vectors; The geological feature vector is synchronized to the second branch for feature space mapping to obtain the geological feature embedding vector; Based on the attention mechanism, the temporal deep feature map and the geological feature embedding vector are fused to construct the feature fusion dataset.

5. The adaptive optimization method for tunneling parameters in pipe jacking construction as described in claim 4, characterized in that, Based on the attention mechanism, feature fusion is performed on the temporal deep feature map and the geological feature embedding vector to construct the feature fusion dataset. The method includes: Based on the attention mechanism, the geological feature embedding vector is used as the query vector, and the feature vector sequence is used as the key and value to perform a dot product operation to obtain multiple attention scores; The multiple attention scores are normalized to generate multiple response weights; The feature vector sequence is weighted and summed according to the multiple response weights to obtain a geologically weighted time series feature vector. The geological weighted time-series feature vector and the geological feature embedding vector are concatenated and fused to generate a fused feature vector; Multi-layer sensing transformation is performed based on the fused feature vector to generate a deep coupling relationship between geology and parameters. Tunneling cycle analysis is then performed according to this deep coupling relationship to generate joint features between geology and parameters. The geological-parameter joint features are added to the feature fusion dataset.

6. The adaptive optimization method for tunneling parameters for pipe jacking construction as described in claim 1, characterized in that, Based on the aforementioned feature fusion dataset, attitude deviation prediction is performed to generate attitude offset prediction values. These attitude offset prediction values ​​are then used as constraints for multi-objective optimization of pipe jacking construction to determine the optimal combination of tunneling parameters. The method includes: The feature fusion dataset is used as the input sequence for temporal convolution to extract local temporal patterns. Long-range dependency capture is performed on the input sequence to obtain the feature depth hidden state; Based on the feature depth hidden state, mapping prediction is performed to obtain the horizontal offset prediction value and the vertical offset prediction value. The horizontal offset prediction value and the vertical offset prediction value are integrated with the attitude deviation to generate the attitude offset prediction value; Set the allowable range of axis deviation for pipe jacking construction, and compare the predicted attitude deviation value with the allowable range of axis deviation; When the predicted attitude offset value is greater than or equal to the allowable range of the axis deviation, an attitude over-limit warning signal is generated. Based on the attitude over-limit early warning signal, the predicted attitude deviation value is used as a constraint to perform multi-objective optimization for pipe jacking construction, and the optimal combination of tunneling parameters is determined.

7. The adaptive optimization method for tunneling parameters in pipe jacking construction as described in claim 6, characterized in that, Based on the attitude over-limit early warning signal, the predicted attitude deviation value is used as a constraint condition to perform multi-objective optimization of pipe jacking construction, and the optimal combination of tunneling parameters is determined. The method includes: Multiple optimization objectives are constructed, and the predicted attitude deviation value is converted into constraints for the multiple optimization objectives. Based on the attitude over-limit warning signal and the constraints, multiple allowable deviation thresholds are set, including horizontal allowable deviation thresholds and vertical allowable deviation thresholds. The predicted horizontal offset value is compared with the allowed horizontal offset threshold to determine the target horizontal offset range; the predicted vertical offset value is compared with the allowed vertical offset threshold to determine the target vertical offset range. Multi-target optimization is performed based on the target horizontal offset interval and the target vertical offset interval to construct candidate tunneling parameter combinations; The interaction between the pipe jacking machine and the soil is analyzed by iterating through the candidate combinations of tunneling parameters to generate tunneling speed response values ​​and surface settlement response values. Based on the tunneling speed response value and the surface settlement response value, a first optimization target value and a second optimization target value are set. The candidate tunneling parameter combinations are sorted non-dominated according to the first optimization target value and the second optimization target value, and multiple non-dominated levels are divided. Based on the candidate tunneling parameter combinations, a screening analysis is performed to determine the parent individuals to undergo crossover mutation and generate the offspring population. The parent individuals and the offspring population are merged and iterated according to multiple non-dominated levels to generate an optimal solution set with multiple non-dominated solutions; Membership is calculated based on the optimal solution set of the multiple non-dominated solutions, and multiple comprehensive satisfaction levels are generated to filter the optimal solution set of the multiple non-dominated solutions, thereby determining the optimal combination of tunneling parameters.

8. The adaptive optimization method for tunneling parameters in pipe jacking construction as described in claim 1, characterized in that, The method includes: simulating tunneling for pipe jacking based on the optimal combination of tunneling parameters, optimizing and controlling the simulated tunneling parameters, and generating control commands. A digital twin is used to simulate the tunneling cycle based on the optimal combination of tunneling parameters, generating simulated tunneling parameters, which include simulated tunneling speed parameters, simulated surface settlement parameters, and simulated attitude change trajectory parameters. Based on the multi-dimensional expected target value set for pipe jacking construction, the simulated tunneling speed parameter, the simulated surface settlement parameter, the simulated attitude change trajectory parameter are compared with the multi-dimensional expected target value; When any one of the simulated tunneling speed parameter, the simulated surface subsidence parameter, or the simulated attitude change trajectory parameter fails to meet the multidimensional expected target value, a control and optimization signal is triggered. The control conversion is performed through the aforementioned regulation and optimization signal to generate a combination of tunneling optimization parameters and construct the regulation command.

9. The adaptive optimization method for tunneling parameters in pipe jacking construction as described in claim 8, characterized in that, The methods for generating control commands and sending them to the controller to execute automatic parameter adjustments include: The combination of tunneling optimization parameters is converted into a parameter setting data package according to the controller's data interface. The parameter setting data packet is sent to the controller for parsing to generate parameter control setting values; The controller is activated to automatically adjust the tunneling parameters for pipe jacking construction according to the parameter control settings.

10. An adaptive optimization system for tunneling parameters used in pipe jacking construction, characterized in that, The system is used to implement the adaptive optimization method for tunneling parameters in pipe jacking construction according to any one of claims 1-9, the system comprising: The multi-dimensional tunneling acquisition module is used to perform real-time multi-dimensional tunneling acquisition based on pipe jacking construction, obtain construction parameter datasets, perform standardized processing, and construct a time-series feature matrix. The feature fusion module is used to construct a geological-parameter coupling model by fusing the time-series feature matrix with geological exploration data to generate a feature fusion dataset. The multi-objective optimization module is used to predict attitude deviation based on the feature fusion dataset, generate attitude offset prediction values, and use the attitude offset prediction values ​​as constraints to perform multi-objective optimization for pipe jacking construction to determine the optimal combination of tunneling parameters. The control and optimization module is used to simulate tunneling for pipe jacking construction based on the optimal combination of tunneling parameters, control and optimize based on the simulated tunneling parameters, and generate control commands to send to the controller for automatic parameter adjustment.

Images