Vertical shaft freezing method construction parameter dynamic regulation and control method based on optical fiber sensing

Abstract

The invention discloses a vertical shaft freezing method construction parameter dynamic regulation and control method based on optical fiber sensing, and the method comprises the following steps: arranging a sensing optical cable on a freezing wall and setting an early warning threshold value, pre-burying a sensor in a well wall and setting an early warning threshold value, collecting the parameters of the freezing wall and the well wall in real time, and analyzing the parameters to form original data; preprocessing and feature extraction are performed on original data; field construction parameters are integrated and then transmitted to a monitoring center; a double-source data set is constructed; a three-dimensional geology-freezing-construction coupling model is established; whether adjustment is needed or not is judged, and the construction process is dynamically optimized; according to the method, construction parameters can be converted from experience driving to data driving, the safety and efficiency of vertical shaft construction in the complex stratum are improved, and the method is suitable for vertical shaft construction under the complex conditions of a deep aquifer and a weakly cemented stratum.

Description

Technical Field

[0001] This invention belongs to the technical field of construction control methods, and relates to a dynamic control method for construction parameters of vertical shaft freezing method based on optical fiber sensing. Background Technology

[0002] Shaft freezing is an important and specialized shaft sinking method used in deep aquifers and weakly cemented strata. Due to its advantages of not being limited by drilling diameter and depth and having good controllability of the construction period, it accounts for over 90% of specialized shaft sinking methods. However, the determination of construction parameters for the current shaft freezing method mainly relies on experience and static calculations. On the one hand, traditional methods do not fully consider the complex and variable factors during construction, such as significant differences in geological conditions, with varying soil types, water content, and permeability in different areas; the real-time state of the frozen wall and the shaft wall is also dynamically changing. As construction progresses, the temperature field distribution, strength characteristics, and stress and strain borne by the shaft wall are constantly changing. On the other hand, existing methods are difficult to adapt to the needs of deep and large-scale shaft construction. Deep strata face complex conditions such as high water pressure, high earth pressure, high ground stress, and high additional loads, while large-scale shafts bring greater workload and difficulty in excavation, support, installation, and transportation.

[0003] As the construction depth of vertical shafts increases, the frozen wall thickness calculated by traditional methods may be extremely unreasonable and fail to meet the actual construction safety requirements. Furthermore, engineering accidents caused by unreasonable parameters are common, resulting in huge economic losses. At the same time, traditional methods cannot adjust construction parameters in real time, leading to low construction efficiency and serious waste of resources, making it difficult to meet the requirements of modern engineering construction for high efficiency, safety, and economy. Summary of the Invention

[0004] The purpose of this invention is to provide a dynamic control method for construction parameters of vertical shaft freezing method based on fiber optic sensing, which solves the problem of the inability to adjust construction parameters in real time in the existing technology.

[0005] The technical solution adopted in this invention is a dynamic control method for construction parameters of vertical shaft freezing method based on optical fiber sensing, which includes the following steps: Step 1: Lay sensor optical cables on the frozen wall and set early warning thresholds; pre-embed sensors in the well wall and set early warning thresholds. Step 2: Collect and analyze the parameters of the frozen wall and wellbore in real time to generate raw data; Step 3: Preprocess and extract features from the raw data; Step 4: Integrate the on-site construction parameters and transmit them to the monitoring center; Step 5: Construct a dual-source dataset, establish a three-dimensional geology-freezing-construction coupled model, and determine the optimal construction parameters under different geological conditions; Step 6: Compare the construction parameters monitored by the sensors with the optimal construction parameters to determine whether adjustments are needed; Step 7: Dynamically optimize the construction process.

[0006] The invention is further characterized by: Step 1 is performed as follows: Step 1.1: Densely distributed temperature sensing optical cables and densely distributed strain sensing optical cables are laid down along the freezing holes of the frozen wall to monitor the temperature of the frozen wall and the strain of the rock mass. At the same time, an audible and visual alarm is installed. Step 1.2: Arrange multiple fiber Bragg grating temperature sensors, multiple fiber Bragg grating concrete strain gauges, multiple fiber Bragg grating rebar stress gauges, multiple inter-wall fiber Bragg grating pressure sensors, and multiple fiber Bragg grating water pressure sensors in an array along the depth direction of the well wall. At the same time, install an audible and visual alarm. Place a fiber Bragg grating demodulator in the computer room on the ground. Connect the sensors in series through their own optical fibers, and then connect them to the fiber Bragg grating demodulator on the ground through the main communication optical cable. Step 1.3: Set the warning threshold for densely distributed temperature sensing optical cables to -30℃ to -25℃, the warning threshold for fiber optic temperature sensors to -10℃ to -8℃, the warning threshold for fiber optic concrete strain gauges to 2000με to 2400με, the warning threshold for fiber optic rebar stress gauges to 30MPa to 32MPa, the warning threshold for wall-mounted fiber optic pressure sensors to 10MPa to 12MPa, and the warning threshold for fiber optic water pressure gauge sensors to 0 to 0.5MPa.

[0007] Step 2 is performed as follows: Step 2.1: Use industrial Ethernet to synchronously collect parameters of the frozen wall and well wall with the operating parameters of the on-site construction equipment. The collection frequency is set to 0.8 times / second to 1.2 times / second. The parameters of the frozen wall and well wall include the temperature of the frozen wall, the rock mass strain of the frozen wall, the temperature of the concrete of the well wall, the concrete strain of the well wall, the deformation stress of the reinforcing steel of the well wall, the contact pressure between the well wall and the frozen wall, and the seepage pressure between the two layers of well walls. The operating parameters of the on-site construction equipment include the flow rate of the freezer, the temperature of the brine inlet and outlet of the pipe, the power of the chiller, the output flow rate of the grouting pump, and the temperature of the grouting liquid. At the same time, alarm indicator lights and buzzers are installed on the control consoles of the freezer and the grouting pump, respectively. Step 2.2: Receive the optical signal transmitted through the main communication optical cable using a fiber optic grating demodulator, convert the wavelength change information generated in the optical signal into an electrical signal, and then calculate the quantitative parameter values ​​of the frozen wall and well wall parameters. Simultaneously mark the acquisition time and the operating parameters of the on-site construction equipment to form raw data.

[0008] Step 3 is performed as follows: Step 3.1: Mark the raw data that exceeds the warning threshold of each sensor as suspected abnormal data. For suspected abnormal data, retrieve the synchronous data of at least 3 adjacent fiber optic sensors of the same type around it for neighborhood verification. For suspected abnormal data with an absolute value of temperature deviation exceeding 2℃, an absolute value of stress deviation exceeding 1MPa, or an absolute value of strain deviation exceeding 100με, determine them as real abnormal data. Step 3.2: Replace the actual outliers using linear interpolation.

[0009] in, x t 1. x t+1 This represents normal data before and after the abnormal event. This refers to the time when actual outlier values ​​were collected. Step 3.3: Normalize the original data using the min-max standardization method:

[0010] in, x The original data, x min , x max This is the theoretical extreme value of the data; Step 3.4 involves downsampling the temperature of the frozen wall, the temperature of the well wall concrete, the temperature of the brine inlet and outlet in the pipe, and the temperature of the grouting liquid. The virtual data points per second are generated by linear interpolation for the frozen wall rock mass strain, the well wall concrete strain, the deformation stress of the well wall reinforcement, the contact pressure between the well wall and the frozen wall, the seepage pressure between the two layers of well walls, the flow rate of the freezer, the power of the chiller, and the output flow rate of the grouting pump.

[0011] Step 4 is performed as follows: Step 4.1: Collect geological parameters during the preliminary exploration stage. Geological parameters include lithology, water content, permeability, compressive strength, and internal friction angle of the rock mass. Set up a monitoring center on the ground. The monitoring center is equipped with a data server, a visualization platform, an audible and visual alarm device, and an alarm control host. The data server is connected to the alarm control host and the visualization platform. The alarm control host is connected to the audible and visual alarm device. Step 4.2: The on-site construction equipment operating parameters and geological parameters are synchronously transmitted to the data server of the ground monitoring center through the data interaction interface. The data preprocessing module unifies the format and aligns the timestamps of the construction equipment operating parameters and geological parameters to form a related dataset containing "geological conditions-equipment operating status". The visualization platform displays the temperature field distribution, stress cloud map and related dataset in real time.

[0012] Step 5 is performed as follows: Step 5.1: Select at least 50 vertical shaft freezing construction projects, covering strata types such as siltstone, fine-grained sandstone, medium-grained sandstone, coarse-grained sandstone, and sandy mudstone, as well as construction scenarios under a water content gradient of 5%-25%. Include successful cases of stable frozen wall formation and crack-free shaft walls, as well as typical cases of localized freezing wall melting and excessive concrete strain in the shaft wall. For each case, collect geological parameters, frozen wall and shaft wall parameters, and on-site construction equipment operating parameters. All monitored values ​​must be within the thresholds specified in the "Coal Mine Shaft and Tunnel Engineering Construction Standard." All data should be timestamped and used as actual case data. Monitored values ​​include shaft wall concrete strain, shaft wall steel reinforcement deformation stress, shaft wall concrete temperature, and frozen wall radial temperature and strain. Step 5.2: Establish a three-dimensional geological-freezing-construction coupled model using COMSOL. Use "free tetrahedral mesh" to partition the entire model. Reduce the mesh size to 0.2m-0.5m around the freezing hole and the contact zone between the well wall and the freezing wall. Use a mesh size of 1m-2m for the strata far from the well. Step 5.3: Use the transient solver to solve the nonlinear equation system using the "Newton-Raphson iteration method" to calculate the parameters of the frozen wall and wellbore under various working conditions. During the calculation process, monitor the convergence curve in real time to ensure that the residual is less than 1e-6. Extract the calculation results and organize them into a dataset in the format of "parameter input-monitoring output" as simulation data. Step 5.4: Integrate the actual case data and the simulation data in a 6:4 ratio to form a dual-source dataset with at least 150,000 data entries; Step 5.5: Select features from the dual-source dataset. Take formation type, water content, and compressive strength as input features, and extract brine temperature, brine flow rate, grouting flow rate, and grout temperature as key variables. Use "whether the well wall concrete temperature, well wall concrete strain, well wall steel reinforcement deformation stress, frozen wall temperature, and frozen wall rock mass strain are all within the warning threshold" as the output label. "Yes" is marked as 1, and "No" is marked as 0. Step 5.6: The gradient boosting tree algorithm is used for training. The dual-source dataset is divided into training and validation sets in an 8:2 ratio. Overfitting is controlled by 5-fold cross-validation. The tree depth and learning rate are optimized by grid search. At least 1000 iterations of training are performed. When the classification accuracy on the validation set is not less than 92%, the K-means clustering algorithm is used to group the output "candidate optimal parameters" to determine the optimal construction parameters under different geological conditions.

[0013] The three-dimensional geological-freezing-construction coupled model is based on the coupling of three fields: heat conduction, porous media seepage, and structural mechanics. It employs COMSOL's built-in coupling algorithm to calculate multiphysics interactions. The heat conduction algorithm is as follows:

[0014] in, Where is the density of the medium, c is the specific heat capacity, T is the temperature, t is the time, k is the thermal conductivity, and Q is the intensity of the internal heat source; Porous media flow algorithm:

[0015] in, For penetration rate, Pore ​​water pressure, For the density of water, Let t be the compressibility coefficient of water, and t be time. Structural mechanics algorithms:

[0016] in, .

[0017] The steps for grouping the "candidate optimal parameters" output by the model using the K-means clustering algorithm are as follows: Step a: Select the construction parameter combinations with output label 1 from the gradient boosting tree model output results as "candidate optimal parameters", and clarify the clustering characteristics as formation type, formation water content, brine temperature, brine flow rate, and grouting temperature; Step b: Based on the strata types of siltstone, fine-grained sandstone, medium-grained sandstone, coarse-grained sandstone, and sandy mudstone covered in the dual-source dataset, as well as the water content gradient of 5%-25%, and combined with the objective of "the same geological conditions correspond to a set of optimal parameter intervals", the K value is determined by the elbow rule, the sum of squared errors within clusters under different K values ​​is calculated, and the K value corresponding to the point where the rate of decrease of the sum of squared errors within clusters drops sharply is selected. Step c: Randomly select K "candidate optimal parameter" samples as initial cluster centers, calculate the Euclidean distance between each remaining parameter sample and each cluster center, assign the sample to the nearest cluster, calculate the mean of each feature of all samples in each cluster, update it as a new cluster center, and repeat the "distance calculation-sample assignment-center update" steps until the center coordinate change is less than 0.1% for 3 consecutive iterations or the number of iterations reaches 100. Step d: For each group of parameters after clustering, calculate the 95% confidence interval for each construction parameter to form a "geological conditions-optimal construction parameter interval" correspondence table. At the same time, verify the validity of the interval by randomly selecting 10% of the parameter combinations in each interval and inputting them into the gradient boosting tree model to verify whether its output label is 1. Ensure that more than 95% of the parameter combinations in the interval can make the monitoring value meet the standard, and finally output the structured grouping results.

[0018] Step 6 specifically involves: Step 6.1: Compare the optimal construction parameters with the current construction parameters and calculate the parameter deviation rate. : ×100% Where m represents the current construction parameters and n represents the optimal construction parameters; When the deviation rate is 5%-10%, an audible and visual alarm is triggered. At the same time, combined with the real-time frozen wall and well wall parameters, the corresponding table of "geological conditions-optimal construction parameter range" is called to generate fine-tuning suggestions. When the deviation rate is greater than 10% or the current construction parameters exceed the warning threshold, a pop-up alarm is triggered, and the out-of-limit parameters, location and control scheme are pushed. Construction is resumed after the parameters return to the warning threshold and the deviation rate is less than 5%.

[0019] Step 7 specifically involves collecting control effect data every 24 hours as a cycle, repeating steps 5 and 6 until the parameter deviation rate stabilizes within three consecutive cycles and no single parameter deviation rate remains greater than 10%, thus completing dynamic control.

[0020] The beneficial effects of this invention are: This invention forms a closed loop through five key stages: sensing, transmission, analysis, control, and optimization. A ring network is constructed by deploying fiber optic sensors on the frozen wall and wellbore to collect real-time monitoring data such as the frozen wall temperature field and wellbore strain, while simultaneously integrating on-site construction parameters and geological data. A dual-source dataset is built based on multiple real-world engineering cases and various numerical simulation scenarios. A gradient boosting tree algorithm is used to train a model that generates a table corresponding to "geological conditions - optimal construction parameter range." During construction, parameters can be matched and dynamically fine-tuned based on current geological conditions and monitoring values. When a warning threshold is exceeded, a tiered alarm is triggered, and a control plan is pushed out. Simultaneously, incremental learning continuously optimizes the model, transforming construction parameters from "experience-driven" to "data-driven," improving the safety and efficiency of vertical shaft construction in complex formations. This invention is suitable for vertical shaft construction in deep aquifers and weakly cemented formations under complex conditions. Attached Figure Description

[0021] Figure 1 This is a flowchart of the present invention. Detailed Implementation

[0022] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments.

[0023] A method for dynamic control of construction parameters in vertical shaft freezing based on fiber optic sensing (see [link]). Figure 1 This includes the following steps: Step 1: Densely distributed temperature and strain sensing optical cables are laid along the freezing holes in the frozen wall to monitor the temperature of the frozen wall and the strain of the rock mass. IP68-rated audible and visual alarms are also installed. Multiple fiber Bragg grating temperature sensors (monitoring concrete hydration heat), multiple fiber Bragg grating concrete strain gauges (monitoring post-pouring strain), multiple fiber Bragg grating rebar stress gauges (monitoring rebar stress), multiple inter-wall fiber Bragg grating pressure sensors (monitoring the contact pressure between the well wall and the frozen wall), and multiple fiber Bragg grating water pressure sensors (monitoring inter-wall seepage) are arranged in an array along the well wall's depth direction. IP68-rated audible and visual alarms are also installed. A fiber Bragg grating demodulator is placed in a computer room on the ground to connect the sensors via a self-connecting... The optical fibers are connected in series and then connected to the fiber optic demodulator on the ground via the main communication optical cable. The warning threshold for the densely distributed temperature sensing optical cable is set to -30℃ to -25℃ to ensure effective support of the freezing curtain. The warning threshold for the fiber optic temperature sensor is set to -10℃ to -8℃ to ensure the stability of the well wall. The warning threshold for the fiber optic concrete strain gauge is set to 2000με to 2400με to ensure the stability of the well wall concrete strength. The warning threshold for the fiber optic steel bar stress gauge is set to 30MPa to 32MPa to ensure the strength of the steel bar. The warning threshold for the inter-wall fiber optic pressure sensor is set to 10MPa to 12MPa to ensure the stability of the well wall. The warning threshold for the fiber optic water pressure sensor is set to 0 to 0.5MPa. The method for setting the early warning threshold of densely distributed strain sensing optical cables is as follows: Based on the design strength of the frozen wall, geological survey reports, and experience from similar projects, the failure intensity of different rock types is monitored through indoor tests using strain optical cables to provide the corresponding strain value at failure and to preliminarily determine the theoretical allowable strain range. Combined with numerical simulation analysis, the thermo-mechanical coupling deformation law of the frozen wall at different construction stages is calculated to clarify the strain development trend of key dangerous sections. In actual monitoring, real-time data needs to be obtained through densely distributed strain optical cables. If the strain of a certain rock layer continues to increase beyond the preset threshold or the local strain gradient suddenly increases, the system will trigger a corresponding early warning and prompt engineering intervention measures to be taken. Step 2: Simultaneously collect parameters of the frozen wall and well wall with the operating parameters of the on-site construction equipment using an industrial Ethernet network (100Mbps transmission rate). The collection frequency is set to 0.8 to 1.2 times per second. Parameters of the frozen wall and well wall include frozen wall temperature, frozen wall rock mass strain, well wall concrete temperature (used to understand the temperature field distribution and hydration heat of the concrete), well wall concrete strain, well wall reinforcement deformation stress, contact pressure between the well wall and the frozen wall, and seepage pressure between the two layers of well walls. Operating parameters of the on-site construction equipment include the freezer flow rate, the temperature of the brine inlet and outlet in the pipe (reflecting the cooling effect of the freezing system), the power of the chiller, the output flow rate of the grouting pump, and the temperature of the grouting liquid (related to the grouting quality of the well wall). This ensures that the frozen wall and well wall participate in the operation of the on-site construction equipment. The parameters are precisely matched in the time dimension, providing a high-timeliness and high-synchronization basic data support for subsequent data preprocessing, model analysis, and construction control. At the same time, alarm indicator lights and buzzers are installed on the control consoles of the freezer and grouting pump, respectively. The optical signal transmitted by the main communication optical cable is received through a fiber optic demodulator, and the wavelength change information generated in the optical signal is converted into an electrical signal. Then, the quantitative parameter values ​​of the frozen wall and well wall parameters are calculated. The acquisition time and the on-site construction equipment operating parameters are marked synchronously to form the raw data. Among them, the flow rate of the freezer is collected by the flow sensor of the freezing system, the temperature of the brine inlet and outlet of the pipe and the temperature of the grouting liquid are collected by temperature sensors, the power of the freezer is collected by a power meter, and the output flow rate of the grouting pump is collected by the flow sensor of the grouting system. Step 3: Mark raw data exceeding the warning threshold of each sensor as suspected anomalous data. For suspected anomalous data, retrieve synchronous data from at least three adjacent fiber optic sensors of the same type for neighborhood verification to eliminate misjudgment caused by single sensor failure (dense distributed sensing optical cables collect data at certain intervals; by averaging the data from two data points 20cm apart, misjudgment can be eliminated). Suspected anomalous data with an absolute temperature deviation exceeding 2℃, an absolute stress deviation exceeding 1MPa, or an absolute strain deviation exceeding 100με are determined as true anomalous values. For data that is suspected anomalous but does not exceed the true anomalous value, it is not determined as a true anomalous value for the time being, but it still needs to be included in the data preprocessing stage. This is because such data may have slight fluctuations or potential anomalous trends, and further analysis is needed in conjunction with subsequent time-series alignment correction and normalization processing steps to ensure that no potential problems that may affect the judgment of construction parameters are overlooked, and to provide a more comprehensive data foundation for subsequent model analysis. Replace the actual outliers with linear interpolation:

[0024] in, x t 1. x t+1 This represents normal data before and after the abnormal event. This is the time when true outliers were collected; it ensures the continuity of the data sequence and avoids the model's inability to capture time-series patterns due to missing values. The original data were normalized using the min-max standardization method:

[0025] in, x The original data, x min , x max This represents the theoretical extreme value of the data; it ensures that different parameters are comparable within the [0,1] interval, guaranteeing that the model learns evenly for each parameter. Due to the difference in sampling frequencies of fiber optic sensors (temperature sensor once per second, strain sensor once every 2 seconds), time-series alignment needs to be performed based on a unified system timestamp (accurate to milliseconds): the freezing wall temperature, well wall concrete temperature, brine inlet and outlet temperatures in the pipe, and grouting liquid temperature are downsampled; the frozen wall rock strain, well wall concrete strain, well wall steel reinforcement deformation stress, well wall-frozen wall contact pressure, seepage pressure between the two well walls, freezer flow rate, chiller power, and grouting pump output flow rate are generated using linear interpolation to create virtual data points per second; this ensures a one-to-one correspondence between the data at the same time point, meets the model's requirements for multi-parameter time-series correlation analysis, and ultimately forms a synchronized dataset in seconds; Step 4: During the preliminary exploration phase, geological parameters are collected, including lithology, water content, permeability, compressive strength, and internal friction angle. Lithology samples are collected using a geological drilling rig; water content is measured using a water content meter; permeability is measured using a permeameter; and compressive strength and internal friction angle are measured using a rock mechanics testing machine. Relevant data is stored in an engineering geological database. A monitoring center is set up on the ground, equipped with a data server (storage capacity ≥ 10TB), a visualization platform, audible and visual alarm devices, and an alarm control host. The data server is connected to the alarm control host and the visualization platform, and the alarm control host is connected to the audible and visual alarm devices. Data exchange is used to communicate with the system. The interface synchronously transmits on-site construction equipment operating parameters and geological parameters to the data server of the ground monitoring center via 5G / wired network. The data preprocessing module unifies the format of the construction equipment operating parameters and geological parameters (converting non-standard data output by different devices into CSV format) and aligns the timestamps (using the system's unified millisecond-level timestamp as a benchmark to associate on-site construction equipment operating parameters and geological parameters for the same time period), forming a related dataset containing "geological conditions-equipment operating status". This provides complete data support for subsequent machine learning models to match optimal construction parameters and achieve dynamic control. The visualization platform displays the temperature field distribution, stress cloud map (refresh frequency 1 time / second), and related datasets in real time. Step 5: Select at least 50 vertical shaft freezing construction projects from existing cases, covering strata types such as siltstone, fine-grained sandstone, medium-grained sandstone, coarse-grained sandstone, and sandy mudstone, as well as construction scenarios under a water content gradient of 5%-25%. Include successful cases of stable frozen wall formation and crack-free shaft walls, as well as typical cases of localized freezing wall melting and excessive concrete strain in the shaft wall. For each case, collect geological parameters (rock mass compressive strength, internal friction angle, formation permeability, formation water content accurate to 0.5%), frozen wall and shaft wall parameters (shaft wall concrete temperature collected every 30 minutes, accuracy ±0.1℃, shaft wall concrete strain resolution 0.1με, shaft wall steel reinforcement deformation stress accuracy ±0.1MPa), and on-site construction equipment operating parameters. All monitored values ​​must be within the thresholds specified in the "Coal Mine Shaft and Tunnel Engineering Construction Standard." All data should be timestamped and used as actual case data. Monitored values ​​include shaft wall concrete strain, shaft wall steel reinforcement deformation stress, shaft wall concrete temperature, and frozen wall radial temperature and strain. The specific steps for creating a 3D coupled geological-freezing-construction model using COMSOL are as follows: Import the shaft design drawings and create a 3D solid model in the geometry module of COMSOL. This model includes the shaft (diameter 5-12m, depth set according to the actual construction section), freezing holes (diameter 8-15m, spacing 0.8-1.2m, quantity consistent with the actual layout), and geological regions (horizontal range 5-10 times the shaft diameter, vertical range covering 50m above and below the construction section to ensure boundary effects do not affect calculations). The model is geometrically partitioned according to previously explored geological parameters (e.g., siltstone, fine-grained sandstone, medium-grained sandstone), and each region is assigned individual physical properties (density, specific heat capacity, thermal conductivity, permeability, elastic modulus). Three-field modules are added: heat conduction, porous media seepage, and solid mechanics. In the multiphysics coupling module, select heat-seepage coupling (temperature affects permeability), heat-structure coupling (temperature affects stress), and seepage-structure coupling (pore water pressure affects effective stress) to establish the interaction relationships between the three fields. Boundary conditions are set as follows: Thermal boundary: The inner wall of the frozen hole is set as a convective heat transfer boundary (the heat transfer coefficient is calculated based on the brine flow rate, and the brine inlet temperature is set to -30℃ to -20℃ to match the simulation conditions); the inner side of the well wall is set as a thermal insulation boundary (ignoring heat transfer with air); the far-field boundary of the formation is set as a constant temperature boundary (taking the original formation temperature, such as 15-25℃); Seepage boundary: The far-field boundary of the formation is set as a constant pressure boundary (taking the original pore water pressure, calculated based on depth); the frozen wall region is set as an impermeable boundary (the permeability of pore water approaches 0 after freezing); the well wall is set as a permeable boundary (considering inter-wall seepage, the permeability coefficient is assigned according to the design parameters of the well wall concrete); Mechanical boundary: far-field boundary of the formation. The boundary is set as a fixed constraint (no displacement), the top of the well wall is set as a free boundary (no constraint at the top during construction), and the contact area between the frozen wall and the well wall is set as a bound constraint (continuous displacement). According to the construction requirements of the "dual-source dataset", physical parameters are assigned to different strata areas (e.g., medium-grained sandstone: density 2600kg / m³, specific heat capacity 900J / (kg·℃), thermal conductivity 1.5W / (m·℃), permeability 0.5m / d; fine-grained sandstone: density 2700kg / m³, specific heat capacity 950J / (kg·℃), thermal conductivity 1.8W / (m·℃), permeability 5m / d). The parameter values ​​are derived from actual engineering cases and geological exploration reports. Set simulation conditions variables, such as formation water content (5%~25%, with a gradient of 2% each), brine temperature (-30℃~-20℃, with an interval of 1℃ each), and brine flow rate (20m³ / h~50m³ / h, with a gradient of 2m³ / h each). Save each condition separately as a calculation task. The three-dimensional geological-freezing-construction coupled model is based on the coupling of three fields: heat conduction, porous media seepage, and structural mechanics. It employs COMSOL's built-in coupling algorithm to calculate multiphysics interactions. The heat conduction algorithm is as follows:

[0026] in, Let C be the density of the medium, c be the specific heat capacity, T be the temperature, t be the time, k be the thermal conductivity, and Q be the intensity of the internal heat source. Based on Fourier's law, the temperature field distribution of the stratum and the frozen wall during the freezing process is calculated using a solid heat conduction module, taking into account the influence of brine circulation cooling (convective heat transfer) and concrete hydration heat (internal heat source) on the temperature field. Porous media flow algorithm:

[0027] in, For penetration rate, Pore ​​water pressure, For the density of water, Let t be the compressibility coefficient of water and t be time. The flow of formation water in porous media is simulated using Darcy's law module. Combining the characteristics of permeability change caused by pore water freezing during the formation of the frozen wall, a freezing water plugging correction factor is introduced (when the temperature is below 0℃, the permeability decreases exponentially with the increase of ice content). Structural mechanics algorithms:

[0028] in, The stress and strain distribution of the frozen wall and the well wall are calculated using the "Solid Mechanics" module. The thermal stress (thermal expansion / contraction) caused by temperature changes and the effective stress changes caused by pore water pressure changes are considered. The "Mohr-Coulomb criterion" is used to determine whether the frozen wall meets the strength requirements. The model was fully meshed using a free tetrahedral mesh. The mesh size was reduced to 0.2m-0.5m around the freezing hole and in the contact zone between the well wall and the freezing wall. A mesh size of 1m-2m was used for the formation far from the wellbore to balance computational accuracy and efficiency. A transient solver (with a time step of 1 hour to match the construction schedule) was used to solve the nonlinear equations using the Newton-Raphson iterative method to calculate the parameters of the freezing wall and wellbore under various working conditions. The convergence curve was monitored in real time during the calculation to ensure that the residual was below 1e-6 and to ensure the convergence of the temperature, pressure, and stress calculation results under the three-field coupling. The calculation results were extracted and organized into a dataset in the format of "parameter input-monitoring output" as simulation data. The model needs to accurately reproduce the wellbore diameter (5m-12m), the arrangement of freezing holes (circumference 8m-15m, hole spacing 0.8m-1.2m), and the engineering characteristics of the brine circulation path. Based on this, more than 1,000 working conditions are simulated: the formation parameter gradient is set manually (e.g., in coarse sandstone sections, the water content gradually increases from 5% to 25%, with each 2% being a gradient), and the construction parameter variables are set (e.g., brine temperature from -30℃ to -20℃, with each 1℃ being an interval, and flow rate from 20m³ / h to 50m³ / h, with each 2m³ / h being a gradient). The model outputs the frozen wall temperature field distribution (accuracy ±0.2℃), well wall strain cloud map (resolution 0.5με), and steel reinforcement stress value (accuracy ±0.2MPa) simulation monitoring data under the corresponding working conditions, and records the set construction parameters simultaneously to form a complete correspondence between "parameter input and monitoring output". Real-world case data and simulated data were integrated in a 6:4 ratio to form a dual-source dataset with at least 150,000 data entries. This ensured the authenticity of the project while supplementing the parameter coverage. During the modeling phase, machine learning algorithms were used to uncover the inherent patterns in the data, achieving a quantitative correlation between "geological conditions, construction parameters, and monitoring values." Features were selected from the dual-source dataset, using stratum type (encoded as a discrete variable), water content (a continuous variable), and compressive strength (a continuous variable) as input features. Saltwater temperature, saltwater flow rate, grouting flow rate, and grout temperature were extracted as key variables. The following parameters were considered: wellbore concrete temperature, wellbore concrete strain, wellbore reinforcement deformation stress, frozen wall temperature, and whether the frozen wall rock mass strain was complete. The output label is "within the warning threshold," with "yes" marked as 1 (corresponding to the candidate optimal parameter) and "no" marked as 0. The Gradient Boosting Tree (GBDT) algorithm is used for training. The dual-source dataset is divided into training and validation sets in an 8:2 ratio. Overfitting is controlled using 5-fold cross-validation. The tree depth and learning rate are optimized through grid search, and at least 1000 iterations are performed. When the classification accuracy on the validation set is not less than 92% (i.e., accurately determining whether a set of parameters can make all monitoring values ​​meet the standard), the K-means clustering algorithm is used to select the construction parameter combinations with an output label of 1 from the output results of the Gradient Boosting Tree model as "candidate optimal parameters." The clustering characteristics are defined as stratigraphic type and stratigraphic content. Based on the stratigraphic types (siltstone, fine-grained sandstone, medium-grained sandstone, coarse-grained sandstone, and sandy mudstone) and the 5%-25% water content gradient covered in the dual-source dataset, and considering the objective of "one set of optimal parameter intervals corresponding to the same geological conditions," the K value was determined using the elbow rule. The sum of squared errors within clusters was calculated for different K values. The K value corresponding to the point where the rate of decrease in the sum of squared errors within a cluster sharply decreased was selected. K "candidate optimal parameter" samples were randomly selected as initial cluster centers. The Euclidean distance between each remaining parameter sample and each cluster center was calculated, and the samples were assigned to the nearest cluster. The mean of each feature of all samples within each cluster was calculated and updated to the new cluster center. The process of "distance calculation - sample allocation - center update" is repeated until the center coordinate change is less than 0.1% for three consecutive iterations or the number of iterations reaches 100. For each group of parameters after clustering, the 95% confidence interval of each construction parameter is calculated (i.e., 95% of the parameter combinations within the interval can make the monitoring value stable within the warning threshold), forming a "geological conditions - optimal construction parameter interval" correspondence table. At the same time, the validity of the interval is verified by randomly selecting 10% of the parameter combinations within each interval and inputting them into the gradient boosting tree model to verify whether its output label is 1, ensuring that more than 95% of the parameter combinations within the interval can make the monitoring value meet the standard. Finally, the structured grouping results are output to determine the optimal construction parameters under different geological conditions. Among them, tree structure-related parameters include: tree depth (optimization range of 6-10 layers, avoiding overfitting due to excessive depth and underfitting due to excessive shallowness, adapting to the complex correlation patterns of at least 150,000 dual-source data) and minimum number of sample splits (optimization value of 20, ensuring that branch nodes have sufficient sample support, avoiding interference from a single sample with the correlation logic between construction parameters and geological conditions). Iterative update of relevant parameters: including learning rate (optimization range 0.05-0.1, controlling the weight contribution of each decision tree to the total predicted value of the model, preventing parameter oscillation caused by excessively rapid model updates during iteration), and number of iterations (fixed at 1000 times, ensuring that the model fully learns the parameter patterns under 12 strata and 5%-25% water cut gradients, and judging whether to stop iteration early by the condition that the accuracy fluctuation of the validation set is <0.5%). Regularization-related parameters include: subsampling ratio (optimization range 0.8-0.9, randomly selecting a portion of training samples to construct the decision tree in each iteration to reduce the risk of overfitting caused by data redundancy) and column sampling ratio (optimization range 0.7-0.8, randomly selecting a portion of input features to train a single tree to avoid over-reliance on key features such as formation water content and brine temperature). During algorithm training, key correlation patterns are automatically identified: for example, in sandy mudstone with a water content of 8%, when the brine temperature is between -26℃ and -24℃ and the flow rate is 30-35 m³ / h, the average temperature of the frozen wall can be stabilized between -28℃ and -25℃ (within the warning threshold), and the well wall concrete strain can be controlled between 2000-2200 με (within the warning threshold); however, in gravel layers with the same water content, due to the higher permeability of the rock mass, the brine temperature needs to be reduced by 2℃ (-28℃ to -26℃) and the flow rate increased by 5 m³ / h (35 m³ / h-40 m³ / h) to maintain the same monitoring effect; the application of the model in actual construction needs to realize the closed-loop logic of "input-matching-fine-tuning-output" to ensure the accuracy and operability of the recommendations. During construction, the system first automatically collects geological verification data for the current construction section (such as "siltstone + 12% water content" determined through advance drilling) and real-time monitoring values ​​transmitted by sensors (such as the current frozen wall temperature of -22℃ and well wall strain of 2350με). Then, it calls a preset "geological conditions - optimal construction parameter range" correspondence table to match the basic parameter range under these geological conditions (such as "saltwater temperature -27℃~-25℃, flow rate 35m³ / h-40m³ / h" for siltstone + 12% water content). Based on this, combined with real-time monitoring... The values ​​are dynamically fine-tuned: If the monitored value is close to the upper limit of the warning threshold (e.g., the wellbore strain of 2350 με is close to the warning threshold of 2400 με), a more conservative parameter is selected from the range (e.g., the brine temperature is set to -27℃ instead of -25℃, and the flow rate is set to 38 m³ / h instead of 35 m³ / h). Through parameter compensation, the monitored value is made to return to the middle of the warning threshold. The final output recommendation should include the specific adjustment value and the expected effect. For example, "The current water content of siltstone is 12%. It is recommended to adjust the brine temperature from -24℃ to -27℃ and increase the flow rate from 32 m³ / h to 38 m³ / h."It is expected that within 3 hours after the adjustment, the frozen wall temperature will drop to -28℃±1℃ (within the warning threshold), and the wellbore strain will gradually decrease to 2200με±50με (within the warning threshold). To avoid discrepancies between the model recommendations and actual field conditions, a dynamic optimization mechanism needs to be established to continuously improve adaptability. On the one hand, the system automatically collects newly generated data from the field every 24 hours, including the current geological verification results (formation water content and lithology correction results updated by advanced drilling), actual construction parameters (adjusted brine temperature / flow rate and grouting parameters), and adjusted monitoring values ​​(frozen wall temperature compliance rate and wellbore strain fluctuation value). Then, it filters the data according to "data integrity + validity" to remove abnormal data caused by sensor malfunctions (such as...). Invalid values ​​with temperature abrupt changes exceeding 5℃ are excluded. Complete data entries with the "geology-parameter-monitoring value" association are retained to form an incremental dataset (the data volume per cycle is approximately 1%-2% of the initial full dataset). The filtered incremental data undergoes the same preprocessing operations as the initial dataset, including repairing missing values ​​with linear interpolation, standardizing dimensions using min-max (ensuring new data feature values ​​are within the [0,1] range), and aligning the time series according to the system's unified millisecond-level timestamps. Simultaneously, the geological feature codes are verified (e.g., maintaining the initial coding rules of "sandstone = 1, siltstone = 2") and the construction parameter field names (e.g., the "saltwater inlet temperature" field is completely consistent with the initial model input fields) to ensure the incremental dataset... The data and model's original input feature dimensions and format are perfectly matched. A lightweight training mode of "partial parameter update + weight fine-tuning" is adopted. Instead of retraining all decision trees in the initial gradient boosting tree model, the key parameters of the model are optimized only based on the incremental dataset: First, the weights of the first 80% of the maturely trained decision trees in the initial model are fixed, and only the last 20% of decision trees that have not fully converged are updated. The residuals are calculated using the incremental dataset, and new weak learners (decision trees) are iteratively trained to supplement the model. Each time, the number of newly added trees is 5%-10% of the initial total number of trees (e.g., if the initial number is 1000 trees, 50-100 trees are added each time). Second, a learning rate decay strategy is used (the initial learning rate is reduced from 0.05-0.1 to 0). (0.01-0.02) Fine-tune the weights of the new trees to avoid excessive interference from the new data on the model's original patterns. Finally, calculate the overall classification accuracy of the model on the "new data + initial validation set" after incremental training. If the accuracy improves by ≥1%, keep the update; otherwise, backtrack and adjust the number of new trees and the learning rate, retrain, and apply the incrementally trained model to the original "geological conditions - optimal construction parameter interval" correspondence table. For the geological scenarios covered by the new data (such as gravel layers with a water content of 22%), recalculate the 95% confidence interval of the candidate optimal parameters in this scenario, and supplement or correct the original interval (such as the original interval of brine temperature -29℃~-27℃, which is adjusted to -30℃~-28℃ after validation with new data).Simultaneously, 3-5 sets of adjusted parameter intervals are selected for field trial construction. The compliance rate of frozen wall temperature and well wall strain within 3 cycles is monitored. The interval is considered valid only if the compliance rate is ≥90%. Finally, the optimized parameter intervals are stored synchronously with the incremental model for dynamic matching of subsequent construction parameters. If a set of parameters can maintain the monitored values ​​more stably in actual application (e.g., no fluctuation for 8 consecutive hours after adjustment), it is added to the optimal parameter interval for the corresponding geological conditions, and its effectiveness is verified by statistical tests (e.g., t-test). On the other hand, for extreme geological conditions (e.g., loose layers with water content >25% or fractured zones containing confined water), it is necessary to combine the geological data revealed on site and add special simulation conditions through COMSOL (e.g., densifying the spacing between freezing holes to 0.6m and increasing the brine flow rate to 1.2m / s) to generate suitable parameter intervals. These intervals are then compared and verified with the field trial construction data (e.g., the compliance rate of monitored values ​​must be ≥90% after 3 cycles of trial construction) to ensure that the model has no blind spots in its coverage of special scenarios. Through this cycle of "data accumulation - model iteration - on-site verification," the accuracy of the model's parameter recommendations can gradually increase from the initial 85% to over 95%, truly realizing the transformation of construction parameters from "experience-based judgment" to "data-driven."

[0029] For simulated working conditions of loose layers with a moisture content >25%, design parameters are formulated around "enhancing the sealing performance of the frozen wall and suppressing frost heave deformation of the loose layer," with specific value ranges as follows: 1. Freezing hole layout parameters: The spacing between freezing holes is set to 0.5m-0.7m (the spacing between conventional formations is 0.8m-1.2m, and the freezing overlap time is shortened by increasing the hole spacing). The diameter of the freezing hole is set to 1.2-1.5 times the wellbore diameter (e.g., when the wellbore diameter is 6m, the diameter is 7.2m-9m). The depth of a single hole covers 5m-8m above and below the loose layer to ensure that the freezing range completely covers the loose layer. 2. Brine circulation parameters: Brine temperature is set to -32℃~-28℃ (normal operating conditions are -30℃~-20℃, lowering the temperature accelerates water freezing), brine flow rate is set to 1m / s-1.3m / s (normal flow rate is 0.6m / s-0.9m / s, increasing the flow rate enhances heat exchange efficiency), and brine flow rate is set to 45m³ / h-55m³ / h (to match high flow rate requirements and ensure sufficient brine supply per orifice). 3. Formation pretreatment parameters: Simulating the pre-grouting reinforcement condition of loose layers, the grouting pressure was set to 1.8MPa-2.5MPa, the grout temperature was set to 22℃-28℃, and the grouting holes and freezing holes were arranged at intervals (1.0m-1.2m). By simulating the change in formation permeability after grouting (from the initial 5m / d-8m / d to 1m / d-2m / d), the effect of pretreatment on the freezing effect was verified. For simulated working conditions in a fractured zone containing confined water, design parameters are based on "resisting high water pressure and preventing leakage from the frozen wall," with specific numerical ranges as follows: 1. Frozen wall reinforcement parameters: The design thickness of the frozen wall is set at 3.5m-4.5m (2m-3m for conventional strata, thickening the frozen wall improves its impermeability). The deviation of the freezing holes is controlled within 0.1% (normally within 0.3%, reducing seepage channels caused by gaps between holes). Simultaneously, 1-2 rings of auxiliary freezing holes (hole spacing 0.6m-0.8m) are added at the upper and lower interfaces of the fractured zone, forming a double-layered freezing structure of "main freezing ring + auxiliary reinforcement ring"; 2. Brine and water pressure balance parameters: The brine temperature is set at -30℃~-26℃, and the brine flow velocity is set at 1.1m / s-1.4m / s (to ensure the frozen wall...). To quickly reach the design strength, the simulation of pressurized water pressure changes is performed simultaneously (set to ±0.5MPa based on the field-revealed value; for example, if the field-measured water pressure is 3MPa, the simulation range is 2.5MPa-3.5MPa) to verify the stress-strain response of the frozen wall under different water pressures. 3. Seepage prevention monitoring parameters: In the simulation, seepage pressure monitoring points between the walls are added (spacing 0.5m-1m), and the grouting and sealing conditions of the frozen holes are simulated simultaneously (grouting pressure 2MPa-3MPa, grout diffusion radius 0.8m-1.2m). The seepage pressure change curves under different grouting parameters are output to determine the optimal matching relationship between grouting and freezing parameters.

[0030] By fusing dual-source data—"real-world case studies + numerical simulations"—a precise correlation between "geological conditions, monitoring values, and construction parameters" is constructed. Ultimately, the optimal construction parameters that can keep monitoring values ​​stable within preset warning thresholds under different geological conditions are identified, providing directly applicable quantitative suggestions for on-site construction. This process requires the comprehensiveness of the data and the accuracy of the correlation. Parameter matching is achieved through algorithmic modeling, and continuous optimization is carried out in conjunction with on-site feedback to ensure that the model output is highly adapted to actual construction needs. Numerical simulation data is used to supplement extreme scenarios and parameter gradients not covered by real-world case studies. The construction of dual-source datasets is the foundation of model effectiveness and must take into account both data coverage and correlation integrity. Step 6: Compare the optimal construction parameters with the current construction parameters and calculate the parameter deviation rate. : ×100% Where m represents the current construction parameters and n represents the optimal construction parameters; When the deviation rate is 5%-10%, an audible and visual alarm is triggered. The audible and visual alarms on the frozen wall and well wall immediately emit a ≥85dB buzzer and flash red light. The alarm control host drives the audible and visual alarm device to light up yellow. The alarm indicator light and buzzer on the control panel of the freezer and grouting pump also light up yellow, indicating a slight over-limit. At the same time, combined with the real-time frozen wall and well wall parameters, the corresponding table of "geological conditions - optimal construction parameter range" is called to generate fine-tuning suggestions. When the deviation rate is greater than 10% or the current construction parameters exceed the warning threshold, a pop-up alarm is triggered. The alarm control host drives the audible and visual alarm device to light up red. The alarm indicator light and buzzer on the control panel of the freezer and grouting pump also light up red, indicating a serious over-limit and pushing the over-limit parameters, location and control scheme, realizing centralized collection and accurate push of alarm information. Construction resumes after the parameters return to the warning threshold and the deviation rate is less than 5%. When the suggestions do not meet the actual problem, manual adjustment is performed. Criteria for judging whether the recommended values ​​do not meet the actual situation: 1. Monitoring data exceeds limits or shows abnormal trends: If the frozen wall temperature, well wall concrete strain, and well wall steel reinforcement deformation stress collected in real time by fiber optic sensors still exceed the warning threshold after adjustment according to the recommended values ​​(e.g., frozen wall temperature > -25℃, well wall strain > 2400με), or the parameter change trend is abnormal (e.g., temperature rises by more than 5℃ in a short period of time, stress continues to increase without signs of stabilization), then the recommended values ​​are judged to be inconsistent with reality and need to be adjusted first; 2. Mismatch between geology and construction scenario: The geological conditions revealed on site do not match the geological data during model training. 3. Abnormal equipment operation status feedback: After implementing the recommended value adjustment, the freezer and grouting pump equipment exhibit abnormal operation (such as overload of the freezer power, pressure fluctuation of the grouting pump exceeding ±1MPa), or there are obvious problems such as water seepage around the well and abnormal noise from the frozen wall. After ruling out equipment failure factors, it is determined that the recommended value does not conform to reality.

[0031] When encountering discrepancies with actual conditions, manual adjustment should be performed according to the "emergency handling - parameter correction - verification and optimization" process. Specific operations are as follows: 1. Emergency handling and parameter suspension: If a safety risk exists (e.g., wellbore strain exceeds the warning threshold), immediately suspend the operation of equipment corresponding to the current recommended value (e.g., reduce chiller load, suspend grouting), prioritize activating audible and visual alarms to alert on-site personnel, and simultaneously retrieve fiber optic monitoring data and equipment operation logs from the past 10 minutes to identify the core cause of the recommended value mismatch; 2. Manual adjustment parameter determination: Adjust parameters based on actual on-site conditions—if mismatch is caused by geological differences, refer to historical successful cases of similar extreme geological conditions (e.g., for loose layers with a water content > 25%, the brine temperature can be lowered by 2℃-3℃, and the spacing between freezing holes can be increased to 0). 0.6m-0.7m); If equipment malfunctions, adjust according to the rated operating range of the equipment (e.g., the grouting pump flow rate should not exceed 80% of the rated value, and the brine flow rate should be controlled within 1m / s-1.2m / s). The parameter adjustment range should be controlled within ±10% of the threshold to avoid large fluctuations that could lead to new risks; 3. Verification and feedback after adjustment: After manual adjustment, continuously monitor for 1-2 data acquisition cycles (10-15 minutes per cycle). If the frozen wall temperature and well wall strain parameters return to the safe range and the trend is stable, and the equipment is operating normally, then the adjustment is confirmed to be effective; At the same time, the "geological conditions-corrected parameters-monitoring results" of this manual adjustment will be entered into the case library as a sample for subsequent incremental learning to optimize the accuracy of the model's parameter suggestions for similar scenarios and form a closed-loop optimization; Step 7: Collect data on the control effect every 24 hours as a cycle, repeat steps 5 and 6, and verify the optimization effect. Use the updated model to predict the optimal parameters for the next day and compare them with the actual monitoring compliance rate. If the compliance rate increases by ≥1%, keep the update; otherwise, backtrack and adjust the learning rate and the number of training trees. Short-term cutoff condition: When the parameter deviation rate is stable and meets the standard for three consecutive periods and no single parameter deviation rate is continuously greater than 10%, it indicates that the model is well adapted to the current geological and working conditions. Incremental updates are suspended, and routine monitoring is switched to complete dynamic control.

[0032] Long-term termination condition: When the current shaft section is completed after the vertical shaft freezing construction (such as excavation to the design depth), or when a major geological change occurs on site (such as encountering an unexplored large confined aquifer) that makes the original model completely unsuitable, the current model update will be stopped. After the new construction section or geological stability, the model will be retrained based on the new dual-source dataset (real-world case + COMSOL special simulation).

[0033] Example 1: A method for dynamic control of construction parameters in vertical shaft freezing based on fiber optic sensing, taking a coal mine vertical shaft freezing construction project with a depth of 500m and a coarse sandstone aquifer thickness of 100m as an example: A densely distributed temperature sensing fiber optic cable and a densely distributed strain sensing fiber optic cable are lowered along the freezing hole. Multiple sets of fiber optic grating sensors are arranged in an array along the shaft depth direction on the shaft wall, including fiber optic grating temperature sensors, fiber optic grating concrete strain gauges, fiber optic grating steel stress gauges, inter-wall fiber optic grating pressure sensors, and fiber optic grating water pressure gauges. The sensor detected that the frozen wall temperature was -20℃ in real time, while the set warning threshold was -25℃ to -30℃. The well wall concrete strain was 2300με, while the warning threshold was 2000με to 2400με. Based on the coarse sandstone freezing data simulated by the machine learning model, it was recommended to lower the brine temperature to -26℃ and increase the brine flow rate to 32m³ / h. At this time, the staff would receive an alarm message and the optimal construction parameters for consideration. They would then use automatic or manual adjustment to lower the frozen wall temperature to the warning threshold range.

[0034] Example 2: A method for dynamic control of construction parameters in vertical shaft freezing based on fiber optic sensing includes the following steps: Step 1: Lay sensor optical cables on the frozen wall and set early warning thresholds; pre-embed sensors in the well wall and set early warning thresholds. Step 2: Collect and analyze the parameters of the frozen wall and wellbore in real time to generate raw data; Step 3: Preprocess and extract features from the raw data; Step 4: Integrate the on-site construction parameters and transmit them to the monitoring center; Step 5: Select at least 50 vertical shaft freezing construction projects, covering strata types such as siltstone, fine-grained sandstone, medium-grained sandstone, coarse-grained sandstone, and sandy mudstone, as well as construction scenarios under a water content gradient of 5%-25%. Include successful cases of stable frozen wall formation and crack-free shaft walls, as well as typical cases of localized freezing wall melting and excessive concrete strain in the shaft wall. For each case, collect geological parameters, frozen wall and shaft wall parameters, and on-site construction equipment operating parameters. All monitored values ​​must be within the thresholds specified in the "Coal Mine Shaft and Tunnel Engineering Construction Standard." All data should be timestamped and used as actual case data. Monitored values ​​include shaft wall concrete strain, shaft wall steel reinforcement deformation stress, shaft wall concrete temperature, and frozen wall radial temperature and strain. A three-dimensional geological-freezing-construction coupled model was established using COMSOL. The model was fully meshed using a free tetrahedral mesh. The mesh size was reduced to 0.2m-0.5m around the freezing hole and in the contact zone between the well wall and the freezing wall. A mesh size of 1m-2m was used for the formation far from the wellbore. The nonlinear equations were solved using the Newton-Raphson iteration method with a transient solver to calculate the parameters of the freezing wall and well wall under various working conditions. The convergence curve was monitored in real time during the calculation to ensure that the residual was below 1e-6. The calculation results were extracted and organized into a dataset in the format of "parameter input-monitoring output" as simulation data. Actual case data and simulation data were integrated in a 6:4 ratio to form a dual-source dataset with at least 150,000 data entries. Features were selected from the dual-source dataset, with formation type, water content, and compressive strength as input features, and brine temperature, brine flow rate, grouting flow rate, and grout temperature as key variables. The output label was "whether the well wall concrete temperature, well wall concrete strain, well wall steel reinforcement deformation stress, frozen wall temperature, and frozen wall rock mass strain are all within the warning threshold". "Yes" was marked as 1, and "no" was marked as 0. The gradient boosting tree algorithm was used for training. The dual-source dataset was divided into training and validation sets in an 8:2 ratio. Five-fold cross-validation was used to control overfitting. The tree depth and learning rate were optimized using grid search. At least 1000 iterations were performed. When the classification accuracy on the validation set was no less than 92%, the K-means clustering algorithm was used to select the construction parameter combinations with an output label of 1 from the gradient boosting tree model output as "candidate optimal parameters." The clustering features were defined as formation type, formation water content, brine temperature, brine flow rate, and grouting temperature. Based on the formation types (siltstone, fine-grained sandstone, medium-grained sandstone, coarse-grained sandstone, and sandy mudstone) and the 5%-25% water content gradient covered in the dual-source dataset, and considering the objective of "one set of optimal parameter intervals corresponding to the same geological conditions," the K value was determined using the elbow rule. The sum of squared errors within clusters under different K values ​​was calculated, and the rate of decrease of the sum of squared errors within clusters was selected. The K value corresponding to the point of reduction is used to randomly select K "candidate optimal parameter" samples as initial cluster centers. The Euclidean distance between each remaining parameter sample and each cluster center is calculated, and the samples are assigned to the nearest cluster. The mean of each feature of all samples in each cluster is calculated and updated as new cluster centers. The "distance calculation-sample assignment-center update" steps are repeated until the center coordinate change is less than 0.1% for three consecutive iterations or the number of iterations reaches 100. For each group of parameters after clustering, the 95% confidence interval of each construction parameter is calculated to form a "geological condition-optimal construction parameter interval" correspondence table. At the same time, the validity of the interval is verified. 10% of the parameter combinations in each interval are randomly selected and input into the gradient boosting tree model to verify whether its output label is 1. This ensures that more than 95% of the parameter combinations in the interval can make the monitoring value meet the standard. Finally, the structured grouping results are output to determine the optimal construction parameters under different geological conditions. Step 6: Compare the optimal construction parameters with the current construction parameters and calculate the parameter deviation rate. : ×100% Where m represents the current construction parameters and n represents the optimal construction parameters; when the deviation rate is 5%-10%, an audible and visual alarm is triggered. At the same time, combined with the real-time frozen wall and well wall parameters, the corresponding table of "geological conditions-optimal construction parameter range" is called to generate fine-tuning suggestions. When the deviation rate is greater than 10% or the current construction parameters exceed the warning threshold, a pop-up alarm is triggered, and the out-of-limit parameters, location and control plan are pushed. Construction resumes after the parameters return to the warning threshold and the deviation rate is less than 5%. Step 7: Collect control effect data every 24 hours as a cycle, repeat steps 5 and 6 until the parameter deviation rate is stable and meets the standard for three consecutive cycles and no single parameter deviation rate is continuously greater than 10%, thus completing dynamic control.

[0035] Example 3: A method for dynamic control of construction parameters in vertical shaft freezing based on fiber optic sensing includes the following steps: Step 1: Deploy densely distributed temperature sensing optical cables and densely distributed strain sensing optical cables along the freezing holes on the frozen wall to monitor the temperature of the frozen wall and the strain of the rock mass. Simultaneously, install audible and visual alarms. Arrange multiple fiber Bragg grating temperature sensors, multiple fiber Bragg grating concrete strain gauges, multiple fiber Bragg grating rebar stress gauges, multiple inter-wall fiber Bragg grating pressure sensors, and multiple fiber Bragg grating water pressure gauges in an array along the depth direction of the well wall. Also, install audible and visual alarms. Place a fiber Bragg grating demodulator in a computer room on the ground to connect all the sensors in series via their built-in optical fibers. Then, it is connected to the ground fiber optic demodulator via the main communication optical cable. The warning threshold for the densely distributed temperature sensing optical cable is set to -30℃ to -25℃, the warning threshold for the fiber optic temperature sensor is set to -10℃ to -8℃, the warning threshold for the fiber optic concrete strain gauge is set to 2000με to 2400με, the warning threshold for the fiber optic rebar stress gauge is set to 30MPa to 32MPa, the warning threshold for the wall-mounted fiber optic pressure sensor is set to 10MPa to 12MPa, and the warning threshold for the fiber optic water pressure sensor is set to 0 to 0.5MPa. Step 2: Use industrial Ethernet to synchronously collect parameters of the frozen wall and well wall with the operating parameters of the on-site construction equipment. The collection frequency is set to 0.8 times / second to 1.2 times / second. The parameters of the frozen wall and well wall include the temperature of the frozen wall, the strain of the frozen wall rock mass, the temperature of the well wall concrete, the strain of the well wall concrete, the deformation stress of the well wall reinforcement, the contact pressure between the well wall and the frozen wall, and the seepage pressure between the two layers of well walls. The operating parameters of the on-site construction equipment include the flow rate of the freezer, the temperature of the brine inlet and outlet in the pipe, the power of the chiller, the output flow rate of the grouting pump, and the temperature of the grouting liquid. At the same time, alarm indicator lights and buzzers are installed on the control consoles of the freezer and the grouting pump, respectively. The optical signal transmitted by the main communication optical cable is received through the fiber optic demodulator, and the wavelength change information generated in the optical signal is converted into an electrical signal. Then, the quantitative parameter values ​​of the frozen wall and well wall are calculated. The collection time and the operating parameters of the on-site construction equipment are marked synchronously to form the raw data. Step 3: Mark the raw data that exceeds the warning threshold of each sensor as suspected abnormal data. For suspected abnormal data, retrieve the synchronous data of at least 3 adjacent fiber optic sensors of the same type for neighborhood verification. Suspected abnormal data with an absolute value of temperature deviation exceeding 2℃, suspected abnormal data with an absolute value of stress deviation exceeding 1MPa, and suspected abnormal data with an absolute value of strain deviation exceeding 100με are determined as real abnormal data. Replace the actual outliers with linear interpolation:

[0036] in, xt 1. x t+1 This represents normal data before and after the abnormal event. This refers to the time when actual outlier values ​​were collected. The original data were normalized using the min-max standardization method:

[0037] in, x The original data, x min , x max The theoretical extreme value of this data is used; the temperature of the frozen wall, the temperature of the well wall concrete, the temperature of the brine inlet and outlet in the pipe, and the temperature of the grouting liquid are downsampled; the frozen wall rock mass strain, the well wall concrete strain, the deformation stress of the well wall steel reinforcement, the contact pressure between the well wall and the frozen wall, the seepage pressure between the two layers of well walls, the flow rate of the freezer, the power of the freezer, and the output flow rate of the grouting pump are generated by linear interpolation to produce virtual data points per second. Step 4: Integrate the on-site construction parameters and transmit them to the monitoring center; Step 5: Select at least 50 vertical shaft freezing construction projects, covering strata types such as siltstone, fine-grained sandstone, medium-grained sandstone, coarse-grained sandstone, and sandy mudstone, as well as construction scenarios under a water content gradient of 5%-25%. Include successful cases of stable frozen wall formation and crack-free shaft walls, as well as typical cases of localized freezing wall melting and excessive concrete strain in the shaft wall. For each case, collect geological parameters, frozen wall and shaft wall parameters, and on-site construction equipment operating parameters. All monitored values ​​must be within the thresholds specified in the "Coal Mine Shaft and Tunnel Engineering Construction Standard." All data should be timestamped and used as actual case data. Monitored values ​​include shaft wall concrete strain, shaft wall steel reinforcement deformation stress, shaft wall concrete temperature, and frozen wall radial temperature and strain. A three-dimensional geological-freezing-construction coupled model was established using COMSOL. The model was fully meshed using a free tetrahedral mesh. The mesh size was reduced to 0.2m-0.5m around the freezing hole and in the contact zone between the well wall and the freezing wall. A mesh size of 1m-2m was used for the formation far from the wellbore. The nonlinear equations were solved using the Newton-Raphson iteration method with a transient solver to calculate the parameters of the freezing wall and well wall under various working conditions. The convergence curve was monitored in real time during the calculation to ensure that the residual was below 1e-6. The calculation results were extracted and organized into a dataset in the format of "parameter input-monitoring output" as simulation data. Real-world case data and simulated data were integrated in a 6:4 ratio to form a dual-source dataset with at least 150,000 data entries. Features were selected from the dual-source dataset, using formation type, water content, and compressive strength as input features, and extracting brine temperature, brine flow rate, grouting flow rate, and grout temperature as key variables. The output label was "whether the well wall concrete temperature, well wall concrete strain, well wall steel reinforcement deformation stress, frozen wall temperature, and frozen wall rock mass strain are all within the warning threshold," with "yes" marked as 1 and "no" marked as 0. The gradient boosting tree algorithm was used for training, and the dual-source dataset was divided into training and validation sets in an 8:2 ratio. Overfitting was controlled by 5-fold cross-validation, and the tree depth and learning rate were optimized by grid search. At least 1,000 iterations of training were performed. When the classification accuracy on the validation set was not less than 92%, the K-means clustering algorithm was used to group the output "candidate optimal parameters" to determine the optimal construction parameters under different geological conditions. Step 6: Compare the optimal construction parameters with the current construction parameters and calculate the parameter deviation rate. : ×100% Where m represents the current construction parameters and n represents the optimal construction parameters; when the deviation rate is 5%-10%, an audible and visual alarm is triggered. At the same time, combined with the real-time frozen wall and well wall parameters, the corresponding table of "geological conditions-optimal construction parameter range" is called to generate fine-tuning suggestions. When the deviation rate is greater than 10% or the current construction parameters exceed the warning threshold, a pop-up alarm is triggered, and the out-of-limit parameters, location and control plan are pushed. Construction resumes after the parameters return to the warning threshold and the deviation rate is less than 5%. Step 7: Dynamically optimize the construction process.

[0038] Example 4: A method for dynamic control of construction parameters in vertical shaft freezing based on fiber optic sensing includes the following steps: Step 1: Deploy densely distributed temperature sensing optical cables and densely distributed strain sensing optical cables along the freezing holes on the frozen wall to monitor the temperature of the frozen wall and the strain of the rock mass. Simultaneously, install audible and visual alarms. Arrange multiple fiber Bragg grating temperature sensors, multiple fiber Bragg grating concrete strain gauges, multiple fiber Bragg grating rebar stress gauges, multiple inter-wall fiber Bragg grating pressure sensors, and multiple fiber Bragg grating water pressure gauges in an array along the depth direction of the well wall. Also, install audible and visual alarms. Place a fiber Bragg grating demodulator in a computer room on the ground to connect all the sensors in series via their built-in optical fibers. Then, it is connected to the ground fiber optic demodulator via the main communication optical cable. The warning threshold for the densely distributed temperature sensing optical cable is set to -30℃ to -25℃, the warning threshold for the fiber optic temperature sensor is set to -10℃ to -8℃, the warning threshold for the fiber optic concrete strain gauge is set to 2000με to 2400με, the warning threshold for the fiber optic rebar stress gauge is set to 30MPa to 32MPa, the warning threshold for the wall-mounted fiber optic pressure sensor is set to 10MPa to 12MPa, and the warning threshold for the fiber optic water pressure sensor is set to 0 to 0.5MPa. Step 2: Use industrial Ethernet to synchronously collect parameters of the frozen wall and well wall with the operating parameters of the on-site construction equipment. The collection frequency is set to 0.8 times / second to 1.2 times / second. The parameters of the frozen wall and well wall include the temperature of the frozen wall, the strain of the frozen wall rock mass, the temperature of the well wall concrete, the strain of the well wall concrete, the deformation stress of the well wall reinforcement, the contact pressure between the well wall and the frozen wall, and the seepage pressure between the two layers of well walls. The operating parameters of the on-site construction equipment include the flow rate of the freezer, the temperature of the brine inlet and outlet in the pipe, the power of the chiller, the output flow rate of the grouting pump, and the temperature of the grouting liquid. At the same time, alarm indicator lights and buzzers are installed on the control consoles of the freezer and the grouting pump, respectively. The optical signal transmitted by the main communication optical cable is received through the fiber optic demodulator, and the wavelength change information generated in the optical signal is converted into an electrical signal. Then, the quantitative parameter values ​​of the frozen wall and well wall are calculated. The collection time and the operating parameters of the on-site construction equipment are marked synchronously to form the raw data. Step 3: Preprocess and extract features from the raw data; Step 4: Integrate the on-site construction parameters and transmit them to the monitoring center; Step 5: Select at least 50 vertical shaft freezing construction projects, covering strata types such as siltstone, fine-grained sandstone, medium-grained sandstone, coarse-grained sandstone, and sandy mudstone, as well as construction scenarios under a water content gradient of 5%-25%. Include successful cases of stable frozen wall formation and crack-free shaft walls, as well as typical cases of localized freezing wall melting and excessive concrete strain in the shaft wall. For each case, collect geological parameters, frozen wall and shaft wall parameters, and on-site construction equipment operating parameters. All monitored values ​​must be within the thresholds specified in the "Coal Mine Shaft and Tunnel Engineering Construction Standard." All data should be timestamped and used as actual case data. Monitored values ​​include shaft wall concrete strain, shaft wall steel reinforcement deformation stress, shaft wall concrete temperature, and frozen wall radial temperature and strain. A three-dimensional geological-freezing-construction coupled model was established using COMSOL. This model centers on the coupling of three fields: heat conduction, porous media seepage, and structural mechanics. COMSOL's built-in coupling algorithm was employed to calculate multiphysics interactions. The heat conduction algorithm is as follows:

[0039] in, Where is the density of the medium, c is the specific heat capacity, T is the temperature, t is the time, k is the thermal conductivity, and Q is the intensity of the internal heat source; Porous media flow algorithm:

[0040] in, For penetration rate, Pore ​​water pressure, For the density of water, Let t be the compressibility coefficient of water, and t be time. Structural mechanics algorithms:

[0041] in, The model was fully meshed using a free tetrahedral mesh. The mesh size was reduced to 0.2m-0.5m around the freezing hole and in the contact zone between the well wall and the freezing wall. A mesh size of 1m-2m was used for the formation far from the wellbore. The nonlinear equations were solved using the Newton-Raphson iteration method with a transient solver to calculate the parameters of the freezing wall and wellbore under various working conditions. The convergence curve was monitored in real time during the calculation to ensure that the residual was below 1e-6. The calculation results were extracted and organized into a dataset in the format of "parameter input-monitoring output" as simulation data. Actual case data and simulation data were integrated in a 6:4 ratio to form a dual-source dataset with at least 150,000 data entries. Features were selected from the dual-source dataset, with formation type, water content, and compressive strength as input features, and brine temperature, brine flow rate, grouting flow rate, and grout temperature as key variables. The output label was "whether the well wall concrete temperature, well wall concrete strain, well wall steel reinforcement deformation stress, frozen wall temperature, and frozen wall rock mass strain are all within the warning threshold". "Yes" was marked as 1, and "no" was marked as 0. The gradient boosting tree algorithm was used for training. The dual-source dataset was divided into training and validation sets in an 8:2 ratio. Five-fold cross-validation was used to control overfitting. The tree depth and learning rate were optimized using grid search. At least 1000 iterations were performed. When the classification accuracy on the validation set was no less than 92%, the K-means clustering algorithm was used to select the construction parameter combinations with an output label of 1 from the gradient boosting tree model output as "candidate optimal parameters." The clustering features were defined as formation type, formation water content, brine temperature, brine flow rate, and grouting temperature. Based on the formation types (siltstone, fine-grained sandstone, medium-grained sandstone, coarse-grained sandstone, and sandy mudstone) and the 5%-25% water content gradient covered in the dual-source dataset, and considering the objective of "one set of optimal parameter intervals corresponding to the same geological conditions," the K value was determined using the elbow rule. The sum of squared errors within clusters under different K values ​​was calculated, and the rate of decrease of the sum of squared errors within clusters was selected. The K value corresponding to the point of reduction is used to randomly select K "candidate optimal parameter" samples as initial cluster centers. The Euclidean distance between each remaining parameter sample and each cluster center is calculated, and the samples are assigned to the nearest cluster. The mean of each feature of all samples in each cluster is calculated and updated as new cluster centers. The "distance calculation-sample assignment-center update" steps are repeated until the center coordinate change is less than 0.1% for three consecutive iterations or the number of iterations reaches 100. For each group of parameters after clustering, the 95% confidence interval of each construction parameter is calculated to form a "geological condition-optimal construction parameter interval" correspondence table. At the same time, the validity of the interval is verified. 10% of the parameter combinations in each interval are randomly selected and input into the gradient boosting tree model to verify whether its output label is 1. This ensures that more than 95% of the parameter combinations in the interval can make the monitoring value meet the standard. Finally, the structured grouping results are output to determine the optimal construction parameters under different geological conditions. Step 6: Compare the construction parameters monitored by the sensors with the optimal construction parameters to determine whether adjustments are needed; Step 7: Dynamically optimize the construction process.

[0042] Example 5: A method for dynamic control of construction parameters in vertical shaft freezing based on fiber optic sensing includes the following steps: Step 1: Deploy densely distributed temperature sensing optical cables and densely distributed strain sensing optical cables along the freezing holes on the frozen wall to monitor the temperature of the frozen wall and the strain of the rock mass. Simultaneously, install audible and visual alarms. Arrange multiple fiber Bragg grating temperature sensors, multiple fiber Bragg grating concrete strain gauges, multiple fiber Bragg grating rebar stress gauges, multiple inter-wall fiber Bragg grating pressure sensors, and multiple fiber Bragg grating water pressure gauges in an array along the depth direction of the well wall. Also, install audible and visual alarms. Place a fiber Bragg grating demodulator in a computer room on the ground to connect all the sensors in series via their built-in optical fibers. Then, it is connected to the ground fiber optic demodulator via the main communication optical cable. The warning threshold for the densely distributed temperature sensing optical cable is set to -30℃ to -25℃, the warning threshold for the fiber optic temperature sensor is set to -10℃ to -8℃, the warning threshold for the fiber optic concrete strain gauge is set to 2000με to 2400με, the warning threshold for the fiber optic rebar stress gauge is set to 30MPa to 32MPa, the warning threshold for the wall-mounted fiber optic pressure sensor is set to 10MPa to 12MPa, and the warning threshold for the fiber optic water pressure sensor is set to 0 to 0.5MPa. Step 2: Use industrial Ethernet to synchronously collect parameters of the frozen wall and well wall with the operating parameters of the on-site construction equipment. The collection frequency is set to 0.8 times / second to 1.2 times / second. The parameters of the frozen wall and well wall include the temperature of the frozen wall, the strain of the frozen wall rock mass, the temperature of the well wall concrete, the strain of the well wall concrete, the deformation stress of the well wall reinforcement, the contact pressure between the well wall and the frozen wall, and the seepage pressure between the two layers of well walls. The operating parameters of the on-site construction equipment include the flow rate of the freezer, the temperature of the brine inlet and outlet in the pipe, the power of the chiller, the output flow rate of the grouting pump, and the temperature of the grouting liquid. At the same time, alarm indicator lights and buzzers are installed on the control consoles of the freezer and the grouting pump, respectively. The optical signal transmitted by the main communication optical cable is received through the fiber optic demodulator, and the wavelength change information generated in the optical signal is converted into an electrical signal. Then, the quantitative parameter values ​​of the frozen wall and well wall are calculated. The collection time and the operating parameters of the on-site construction equipment are marked synchronously to form the raw data. Step 3: Mark the raw data that exceeds the warning threshold of each sensor as suspected abnormal data. For suspected abnormal data, retrieve the synchronous data of at least 3 adjacent fiber optic sensors of the same type for neighborhood verification. Suspected abnormal data with an absolute value of temperature deviation exceeding 2℃, suspected abnormal data with an absolute value of stress deviation exceeding 1MPa, and suspected abnormal data with an absolute value of strain deviation exceeding 100με are determined as real abnormal data. Replace the actual outliers with linear interpolation:

[0043] in, x t 1. xt+1 This represents normal data before and after the abnormal event. The data represents the actual outlier collection time; the raw data was normalized using the min-max standardization method.

[0044] in, x The original data, x min , x max The theoretical extreme value of this data is used; the temperature of the frozen wall, the temperature of the well wall concrete, the temperature of the brine inlet and outlet in the pipe, and the temperature of the grouting liquid are downsampled; the frozen wall rock mass strain, the well wall concrete strain, the deformation stress of the well wall steel reinforcement, the contact pressure between the well wall and the frozen wall, the seepage pressure between the two layers of well walls, the flow rate of the freezer, the power of the freezer, and the output flow rate of the grouting pump are generated by linear interpolation to produce virtual data points per second. Step 4: Collect geological parameters during the preliminary exploration phase. These parameters include lithology, water content, permeability, compressive strength, and internal friction angle. Set up a monitoring center on the ground, equipped with a data server, visualization platform, audible and visual alarm devices, and alarm control host. The data server is connected to the alarm control host and visualization platform, and the alarm control host is connected to the audible and visual alarm devices. Through the data interaction interface, the operating parameters of the construction equipment and geological parameters are synchronously transmitted to the data server in the ground monitoring center. The data preprocessing module unifies the format and aligns the timestamps of the operating parameters of the construction equipment and geological parameters to form a related dataset containing "geological conditions - equipment operating status". The visualization platform displays the temperature field distribution, stress cloud map, and related dataset in real time. Step 5: Select at least 50 vertical shaft freezing construction projects, covering strata types such as siltstone, fine-grained sandstone, medium-grained sandstone, coarse-grained sandstone, and sandy mudstone, as well as construction scenarios under a water content gradient of 5%-25%. Include successful cases of stable frozen wall formation and crack-free shaft walls, as well as typical cases of localized freezing wall melting and excessive concrete strain in the shaft wall. For each case, collect geological parameters, frozen wall and shaft wall parameters, and on-site construction equipment operating parameters. All monitored values ​​must be within the thresholds specified in the "Coal Mine Shaft and Tunnel Engineering Construction Standard." All data should be timestamped and used as actual case data. Monitored values ​​include shaft wall concrete strain, shaft wall steel reinforcement deformation stress, shaft wall concrete temperature, and frozen wall radial temperature and strain. A three-dimensional geological-freezing-construction coupled model was established using COMSOL. The model was fully meshed using a free tetrahedral mesh. The mesh size was reduced to 0.2m-0.5m around the freezing hole and in the contact zone between the well wall and the freezing wall. A mesh size of 1m-2m was used for the formation far from the wellbore. The nonlinear equations were solved using the Newton-Raphson iteration method with a transient solver to calculate the parameters of the freezing wall and well wall under various working conditions. The convergence curve was monitored in real time during the calculation to ensure that the residual was below 1e-6. The calculation results were extracted and organized into a dataset in the format of "parameter input-monitoring output" as simulation data. Real-world case data and simulated data were integrated in a 6:4 ratio to form a dual-source dataset with at least 150,000 data entries. Features were selected from the dual-source dataset, using formation type, water content, and compressive strength as input features, and extracting brine temperature, brine flow rate, grouting flow rate, and grout temperature as key variables. The output label was "whether the well wall concrete temperature, well wall concrete strain, well wall steel reinforcement deformation stress, frozen wall temperature, and frozen wall rock mass strain are all within the warning threshold," with "yes" marked as 1 and "no" marked as 0. The gradient boosting tree algorithm was used for training, and the dual-source dataset was divided into training and validation sets in an 8:2 ratio. Overfitting was controlled by 5-fold cross-validation, and the tree depth and learning rate were optimized by grid search. At least 1,000 iterations of training were performed. When the classification accuracy on the validation set was not less than 92%, the K-means clustering algorithm was used to group the output "candidate optimal parameters" to determine the optimal construction parameters under different geological conditions. Step 6: Compare the optimal construction parameters with the current construction parameters and calculate the parameter deviation rate. : ×100% Where m represents the current construction parameters and n represents the optimal construction parameters; when the deviation rate is 5%-10%, an audible and visual alarm is triggered. At the same time, combined with the real-time frozen wall and well wall parameters, the corresponding table of "geological conditions-optimal construction parameter range" is called to generate fine-tuning suggestions. When the deviation rate is greater than 10% or the current construction parameters exceed the warning threshold, a pop-up alarm is triggered, and the out-of-limit parameters, location and control plan are pushed. Construction resumes after the parameters return to the warning threshold and the deviation rate is less than 5%. Step 7: Dynamically optimize the construction process.

[0045] Example 6: A method for dynamic control of construction parameters in vertical shaft freezing based on fiber optic sensing includes the following steps: Step 1: Lay sensor optical cables on the frozen wall and set early warning thresholds; pre-embed sensors in the well wall and set early warning thresholds. Step 2: Collect and analyze the parameters of the frozen wall and wellbore in real time to generate raw data; Step 3: Preprocess and extract features from the raw data; Step 4: Integrate the on-site construction parameters and transmit them to the monitoring center; Step 5: Select at least 50 vertical shaft freezing construction projects, covering strata types such as siltstone, fine-grained sandstone, medium-grained sandstone, coarse-grained sandstone, and sandy mudstone, as well as construction scenarios under a water content gradient of 5%-25%. Include successful cases of stable frozen wall formation and crack-free shaft walls, as well as typical cases of localized freezing wall melting and excessive concrete strain in the shaft wall. For each case, collect geological parameters, frozen wall and shaft wall parameters, and on-site construction equipment operating parameters. All monitored values ​​must be within the thresholds specified in the "Coal Mine Shaft and Tunnel Engineering Construction Standard." All data should be timestamped and used as actual case data. Monitored values ​​include shaft wall concrete strain, shaft wall steel reinforcement deformation stress, shaft wall concrete temperature, and frozen wall radial temperature and strain. A three-dimensional geological-freezing-construction coupled model was established using COMSOL. The model was fully meshed using a free tetrahedral mesh. The mesh size was reduced to 0.2m-0.5m around the freezing hole and in the contact zone between the well wall and the freezing wall. A mesh size of 1m-2m was used for the formation far from the wellbore. The nonlinear equations were solved using the Newton-Raphson iteration method with a transient solver to calculate the parameters of the freezing wall and well wall under various working conditions. The convergence curve was monitored in real time during the calculation to ensure that the residual was below 1e-6. The calculation results were extracted and organized into a dataset in the format of "parameter input-monitoring output" as simulation data. Real-world case data and simulated data were integrated in a 6:4 ratio to form a dual-source dataset with at least 150,000 data entries. Features were selected from the dual-source dataset, using formation type, water content, and compressive strength as input features, and extracting brine temperature, brine flow rate, grouting flow rate, and grout temperature as key variables. The output label was "whether the well wall concrete temperature, well wall concrete strain, well wall steel reinforcement deformation stress, frozen wall temperature, and frozen wall rock mass strain are all within the warning threshold," with "yes" marked as 1 and "no" marked as 0. The gradient boosting tree algorithm was used for training, and the dual-source dataset was divided into training and validation sets in an 8:2 ratio. Overfitting was controlled by 5-fold cross-validation, and the tree depth and learning rate were optimized by grid search. At least 1,000 iterations of training were performed. When the classification accuracy on the validation set was not less than 92%, the K-means clustering algorithm was used to group the output "candidate optimal parameters" to determine the optimal construction parameters under different geological conditions. Step 6: Compare the construction parameters monitored by the sensors with the optimal construction parameters to determine whether adjustments are needed; Step 7: Dynamically optimize the construction process.

Claims

1. A method for dynamic regulation and control of construction parameters of shaft freezing method based on optical fiber sensing, characterized in that, The method comprises the following steps: Step 1: arranging a sensing optical cable on the frozen wall and setting a pre-warning threshold value, embedding a sensor on the well wall and setting a pre-warning threshold value; Step 2: collecting and analyzing the parameters of the frozen wall and the well wall in real time to form raw data; Step 3: preprocessing and feature extraction of the raw data; Step 4: transmitting the integrated on-site construction parameters to the monitoring center; Step 5: constructing a dual-source data set, establishing a three-dimensional geological-frozen-construction coupling model, and determining the optimal construction parameters under different geological conditions; Step 6: comparing the construction parameters monitored by the sensor with the optimal construction parameters to determine whether adjustment is needed; Step 7: dynamically optimizing the construction process.

2. The optical fiber sensing based construction parameter dynamic regulation and control method for shaft freezing method according to claim 1, characterized in that, The step 1 is performed according to the following steps: Step 1.1, lowering a densely distributed temperature sensing optical cable and a densely distributed strain sensing optical cable along the frozen hole on the frozen wall to monitor the temperature of the frozen wall and the rock mass strain, and simultaneously installing a sound-light alarm; Step 1.2, arranging multiple fiber grating type temperature sensors, multiple fiber grating type concrete strain gauges, multiple fiber grating type steel stress gauges, multiple inter-wall fiber grating type pressure sensors, and multiple fiber grating type water pressure sensor in an array along the depth direction of the well wall, simultaneously installing a sound-light alarm, and placing a fiber grating demodulator in a computer room on the ground, connecting the sensors in series through the self-provided optical fiber, and then connecting the main communication optical cable with the fiber grating demodulator on the ground; Step 1.3, setting the pre-warning threshold value of the densely distributed temperature sensing optical cable to -30℃~-25℃, setting the pre-warning threshold value of the fiber grating type temperature sensor to -10℃~-8℃, setting the pre-warning threshold value of the fiber grating type concrete strain gauge to 2000με~2400με, setting the pre-warning threshold value of the fiber grating type steel stress gauge to 30MPa~32Mpa, setting the pre-warning threshold value of the inter-wall fiber grating type pressure sensor to 10MPa~12Mpa, and setting the pre-warning threshold value of the fiber grating type water pressure sensor to 0~0.5MPa.

3. The method for dynamic control of construction parameters of vertical shaft freezing method based on fiber optic sensing according to claim 2, characterized in that, The step 2 is performed according to the following steps: Step 2.1, synchronously collecting the parameters of the frozen wall and the well wall and the on-site construction equipment operation parameters by using an industrial Ethernet, and setting the collection frequency to 0.8 times / s~1.2 times / s, the parameters of the frozen wall and the well wall including the temperature of the frozen wall, the rock mass strain of the frozen wall, the temperature of the well wall concrete, the strain of the well wall concrete, the deformation stress of the well wall steel, the contact pressure between the well wall and the frozen wall, and the water seepage pressure between the double-layer well walls, the on-site construction equipment operation parameters including the flow rate of the freezer, the temperature of the salt water inlet and outlet in the pipe, the power of the freezer, the output flow rate of the grouting pump, and the temperature of the grouting liquid, and simultaneously installing an alarm indicator and a buzzer on the control panel of the freezer and the grouting pump; Step 2.2, receiving the optical signal transmitted by the main communication optical cable through the fiber grating demodulator, converting the wavelength change information in the optical signal into an electrical signal, calculating the quantitative parameter value of the parameters of the frozen wall and the well wall, synchronously marking the collection time and the on-site construction equipment operation parameters, and forming raw data.

4. The method for dynamic control of construction parameters of vertical shaft freezing method based on fiber optic sensing according to claim 3, characterized in that, The step 3 is performed according to the following steps: Step 3.1, mark the original data exceeding the warning threshold of each sensor as suspected abnormal data, and call the synchronous data of at least 3 adjacent sensors of the same type around the suspected abnormal data for neighborhood verification, and judge the suspected abnormal data with temperature deviation absolute value exceeding 2℃, stress deviation absolute value exceeding 1MPa and strain deviation absolute value exceeding 100με as real abnormal values; Step 3.2, replace the real abnormal values with linear interpolation method: wherein, x t 1、 x t+1 is normal data before and after the abnormal moment, is the real abnormal value collection time; Step 3.3, normalize the original data by min-max standardization method: wherein x is the original data, x min , x max is the theoretical extreme value of the data; Step 3.4, reduce the sampling of the freezing wall temperature, well wall concrete temperature, temperature of salt water inlet and outlet in the pipe, and temperature of grouting liquid, and generate virtual data points per second for the freezing wall rock mass strain, well wall concrete strain, well wall steel deformation stress, well wall and freezing wall contact pressure, double-layer well wall seepage water pressure, freezing device flow, power of refrigeration machine, and output flow of grouting pump by linear interpolation.

5. The method for dynamic control of construction parameters of vertical shaft freezing method based on fiber optic sensing according to claim 4, characterized in that, The step 4 is performed as follows: Step 4.1, collect geological parameters in the early exploration stage, the geological parameters include stratum lithology, stratum water content, stratum permeability, rock mass compressive strength, and internal friction angle, and a monitoring center is arranged on the ground, the monitoring center is provided with a data server, a visualization platform, an audible and light alarm device, and an alarm control host, the data server is connected with the alarm control host and the visualization platform, and the alarm control host is signal connected with the audible and light alarm device; Step 4.2, synchronously transmit the field construction equipment operation parameters and the geological parameters to the data server of the ground monitoring center through a data interaction interface, uniformly format and time stamp align the construction equipment operation parameters and the geological parameters through a data preprocessing module, form an associated data set containing "geological conditions-equipment operation state", and the visualization platform displays the temperature field distribution, the stress nephogram, and the associated data set in real time.

6. The method according to any one of claims 1-5, wherein the method is characterized by, The step 5 is performed as follows: Step 5.1, select at least 50 freezing construction projects of vertical shafts, which cover siltstone, fine-grained sandstone, medium-grained sandstone, coarse-grained sandstone, sandy mudstone stratum types and construction scenes under a 5%-25% water content gradient, and simultaneously contain successful cases of freezing wall stable forming and well wall without cracks and typical cases of local thawing of freezing wall and over-limit strain of well wall concrete, collect geological parameters, freezing wall and well wall parameters and field construction equipment operation parameters of each case, and the monitoring values are within the threshold values specified in the "Coal Mine Roadway Engineering Construction Standard", all data are time stamped, as actual case data, and the monitoring values include well wall concrete strain, well wall steel deformation stress, well wall concrete temperature, freezing wall radial temperature and strain; Step 5.2, establish a three-dimensional geological-freezing-construction coupling model through COMSOL, divide the whole model by using "free tetrahedral mesh", and reduce the mesh size of the freezing hole periphery and the contact zone between the well wall and the freezing wall to 0.2m-0.5m, and use a mesh with a size of 1m-2m for the stratum far away from the shaft. Step 5.3, using the transient solver to solve the nonlinear equations using the "Newton-Raphson iteration method", the frozen wall and well wall parameters under each working condition are calculated, and the convergence curve is monitored in real time during the calculation to ensure that the residual is less than 1e-6, the calculation results are extracted and arranged into a data set in the format of "parameter input-monitoring output", which is used as simulation data; Step 5.4, integrate the actual case data and simulation data according to 6:4 to form a dual-source data set with at least 150,000 data entries; Step 5.5, filter features from the dual-source data set, use formation type, water cut, and compressive strength as input features, extract brine temperature, brine flow, grouting flow, and slurry temperature as key variables, and use well wall concrete temperature, well wall concrete strain, well wall steel deformation stress, frozen wall temperature, and frozen wall rock mass strain as output labels, with "yes" marked as 1 and "no" marked as 0; Step 5.6, use the gradient boosting tree algorithm to train, divide the dual-source data set into training set and validation set according to 8:2, control overfitting with 5-fold cross-validation, optimize tree depth and learning rate with grid search, and perform at least 1000 iterations of training, when the classification accuracy on the validation set is not less than 92%, use K-means clustering algorithm to group the output "candidate optimal parameters", and determine the optimal construction parameters under different geological conditions.

7. The optical fiber sensing based construction parameter dynamic regulation and control method for shaft freezing method according to claim 6, characterized in that, The three-dimensional geology-freezing-construction coupling model takes "heat conduction-porous medium seepage-structural mechanics" three-field coupling as the core, uses the built-in coupling algorithm of COMSOL to realize the calculation of mutual interaction of multiple physical fields, the heat conduction algorithm is: wherein wherein p is the medium density, c is the specific heat capacity, T is the temperature, t is the time, k is the thermal conductivity, and Q is the internal heat source intensity. Porous medium seepage algorithm: wherein, is the permeability, is the pore water pressure, is the water density, is the water compressibility, and t is time; Structural mechanics algorithm: wherein .

8. The optical fiber sensing based construction parameter dynamic regulation and control method for shaft freezing method according to claim 7, characterized in that, The steps of grouping the "candidate optimal parameters" output by the model are: Step a, filter out the construction parameter combinations with output label 1 from the gradient boosting tree model output results as "candidate optimal parameters", and specify the clustering features as formation type, formation water cut, brine temperature, brine flow, and grouting temperature; Step b, according to the siltstone, fine-grained sandstone, medium-grained sandstone, coarse-grained sandstone, and sandy mudstone formation types and 5%-25% water cut gradient covered in the dual-source data set, combined with the goal of "one set of optimal parameter interval corresponding to the same geological condition", determine the K value by elbow rule, calculate the within-cluster sum of squares for different K values, and select the K value corresponding to the point where the within-cluster sum of squares decreases sharply; Step c, randomly select K "candidate optimal parameter" samples as initial cluster centers, calculate the Euclidean distance between each parameter sample and each cluster center, assign the sample to the nearest cluster, calculate the mean of each feature of all samples in each cluster, and update the new cluster center, repeat the "distance calculation-sample assignment-center update" steps until the center coordinates change by less than 0.1% for 3 consecutive iterations or the iteration number reaches 100. Step d: For each group of parameters after clustering, calculate the 95% confidence interval of each construction parameter respectively to form the "geological condition-optimal construction parameter interval" corresponding table, and verify the effectiveness of the interval. Randomly select 10% of the parameter combinations within each interval and input them into the gradient boosting tree model to verify whether the output label is 1. Ensure that more than 95% of the parameter combinations within the interval can make the monitoring value meet the standard. Finally, output the structured grouping results.

9. The optical fiber sensing based construction parameter dynamic regulation and control method for shaft freezing method according to claim 8, characterized in that, The step 6 is specifically: Step 6.1, compare the optimal construction parameters with the current construction parameters, calculate the parameter deviation rate : ×100% Wherein, m is the current construction parameter, and n is the optimal construction parameter. When the deviation rate is 5%-10%, trigger the audible and light alarm, and at the same time, combine the real-time frozen wall and well wall parameters to call the "geological condition-optimal construction parameter interval" corresponding table to generate fine-tuning suggestions. When the deviation rate is greater than 10% or the current construction parameter exceeds the warning threshold, trigger the pop-up window alarm and push the out-of-limit parameters, location and control scheme. After the parameters return to the warning threshold and the deviation rate is less than 5%, resume the construction.

10. The optical fiber sensing based construction parameter dynamic regulation and control method for shaft freezing method according to claim 9, characterized in that, The step 7 is specifically: collect the control effect data every 24 hours as a cycle, and repeat steps 5 and 6 until the parameter deviation rate is stable and meets the standard in the last 3 cycles and no single parameter deviation rate is greater than 10%. Complete the dynamic control.

Images