Tunnel surrounding rock stress parameter inversion and deep learning stress prediction method
By combining PSO-XGBoost with LSTM, the problems of low computational efficiency and dynamic response in the inversion of tunnel surrounding rock stress parameters were solved, achieving efficient and accurate prediction of tunnel surrounding rock stress, and improving construction safety and the reliability of engineering decisions.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- TIANJIN UNIV
- Filing Date
- 2025-11-21
- Publication Date
- 2026-04-24
AI Technical Summary
Existing methods for inverting stress parameters of tunnel surrounding rock are inefficient, prone to getting trapped in local optima, and unable to dynamically respond to changes in the mechanical properties of the surrounding rock, thus affecting the safety of tunnel construction.
A fusion algorithm of particle swarm optimization and XGBoost model (PSO-XGBoost) combined with long short-term memory neural network (LSTM) is used to invert tunnel surrounding rock parameters and predict stress. Training samples are generated through finite element modeling and orthogonal experimental method, and multi-step prediction and accuracy evaluation are performed using measured stress data.
It significantly improves the efficiency and global convergence of surrounding rock mechanical parameter inversion, and enhances the accuracy of tunnel construction safety assessment and engineering decision-making.
Smart Images
Figure CN121920119A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of geotechnical mechanics parameter inversion technology, and in particular to a method for inverting stress parameters of tunnel surrounding rock and predicting stress through deep learning. Background Technology
[0002] Tunnel engineering, as a core area of underground space development, is widely used in water conservancy projects, transportation tunnels, and mining construction. With the development of deep mining technology, soft rock tunnels face complex geomechanical challenges such as high stress disturbance and significant rheological properties, making surrounding rock stability assessment a crucial aspect of engineering safety. Among related technologies, a complete technical system has been constructed, from material property characterization to stress distribution prediction, through the collaborative operation of finite element modeling, constitutive model selection, and parameter inversion. Specifically, this system covers the entire process from the Mohr-Coulomb model (used to simulate surrounding rock and concrete materials) to the classical metallic elastoplastic constitutive model (used for support structures such as anchor bolts and I-beams), including the elastic modulus. Poisson's ratio internal friction angle Cohesion The determination of key parameters and the sample generation strategy based on orthogonal experimental method form a closed-loop verification mechanism of parameters-stress-model.
[0003] However, existing parameter inversion methods, which directly employ traditional trial-and-error algorithms or single machine learning models, have not effectively addressed multivariate nonlinear optimization problems. This can lead to technical bottlenecks such as low computational efficiency and slow convergence. Specifically, traditional methods are prone to getting trapped in local optima when the parameter search space is large, and require numerous finite element calculations and iterations to obtain samples, significantly increasing computational costs. Furthermore, the risk of mismatch between experience-based parameter selection and limited monitoring point data leads to deviations between the inversion results and the actual stress distribution. Due to the high real-time requirements during the construction phase, existing technologies typically employ fixed parameter assumptions, but these have limitations such as an inability to dynamically respond to changes in the mechanical properties of the surrounding rock, thus affecting the reliability of numerical simulations. This deficiency is particularly prominent in deep soft rock engineering, potentially causing major safety hazards such as support structure failure and surrounding rock instability, thus hindering the technological development of intelligent monitoring and dynamic decision-making in underground engineering. Summary of the Invention
[0004] The present invention aims to at least partially solve one of the technical problems in the related art.
[0005] Therefore, the first objective of this invention is to propose a method for inverting stress parameters of tunnel surrounding rock and predicting stress using deep learning.
[0006] The second objective of this invention is to provide a device for inverting stress parameters of tunnel surrounding rock and predicting stress through deep learning.
[0007] The third objective of this invention is to provide an electronic device.
[0008] The fourth objective of this invention is to provide a computer-readable storage medium.
[0009] The fifth objective of this invention is to provide a computer program product.
[0010] To achieve the above objectives, a first aspect of the present invention proposes a method for inverting stress parameters of tunnel surrounding rock and predicting stress using deep learning, comprising: S1. Establish a finite element calculation model based on construction and survey data, determine the parameters to be inversely determined based on engineering experience, generate parameter combinations using orthogonal experimental method and perform finite element calculation, obtain training samples to construct a dataset. S2 uses the dataset constructed in S1 to train a tunnel surrounding rock parameter inversion model based on a fusion algorithm of particle swarm optimization and XGBoost model. S3. Input the measured stress data into the inversion model to obtain the inversion parameters and substitute them into the finite element model to calculate the stress, and generate the surrounding rock stress cloud map and stress curve. S4 uses a long short-term memory neural network algorithm to segment stress curve data into training and testing sets, performs multi-step predictions, and evaluates the model's prediction accuracy.
[0011] In one embodiment of the present invention, step S1 includes: S11, When determining the range of the inverse parameter, according to the geotechnical engineering code, the range of the elastic modulus E is limited to [10GPa, 50GPa], and the Poisson's ratio... The value range is limited to [0.2, 0.4], and the internal friction angle is... The value range of is limited to [25°, 45°], and the value range of cohesion c is limited to [0.5MPa, 3.0MPa]; S12, when generating parameter combinations through orthogonal experimental design, uses an L9(3^4) orthogonal array to design the experiment, with 3 levels for each parameter to ensure the uniform distribution of parameter combinations.
[0012] In one embodiment of the present invention, step S2 includes: S21, When preprocessing the parameter samples, the Z-score standardization method is used to convert the data into standardized data with a mean of 0 and a standard deviation of 1. S22. When selecting feature measurement points, based on the stress sensitivity analysis results, areas with stress change gradients greater than 0.5 MPa / m are preferentially selected as feature measurement points.
[0013] In one embodiment of the present invention, step S4 includes: S41, When adjusting the sliding time window length, the window length is set to an integer multiple of the construction cycle according to the stress curve fluctuation characteristics, and the ratio of the window length to the predicted step length is in the range of [2, 5]. S42, when performing multi-step prediction, at test set time... At that time, a recursive prediction strategy is adopted, using the model output of the previous time step as the input of the current time step to form a prediction sequence. .
[0014] In one embodiment of the present invention, it further includes: S5. After substituting the inversion parameters into the finite element model, the inversion parameters are verified in multiple stages based on the stress sensitivity analysis results, specifically including: S51, Calculate the relative error between the stress values and the measured values at each stage. ; S52, when relative error If the value is less than 5% for three consecutive construction stages, the inversion parameters are deemed valid and the calculation proceeds to the next stage.
[0015] To achieve the above objectives, a second aspect of the present invention provides a device for inverting stress parameters of tunnel surrounding rock and predicting stress using deep learning, comprising: The finite element modeling and parameter combination generation module is used to establish a finite element calculation model based on construction and survey data, determine the parameters to be inversely determined based on engineering experience, generate parameter combinations using the orthogonal experimental method, perform finite element calculations, and obtain training samples to construct a dataset. The fusion algorithm training module is used to train the tunnel surrounding rock parameter inversion model using the dataset constructed by S1 and based on the fusion algorithm of particle swarm optimization and XGBoost model. The parameter inversion and stress calculation module is used to input measured stress data into the inversion model, obtain inversion parameters, and substitute them into the finite element model to perform stress calculation, generating surrounding rock stress cloud map and stress curve. The multi-step prediction and accuracy evaluation module is used to segment stress curve data into training and testing sets based on the long short-term memory neural network algorithm, perform multi-step prediction, and evaluate the model's prediction accuracy.
[0016] To achieve the above objectives, a third aspect of the present invention provides an electronic device, comprising: a processor, and a memory communicatively connected to the processor; The memory stores computer-executed instructions; The processor executes computer execution instructions stored in the memory to implement the method as described in any one of the first aspects.
[0017] To achieve the above objectives, a fourth aspect of the present invention provides a computer-readable storage medium storing computer-executable instructions that, when executed by a processor, are used to implement the method as described in any one of the first aspects.
[0018] To achieve the above objectives, a fifth aspect of the present invention provides a computer program product that, when executed by a processor, implements the method described in any one of the first aspects.
[0019] The technical solutions provided by the embodiments of the present invention bring at least the following beneficial effects: The embodiments of the present invention significantly improve the efficiency and global convergence of the inversion of surrounding rock mechanical parameters, reduce the number of finite element calculations, and enhance the reliability of the inversion results, thereby improving the accuracy of tunnel construction safety assessment and engineering decision-making.
[0020] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description
[0021] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein: Figure 1 This is a flowchart illustrating a method for inverting stress parameters in tunnel surrounding rock and predicting stress using deep learning, provided in an embodiment of the present invention. Figure 2 This is a flowchart illustrating the process architecture of a method for inverting stress parameters in tunnel surrounding rock and predicting stress using deep learning, provided in an embodiment of the present invention. Figure 3 This is a three-dimensional finite element model of the surrounding rock of a tunnel provided in an embodiment of the present invention; Figure 4 This is a schematic diagram of the inversion technology process of a tunnel surrounding rock stress parameter inversion and deep learning stress prediction method provided in an embodiment of the present invention; Figure 5 This is a schematic diagram comparing the stress calculated using inversion parameters with the actual monitored stress, provided in an embodiment of the present invention. Figure 6 This is a schematic diagram comparing the LSTM predicted sequence and the actual sequence provided in an embodiment of the present invention; Figure 7 This is a schematic diagram of the structure of a tunnel surrounding rock stress parameter inversion and deep learning stress prediction device provided in an embodiment of the present invention. Detailed Implementation
[0022] Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.
[0023] Figure 1 This is a schematic flowchart illustrating a method for inverting stress parameters in tunnel surrounding rock and predicting stress using deep learning, provided in an embodiment of the present invention. Figures 1-6 As shown, the method includes the following steps: S1. Establish a finite element calculation model based on construction and survey data, determine the parameters to be inversely related based on engineering experience, generate parameter combinations using orthogonal experimental method and perform finite element calculations, and obtain training samples to construct a dataset.
[0024] Specifically, in step S1, establishing a finite element calculation model based on on-site construction and survey data is the fundamental step in the entire parameter inversion and stress prediction process. This step first constructs a three-dimensional finite element model that conforms to the actual engineering conditions using information such as geological exploration reports, geotechnical mechanics test data, and construction progress records. In the model, the surrounding rock and concrete materials are typically simulated using the Mohr-Coulomb constitutive model, while structural materials such as anchor bolts, I-beams, and secondary lining reinforcement are simulated using the classic metal elastoplastic constitutive model to more accurately reflect their mechanical response under complex stress states.
[0025] After determining the model structure, based on engineering experience and stress sensitivity analysis results, the parameters to be inversely selected, including the elastic modulus, were chosen. Poisson's ratio internal friction angle and cohesion These parameters are key variables affecting the stress distribution in the surrounding rock, and their value ranges are usually set according to standards such as the "Code for Geotechnical Investigation" (GB 50021) and the "Code for Design of Highway Tunnels" (JTG D70). For example, the elastic modulus... The range of values can be set to Poisson's ratio Usually in Between, internal friction angle Generally Cohesion The range is .
[0026] Furthermore, an orthogonal experimental design was employed to construct parameter combinations, reducing the number of experiments and improving the coverage efficiency of the parameter space. First, the experimental objective was defined, and the number of levels for each parameter was determined (e.g., 3 or 5 levels). Then, independent parameter combinations were generated using an orthogonal array. After each set of parameters was input into the finite element model, the stress response at the corresponding measurement points was calculated, forming training samples. Representative parameter sets were selected through statistical analysis to serve as the data foundation for subsequent PSO-XGBoost model training.
[0027] The technical value of this step lies in constructing a high-quality training dataset through systematic parameter combination design and finite element simulation, providing reliable input for the subsequent training of the parameter inversion model. Simultaneously, the introduction of the orthogonal experimental method significantly reduces computational costs and improves parameter search efficiency, laying a solid foundation for achieving efficient and accurate surrounding rock parameter inversion.
[0028] Furthermore, S1 includes: S11, When determining the range of the inverse parameter, according to the geotechnical engineering code, the range of the elastic modulus E is limited to [10GPa, 50GPa], and the Poisson's ratio... The value range is limited to [0.2, 0.4], and the internal friction angle is... The value range of is limited to [25°, 45°], and the value range of cohesion c is limited to [0.5MPa, 3.0MPa].
[0029] Specifically, in step S1031, determining the range of values for the inverse parameter is a crucial step in constructing the finite element training sample set. This technical implementation is based on geotechnical engineering specifications and material mechanical property analysis. Specifically, the elastic modulus... The range of values is limited to This range references the typical elastic modulus ranges for soft and medium-hard rock materials in the "Code for Geotechnical Investigation" (GB 50021), ensuring the selected parameters are engineering-appropriate and representative. Poisson's ratio The value range is [0.2, 0.4]. This range conforms to the Poisson's ratio characteristics of common soil and rock materials, thus avoiding the non-physical disturbance of extreme values to the finite element calculation results.
[0030] internal friction angle The value range is set to This parameter is one of the core parameters in the Mohr-Coulomb constitutive model. Its range is set according to the frictional characteristics of different rock mass strength grades in the "Code for Design of Geotechnical Engineering" (GB 50010) to ensure that the model has sufficient physical meaning when simulating the failure mechanism of surrounding rock. Cohesion The range of values is This range reflects the shear strength performance of soft rock under different geological conditions, and helps to capture the deformation and failure behavior of surrounding rock under different stress states.
[0031] In terms of implementation, this step utilizes orthogonal experimental design to divide the aforementioned parameters into several levels within their respective ranges, forming a parameter combination matrix. Each parameter combination is independent and covers different levels, thus achieving efficient sampling of the parameter space within a limited number of experiments. This method is widely used in sensitivity analysis of geotechnical engineering parameters, significantly reducing computational load and improving sample representativeness.
[0032] This step plays a fundamental role in the overall technical solution, providing high-quality and diverse finite element calculation samples for the subsequent training of the PSO-XGBoost model. By reasonably limiting the parameter range, the convergence speed and accuracy of parameter inversion can be effectively improved, providing reliable data support for modeling the mechanical behavior of tunnel surrounding rock, thereby enhancing the accuracy and engineering applicability of stress prediction during construction.
[0033] S12, when generating parameter combinations through orthogonal experimental design, uses an L9(3^4) orthogonal array to design the experiment, with 3 levels for each parameter to ensure the uniform distribution of parameter combinations.
[0034] Specifically, in step S103, L9(3) is used. 4 Orthogonal array design experiments are used to generate parameter combinations and ensure their uniform distribution in the parameter space. This method is based on the principle of orthogonal experimental design (OED), which selects a limited number of experimental points to cover multiple levels of multiple parameters with fewer experiments, thereby significantly reducing computational costs while ensuring experimental comprehensiveness.
[0035] In the specific implementation, L9(3 4 Orthogonal arrays are suitable for experimental designs with four parameters, each with three levels. The parameters to be inverseed in this invention include the elastic modulus. Poisson's ratio internal friction angle and cohesion Each parameter has a reasonable value range set based on engineering experience and stress sensitivity analysis results. For example, the elastic modulus... Possibly set as , , Other parameters are also set to three levels in a similar manner. By arranging and combining orthogonal arrays, nine sets of parameter combinations can be generated. Each set of parameter combinations satisfies the orthogonality requirement, that is, any two parameter level combinations appear with the same frequency in all experiments, thus ensuring uniform sampling of the parameter space.
[0036] During the finite element analysis (FEM) phase, each set of parameters is input into the established 3D finite element model. The model employs the Mohr-Coulomb constitutive model and a metal elastoplastic model to simulate the mechanical response of structures such as surrounding rock, concrete, and anchor bolts. After the calculation is completed, stress data from pre-selected measurement points are extracted to form a training sample set. This sample set is used for subsequent training of the PSO-XGBoost model, providing high-quality input-output pairs for parameter inversion.
[0037] This step plays a crucial role in the entire technical solution. By using the orthogonal experimental method, not only is the computational load of the finite element simulation reduced, but the representativeness and independence of the parameter combinations are also ensured, thereby improving the generalization ability and prediction accuracy of the inversion model. Furthermore, the uniformly distributed parameter combinations help reveal the sensitivity of each parameter to the surrounding rock stress, providing data support for subsequent optimization.
[0038] S2 uses the dataset constructed in S1 to train a tunnel surrounding rock parameter inversion model based on a fusion algorithm of particle swarm optimization and XGBoost model.
[0039] Specifically, in step S2, based on the finite element training samples constructed in S1, a fusion algorithm of particle swarm optimization (PSO) and XGBoost (PSO-XGBoost) is used to train the tunnel surrounding rock parameter inversion model. This fusion algorithm optimizes the hyperparameters of XGBoost through PSO, thereby improving the prediction accuracy and convergence efficiency of the model in the parameter inversion task.
[0040] At the technical implementation level, the PSO-XGBoost algorithm first addresses the key hyperparameters of the XGBoost model (such as the learning rate). Maximum tree depth Subsampling rate Regularization coefficient and The parameters (such as velocity and position) are encoded to form a search space for particles. Each particle represents a set of hyperparameter combinations, and its fitness function is defined as the model's prediction error on the training set, such as mean squared error (MSE) or root mean square error (RMSE). The PSO algorithm searches for the optimal hyperparameter combination by iteratively updating the particle's velocity and position, thereby constructing the optimal XGBoost model.
[0041] In terms of parameters, the PSO algorithm typically sets the number of particles to 20-50, the number of iterations to 50-100, and the inertia weight. The cognitive factor is generally dynamically adjusted within the range of [0.4, 0.9]. social factors The learning rate of XGBoost is typically set to 2. Set as Maximum tree depth The subsampling rate is 3~8. for Regularization coefficient and The strengths of L2 and L1 regularization are controlled separately, typically set to... .
[0042] In application scenarios, this fusion algorithm is suitable for inverting surrounding rock parameters during tunnel construction under complex geological conditions. By inputting stress data from pre-selected measurement points in the finite element simulation, the model can invert the elastic modulus of the surrounding rock. Poisson's ratio internal friction angle and cohesion Key mechanical parameters, etc. The inversion results can be further input into the finite element model for verification to ensure that the error between them and the measured stress data is within an acceptable range (e.g., RMSE < 5 MPa).
[0043] The technical advantage of this step lies in significantly improving the model's search efficiency and generalization ability in nonlinear, high-dimensional parameter spaces by optimizing the hyperparameter selection of XGBoost using PSO. This accelerates the parameter inversion process and improves inversion accuracy. Compared to traditional trial-and-error methods or single machine learning models, the PSO-XGBoost fusion algorithm can still achieve stable convergence even with a limited number of finite element samples, providing reliable data support for subsequent stress prediction and engineering decision-making.
[0044] Furthermore, S2 includes: S21. When preprocessing the parameter samples, the Z-score standardization method is used to convert the data into standardized data with a mean of 0 and a standard deviation of 1.
[0045] S22. When selecting feature measurement points, based on the stress sensitivity analysis results, areas with stress change gradients greater than 0.5 MPa / m are preferentially selected as feature measurement points.
[0046] Specifically, during parameter sample preprocessing, the Z-score standardization method is used to convert the data into standardized data with a mean of 0 and a standard deviation of 1. This method is based on statistical principles and improves the convergence speed and prediction accuracy of the subsequent PSO-XGBoost parameter inversion model by eliminating the influence of different parameter dimensions and numerical ranges on model training.
[0047] The core formula for Z-score standardization is:
[0048] in, This represents the original parameter sample value. This is the mean of the sample values for this parameter. Its standard deviation, The data is standardized. In this invention, the parameter to be inversely evaluated includes the elastic modulus. Poisson's ratio internal friction angle and cohesion These parameters exhibit significant numerical differences under varying engineering geological conditions, for example... The range of values may be ,and Usually in If these parameters are not standardized, scale differences will cause instability in the gradient descent process when input into the XGBoost model, affecting the model's generalization ability.
[0049] In its implementation, Z-score normalization first performs statistical calculations on each parameter sample column to obtain its mean. and standard deviation Subsequently, a linear transformation is performed on each sample point to center and normalize its distribution. This process can be efficiently implemented using numerical computing libraries such as NumPy or Pandas, ensuring the real-time performance and accuracy of data processing.
[0050] In addition, preprocessing also includes the detection and removal of outliers, typically using 3... The principle is to eliminate those that exceed [the limit]. The sample points are selected within a certain range to prevent noise from interfering with model training. The standardized parameter samples will serve as input features for the PSO-XGBoost model, calculating the surrounding rock stress at the pre-selected measurement points under the same finite element simulation conditions, thus providing a high-quality data foundation for subsequent model training.
[0051] This step plays a crucial bridging role in the entire technical solution, ensuring that the parameter samples have good statistical properties before being input into the machine learning model, thereby improving the stability and predictive ability of the parameter inversion model and providing reliable data support for subsequent stress analysis and prediction.
[0052] S3. Input the measured stress data into the inversion model to obtain the inversion parameters, and substitute them into the finite element model to calculate the stress, generating the surrounding rock stress cloud map and stress curve.
[0053] Specifically, in step S3, the measured stress data is input into the trained PSO-XGBoost parameter inversion model to obtain the inverted surrounding rock mechanical parameters, which are then substituted into the finite element model for stress calculation, ultimately generating a surrounding rock stress cloud map and stress curve. This step is a crucial step in realizing the quantitative analysis and visualization of the mechanical state of the tunnel surrounding rock.
[0054] At the technical implementation level, firstly, measured stress data is collected from stress monitoring points set up at the construction site. This data typically includes time-series stress response values, such as principal stresses. or shear stress Data acquisition needs to meet certain sampling frequency and accuracy requirements, typically using high-precision strain gauges or fiber Bragg grating sensors, with a sampling frequency of no less than 1Hz and an error controlled within ±1%. Subsequently, these measured data are input into the PSO-XGBoost inversion model, which optimizes the hyperparameters of XGBoost using the Particle Swarm Optimization (PSO) algorithm, thereby improving the model's generalization ability and inversion accuracy. The model output is the inverted material parameters, including the elastic modulus. Poisson's ratio internal friction angle and cohesion These parameters will be used to update the material properties of the finite element model.
[0055] In the finite element model, the Mohr-Coulomb constitutive model is used to simulate the surrounding rock and concrete materials, while the metal elastoplastic model is used for support structures such as anchor bolts and I-beams. After substituting the inversion parameters into the model, the stress field simulation for each stage of construction is performed again, and the stress distribution at each node is calculated. Stress cloud maps are generated using post-processing software (such as ANSYS, ABAQUS, or MIDAS) to visually reflect the stress state of the surrounding rock; at the same time, the stress-time curves of characteristic measuring points are extracted for mechanical behavior analysis during construction.
[0056] Regarding parameter indices, the output parameters of the inversion model must meet engineering specifications, such as the elastic modulus. The inversion error should be controlled within ±10%, and the internal friction angle With cohesion The inversion error should be less than ±15%. The mesh of the finite element model should use hexahedral or tetrahedral elements, with a mesh size not exceeding 0.5 m, to ensure the accuracy of the stress field calculation.
[0057] This step has significant application value in practical engineering. By combining the inversion model with finite element calculations, the stress state of areas without monitoring points can be predicted, providing data support for construction safety early warning and support optimization design. Simultaneously, the generated stress cloud maps and curves can serve as important bases for real-time decision-making by construction managers, improving the safety and intelligence level of tunnel construction.
[0058] S4 uses a long short-term memory neural network algorithm to segment stress curve data into training and testing sets, performs multi-step predictions, and evaluates the model's prediction accuracy.
[0059] Specifically, in step S4, based on the Long Short-Term Memory (LSTM) neural network algorithm, the surrounding rock stress curve data obtained from finite element calculations are divided into training and testing sets in a 4:1 ratio, and multi-step predictions are performed to evaluate the model's prediction accuracy. This step is the core component of deep learning stress prediction in this invention. Its technical implementation principle is based on the modeling capability of LSTM for time series data, which can effectively capture the nonlinear dynamic characteristics of stress changing over time.
[0060] At the technical implementation level, the stress response curves of the surrounding rock at multiple construction stages are first calculated using a finite element model. These curves, typically in units of time steps, record the stress changes at monitoring points under different construction conditions. To construct a supervised learning model, the original stress curve data needs to be converted into an input-output pair format. This involves extracting historical stress sequences as input features using a sliding time window approach, and using current or future stress values as output labels. For example, if the sliding time window length is set to... The input is The output is or This enables multi-step prediction.
[0061] At the parameter level, the dataset is split in a 4:1 ratio, meaning 80% is used for training and 20% for testing, which aligns with common data partitioning standards in deep learning model training. The hyperparameters of the LSTM model include the number of hidden layers, the number of neurons per layer, and the sliding time window length. The learning rate, batch size, etc. In some implementations, the hidden layers can be set to 2-3 layers, with each layer containing 64-128 neurons, and the sliding time window length... Typically, 5-10 steps are taken to balance the model's memory capacity and computational efficiency. During model training, mean squared error (MSE) and root mean square error (RMSE) are used as evaluation metrics, defined as follows:
[0062]
[0063] in, This represents the actual stress value. These are the model's predicted values. The number of test samples is [number]. By continuously adjusting the hyperparameters and calculating the above metrics, the optimal LSTM model can be obtained.
[0064] At the application level, this step is primarily used for predicting stress evolution during tunnel construction. In actual engineering, stress changes during the construction phase exhibit significant time dependence. LSTM models can predict stress responses several steps into the future based on historical stress data, thus providing data support for construction safety early warning and structural stability assessment. This is especially important during the testing phase, because... Since data at any given time is unavailable, the model must employ a recursive prediction method, using the output of the previous time step as the input for the next time step to maintain the continuity of the time window.
[0065] The technical effect of this step is that by modeling and predicting the finite element calculation results using a deep learning model, the accuracy and stability of stress curve prediction are improved, and reliable stress estimation is provided for areas where no monitoring points are set up during construction, thereby enhancing the comprehensiveness and foresight of tunnel structure safety analysis.
[0066] The tunnel surrounding rock stress parameter inversion and deep learning stress prediction method of this invention improves the efficiency and accuracy of surrounding rock parameter inversion, realizes multi-step prediction of stress curves, and enhances the safety of tunnel construction and the reliability of engineering decisions.
[0067] Furthermore, S4 includes: S41, When adjusting the sliding time window length, the window length is set to an integer multiple of the construction cycle according to the stress curve fluctuation characteristics, and the ratio of the window length to the predicted step length is in the range of [2, 5].
[0068] Specifically, the adjustment of the sliding time window length is based on the matching relationship between the fluctuation characteristics of the stress curve and the construction period. In particular, the sliding time window length... The settings need to meet two key conditions: First Construction period Integer multiples of, i.e. ,in Second, the window length and prediction step size. The ratio should be controlled within the interval [2, 5], that is... This setting aims to ensure that the time-series input data fully reflects the periodic characteristics of stress changes during construction, while also taking into account the model's ability to predict future stress states.
[0069] At the technical implementation level, the length of the sliding time window determines the amount of historical information that the LSTM neural network relies on when performing sequence modeling. If the window length is too short, it may fail to capture the stress evolution pattern within the construction cycle; if the window is too long, it will increase the computational complexity of the model and may lead to overfitting. Therefore, setting the window length to an integer multiple of the construction cycle allows the model to fully cover the stress response of one or more construction stages within each window, thereby improving the model's adaptability to periodic disturbances. In addition, controlling the ratio of the window length to the prediction step size within the range of [2, 5] is to establish a reasonable temporal dependency between the model input and output, ensuring the continuity and stability of the prediction results in the time dimension.
[0070] In practical operation, the length of the sliding time window The timeframe is typically set based on the cycle of key processes such as excavation, support, and secondary lining in the construction schedule. For example, if the construction cycle is 7 days, then a timeframe of 7 days can be set. , or Heaven, corresponding to , or Predicting step size This is set according to project management needs, and is typically 1 to 3 days. Adjustments can be made... and The ratio of [value] can optimize the error accumulation effect of the model in multi-step prediction.
[0071] This step is of great significance in practical engineering. During tunnel construction, the stress in the surrounding rock exhibits periodic changes with the construction sequence. By appropriately setting the sliding time window, the LSTM model can more accurately learn the temporal characteristics of the stress curve, improving the prediction accuracy of future stress states. Combined with subsequent multi-step prediction mechanisms (i.e., The prediction at a given moment depends on the model output at the previous moment. This setting helps to achieve dynamic prediction of stress changes in the later stages of construction, providing a reliable basis for construction safety early warning and engineering decision-making.
[0072] S42, when performing multi-step prediction, at test set time... At that time, a recursive prediction strategy is adopted, using the model output of the previous time step as the input of the current time step to form a prediction sequence. .
[0073] Specifically, when performing multi-step prediction, at time step of the test set... At that time, a recursive prediction strategy is adopted, using the model output of the previous time step as the input of the current time step to form a prediction sequence. This step is based on the time series prediction mechanism of LSTM (Long Short-Term Memory) neural network, which aims to address the problem of unavailable data for future moments during the testing phase, thereby enabling continuous prediction of the surrounding rock stress curve.
[0074] At the technical implementation level, the LSTM model, through its internal memory units and gating mechanism, can effectively capture the long-term dependencies of stress data over time. During training, the model learns to extract features from historical stress data and map them to nonlinear functions of future stress values. During testing, due to… Since the data at time point 1 is not provided, the model cannot directly obtain subsequent input sequences. Therefore, a recursive prediction strategy is adopted, that is, utilizing the model's... Output at time As The inputs at each time point are recursively processed to form a sequence of length . The predicted sequence. This strategy ensures continuous sliding of the time window, thereby maintaining the model's adaptability to the time series structure.
[0075] At the parameter level, the step size of recursive prediction Typically set based on the engineering monitoring cycle and predicted needs, for example or This corresponds to stress prediction over 5 or 10 time steps. The input dimension of the model is determined by the length of the sliding time window. Decision, generally set as to This is to fully reflect the dynamic characteristics of stress changes. Prediction errors are evaluated using mean square error (MSE) and root mean square error (RMSE), calculated using the following formulas:
[0076]
[0077] At the application level, this step is mainly used for stress trend prediction in the later stages of tunnel construction or operation. For example, after the excavation is completed, finite element simulation is performed using the inverted surrounding rock mechanical parameters to obtain the stress curve as model input. The LSTM model then recursively predicts stress changes at several future moments, providing early warning information and decision support for construction managers.
[0078] The technical advantage of this step lies in its effective resolution of the problem of missing future data during the testing phase through a recursive prediction mechanism, thereby improving the model's generalization ability and predictive continuity in practical engineering. Simultaneously, by combining the high-precision parameters provided by the PSO-XGBoost inversion model, the reliability of stress prediction is further enhanced, providing crucial data support for tunnel structure safety assessment and construction risk control.
[0079] The tunnel surrounding rock stress parameter inversion and deep learning stress prediction method of this invention improves the efficiency and accuracy of surrounding rock parameter inversion, realizes multi-step prediction of stress curves, and enhances the safety of tunnel construction and the reliability of engineering decisions.
[0080] S5. After substituting the inversion parameters into the finite element model, the inversion parameters are verified in multiple stages based on the stress sensitivity analysis results, specifically including: Specifically, the key step in this invention to achieve parameter inversion accuracy control and model reliability assessment is to perform multi-stage verification of the inversion parameters after substituting them into the finite element model based on the stress sensitivity analysis results. The technical principle behind this step is based on the comparative analysis of finite element numerical simulation and measured stress data, ensuring the applicability and stability of the inversion parameters under different construction stages through a multi-stage verification mechanism.
[0081] At the technical implementation level, the material parameters (such as elastic modulus) obtained by inversion through the PSO-XGBoost algorithm will first be... Poisson's ratio internal friction angle Cohesion The stress data is input into the established finite element model. This model typically uses the Mohr-Coulomb constitutive model to simulate the surrounding rock and concrete materials, while metal components (such as anchor bolts, I-beams, and secondary lining reinforcement) are modeled using the classic metal elastoplastic constitutive model. Subsequently, during the construction stages corresponding to the measured stress data (such as excavation, support, and secondary lining), stress calculations are performed on the model stage by stage, generating stress cloud maps and stress-time curves for the surrounding rock.
[0082] At the parameter level, the validation process needs to focus on several key indicators, including but not limited to peak stress error, stress distribution pattern matching degree, and time series consistency. Specifically, the mean square error (MSE) and root mean square error (RMSE) between the output stress of the inversion model and the measured stress can be calculated, and their definitions are as follows:
[0083]
[0084] in, This represents the measured stress value. This represents the stress value calculated using the finite element model. The number of samples. By setting an error threshold (e.g.) This can determine whether the inversion parameters meet the engineering accuracy requirements.
[0085] In terms of application scenarios, this step is suitable for parameter verification and model correction throughout the entire tunnel construction process. For example, during the excavation stage, if the arch stress predicted by the inversion model deviates significantly from the measured value, it is necessary to readjust the range of inversion parameters or optimize the training strategy of the PSO-XGBoost model. After the support structure is applied, the consistency between the support effect and the surrounding rock response can be further verified by comparing the stress distribution cloud maps of the model and the measured values.
[0086] The technical effect of this step is that, through a multi-stage verification mechanism, the applicability of the inversion parameters and the prediction accuracy of the model are effectively improved, providing high-quality input data for subsequent LSTM-based stress curve prediction, thereby enhancing the safety assessment capability and engineering decision support level during tunnel construction.
[0087] S51, Calculate the relative error between the stress values and the measured values at each stage. .
[0088] Specifically, calculating the relative error (RE) between the stress values and the measured values at each stage is an important technical means to evaluate the accuracy and reliability of the inversion parameter model. Its mathematical expression is:
[0089] in, Indicates the first The stress values obtained from the finite element simulation at each monitoring point This represents the stress value measured at that point during actual construction. This formula quantifies the degree of error in the model prediction by calculating the relative deviation between the simulated and measured values, thus providing a basis for subsequent model correction and parameter optimization.
[0090] At the technical implementation level, this step first uses the PSO-XGBoost parameter inversion model trained in step S2 as input, and inputs the stress data from the actual monitoring points to invert the mechanical parameters of the surrounding rock (such as elastic modulus). Poisson's ratio internal friction angle Cohesion Subsequently, the parameters obtained from the inversion are re-imported into the finite element model to perform stress simulation calculations for each stage of construction. The simulation results are matched one-to-one with the measured stress data at the same time point and spatial location. The relative error is calculated point by point using the above formula, and finally an error distribution map or error statistics table is generated.
[0091] At the parameter level, relative error Typically expressed as a percentage, its numerical range directly affects the model's reliability. In engineering practice, it is generally required that... This is to ensure a high degree of consistency between the simulation results and the measured data. If the error exceeds this threshold, iterative optimization of the inversion model is required, adjusting parameters such as the number of particles, number of iterations, and inertia weights in the PSO algorithm, or recalibrating hyperparameters such as the tree depth and learning rate of XGBoost to improve the model's fitting ability.
[0092] In application scenarios, this step is widely used for real-time stress monitoring and parameter correction during tunnel construction. For example, during critical construction stages such as excavation, support, and secondary lining, actual stress data is collected by deploying stress monitoring points (such as anchor stress gauges and surrounding rock pressure cells) and compared with simulation results. This method is particularly suitable for engineering scenarios with significant surrounding rock deformation and complex stress distribution in soft rock strata, effectively identifying the sources of model errors and improving the accuracy of finite element simulations.
[0093] From a technical perspective, this step provides quantitative evidence for the verification of the inversion model, ensuring the engineering applicability of the inversion parameters. Simultaneously, error analysis can identify potential issues in the finite element model, such as deviations in boundary condition settings and inappropriate selection of material constitutive models, thus providing reliable data support for subsequent construction safety assessments and support scheme optimization.
[0094] S52, when relative error If the value is less than 5% for three consecutive construction stages, the inversion parameters are deemed valid and the calculation proceeds to the next stage.
[0095] Specifically, in the tunnel surrounding rock mechanical parameter inversion method of the present invention, when the relative error If the error rate is less than 5% for three consecutive construction stages, the inversion parameters are deemed valid and the calculation proceeds to the next stage. This step is a crucial link in the parameter inversion model verification and iterative optimization process, and its technical implementation is based on an error assessment mechanism and dynamic feedback from the construction stages.
[0096] From a technical implementation perspective, this step involves comparing the stress values output by the inversion model with the measured stress values at each construction stage to calculate the relative error. Its definition is:
[0097] in, Indicates the first Stress values calculated by the finite element model during each construction stage This represents the measured stress value in the corresponding stage. This error assessment is performed simultaneously at multiple feature measurement points to ensure the accuracy of the model in spatial distribution. In three consecutive stages, if all feature points... If all values are less than 5%, it indicates that the current inversion parameters can reflect the actual mechanical behavior of the surrounding rock well, and the model has good convergence and stability.
[0098] At the parameter level, the 5% relative error threshold is set based on engineering experience and error sensitivity analysis, and is typically used to measure the consistency between numerical simulations and measured data. This threshold is widely accepted in geotechnical engineering and meets the requirements for model validation error control in the "Code for Numerical Analysis of Geotechnical Engineering" (GB 50021). Furthermore, the setting of three consecutive stages is to avoid misjudgments due to data fluctuations or measurement errors in a single time period, thereby enhancing the robustness of parameter determination.
[0099] In practical applications, this step is typically deployed in a real-time monitoring system during the construction phase. It combines the finite element model with data from on-site stress sensors to dynamically validate the inversion model. For example, during construction stages such as excavation, support, and secondary lining, the system automatically collects stress data from monitoring points and compares it with the model's predicted values to determine if the current inversion parameters are still applicable. If valid, the system proceeds to the next stage of parameter updating and model training; otherwise, the parameter combinations need to be readjusted and a new round of finite element calculations performed.
[0100] The technical advantage of this step lies in its ability to effectively filter out unreasonable parameter combinations by setting strict error thresholds and stage continuity requirements, ensuring that the mechanical parameters output by the inversion model have engineering applicability and physical rationality. Simultaneously, this mechanism provides high-quality input data for subsequent LSTM-based stress curve prediction, thereby improving the accuracy and reliability of the entire prediction model and providing solid data support for construction safety assessment and engineering decision-making.
[0101] The tunnel surrounding rock stress parameter inversion and deep learning stress prediction method of this invention effectively screens and confirms the reliability of inversion parameters through a multi-stage verification mechanism combined with stress sensitivity analysis, significantly improving the prediction consistency and engineering applicability of the finite element model, thereby further enhancing the stress prediction accuracy and structural stability assessment capability throughout the tunnel construction process.
[0102] To achieve the above embodiments, the present invention also proposes a device for inverting stress parameters of tunnel surrounding rock and for deep learning stress prediction. Figure 7 This is a schematic diagram of a device for inverting stress parameters of tunnel surrounding rock and predicting stress using deep learning, provided as an embodiment of the present invention. Figure 7 As shown, the device includes: The finite element modeling and parameter combination generation module is used to establish a finite element calculation model based on construction and survey data, determine the parameters to be inversely determined based on engineering experience, generate parameter combinations using the orthogonal experimental method, perform finite element calculations, and obtain training samples to construct a dataset. The fusion algorithm training module is used to train the tunnel surrounding rock parameter inversion model using the dataset constructed by S1 and based on the fusion algorithm of particle swarm optimization and XGBoost model. The parameter inversion and stress calculation module is used to input measured stress data into the inversion model, obtain inversion parameters, and substitute them into the finite element model to perform stress calculation, generating surrounding rock stress cloud map and stress curve. The multi-step prediction and accuracy evaluation module is used to segment stress curve data into training and testing sets based on the long short-term memory neural network algorithm, perform multi-step prediction, and evaluate the model's prediction accuracy.
[0103] Regarding the apparatus in the above embodiments, the specific manner in which each module performs its operation has been described in detail in the embodiments related to the method, and will not be elaborated upon here.
[0104] To implement the above embodiments, the present invention also proposes an electronic device, including: a processor and a memory communicatively connected to the processor; the memory stores computer execution instructions; the processor executes the computer execution instructions stored in the memory to implement the method provided in the foregoing embodiments.
[0105] To implement the above embodiments, the present invention also proposes a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement the methods provided in the foregoing embodiments.
[0106] To implement the above embodiments, the present invention also proposes a computer program product, including a computer program that, when executed by a processor, implements the methods provided in the foregoing embodiments.
[0107] The collection, storage, use, processing, transmission, provision, and disclosure of user personal information involved in this invention all comply with the provisions of relevant laws and regulations and do not violate public order and good morals.
[0108] It should be noted that personal information collected from users should be used for legitimate and reasonable purposes and should not be shared or sold outside of these legitimate uses. Furthermore, such collection / sharing should only be conducted after receiving the user's informed consent, including but not limited to notifying the user to read the user agreement / user notice and sign an agreement / authorization that includes authorization of relevant user information before the user uses the function. In addition, any necessary steps must be taken to protect and safeguard access to such personal information data and ensure that others with access to personal information data comply with their privacy policies and procedures.
[0109] This invention is intended to provide implementation schemes for users to selectively prevent the use or access to personal information data. That is, this disclosure is intended to provide hardware and / or software to prevent or block access to such personal information data. Once personal information data is no longer needed, risks can be minimized by restricting data collection and deleting data. Furthermore, where applicable, such personal information can be de-identified to protect user privacy.
[0110] In the foregoing descriptions of the embodiments, the terms "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.
[0111] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly specified.
[0112] Any process or method description in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing custom logic functions or processes, and the scope of preferred embodiments of the invention includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as should be understood by those skilled in the art to which embodiments of the invention pertain.
[0113] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include: an electrical connection having one or more wires (electronic device), a portable computer disk drive (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Alternatively, the computer-readable medium may be paper or other suitable media on which the program can be printed, since the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in a computer memory.
[0114] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware as in another embodiment, it can be implemented using any of the following techniques known in the art, or a combination thereof: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.
[0115] Those skilled in the art will understand that all or part of the steps of the methods in the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, the program includes one or a combination of the steps of the method embodiments.
[0116] Furthermore, the functional units in the various embodiments of the present invention can be integrated into a processing module, or each unit can exist physically separately, or two or more units can be integrated into a module. The integrated module can be implemented in hardware or as a software functional module. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium.
[0117] The storage medium mentioned above can be a read-only memory, a disk, or an optical disk, etc. Although embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present invention.
[0118] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this invention can be achieved, and this is not limited herein.
[0119] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.
Claims
1. A method for inverting stress parameters of tunnel surrounding rock and predicting stress using deep learning, characterized in that, Includes the following steps: S1. Establish a finite element calculation model based on construction and survey data, determine the parameters to be inversely determined based on engineering experience, generate parameter combinations using orthogonal experimental method and perform finite element calculation, obtain training samples to construct a dataset. S2 uses the dataset constructed in S1 to train the tunnel surrounding rock parameter inversion model based on the fusion algorithm of particle swarm optimization and XGBoost model; S3. Input the measured stress data into the inversion model to obtain the inversion parameters and substitute them into the finite element model to calculate the stress, and generate the surrounding rock stress cloud map and stress curve. S4 uses a long short-term memory neural network algorithm to segment stress curve data into training and testing sets, performs multi-step predictions, and evaluates the model's prediction accuracy.
2. The method according to claim 1, characterized in that, Step S1 includes: S11, When determining the range of the inverse parameter, according to the geotechnical engineering code, the range of the elastic modulus E is limited to [10GPa, 50GPa], and the Poisson's ratio... The value range is limited to [0.2, 0.4], and the internal friction angle is... The value range of is limited to [25°, 45°], and the value range of cohesion c is limited to [0.5MPa, 3.0MPa]; S12, when generating parameter combinations through orthogonal experimental design, uses an L9(3^4) orthogonal array to design the experiment, with 3 levels for each parameter to ensure the uniform distribution of parameter combinations.
3. The method according to claim 1, characterized in that, Step S2 includes: S21, When preprocessing the parameter samples, the Z-score standardization method is used to convert the data into standardized data with a mean of 0 and a standard deviation of 1. S22. When selecting characteristic measurement points, based on the stress sensitivity analysis results, areas with stress change gradients greater than 0.5 MPa / m are preferentially selected as characteristic measurement points.
4. The method according to claim 1, characterized in that, Step S4 includes: S41, When adjusting the sliding time window length, the window length is set to an integer multiple of the construction cycle according to the stress curve fluctuation characteristics, and the ratio of the window length to the predicted step length is in the range of [2, 5]. S42, when performing multi-step prediction, at test set time... At that time, a recursive prediction strategy is adopted, using the model output of the previous time step as the input of the current time step to form a prediction sequence. .
5. The method according to claim 1, characterized in that, Also includes: S5. After substituting the inversion parameters into the finite element model, the inversion parameters are verified in multiple stages based on the stress sensitivity analysis results, specifically including: S51, Calculate the relative error between the stress values and the measured values at each stage. ; S52, when relative error If the value is less than 5% for three consecutive construction stages, the inversion parameters are deemed valid and the calculation proceeds to the next stage.
6. A device for inverting stress parameters of tunnel surrounding rock and predicting stress using deep learning, characterized in that, include: The finite element modeling and parameter combination generation module is used to establish a finite element calculation model based on construction and survey data, determine the parameters to be inversely determined based on engineering experience, generate parameter combinations using the orthogonal experimental method, perform finite element calculations, and obtain training samples to construct a dataset. The fusion algorithm training module is used to train the tunnel surrounding rock parameter inversion model using the dataset constructed by S1 and based on the fusion algorithm of particle swarm optimization and XGBoost model. The parameter inversion and stress calculation module is used to input measured stress data into the inversion model, obtain inversion parameters, and substitute them into the finite element model to perform stress calculation, generating surrounding rock stress cloud map and stress curve. The multi-step prediction and accuracy evaluation module is used to segment stress curve data into training and testing sets based on the long short-term memory neural network algorithm, perform multi-step prediction, and evaluate the model's prediction accuracy.
7. An electronic device, characterized in that, include: A processor, and a memory communicatively connected to the processor; The memory stores computer-executed instructions; The processor executes computer execution instructions stored in the memory to implement the method as described in any one of claims 1-5.
8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the method as described in any one of claims 1-5.
9. A computer program product, characterized in that, Includes a computer program that, when executed by a processor, implements the method of any one of claims 1-5.