Function decoupling regulation and control system and method for hot and humid environment of TBM tunneling face of deep mine

Abstract

The invention relates to the technical field of mine tunneling, in particular to a function decoupling regulation and control system and method for the hot and humid environment of a deep mine TBM tunneling face. According to the method, multi-source environment parameters are obtained, the hot and humid state and air quality data are generated through fusion processing, and the current hot and humid grade and the oxygen demand are recognized; outputting an environment grade result and an oxygen supply demand signal; and constructing an oxygen supply-cooling-environment coupling thermodynamic model, and estimating the functional coupling degree of the two systems in a future time period in combination with a dynamic prediction method. The self-adaptive decoupling controller is used for dynamically optimizing energy consumption distribution of the oxygen supply and cooling system according to a prediction result, regulation and control parameters are generated, control signals of oxygen supply equipment and cooling equipment are output according to the regulation and control parameters, and dynamic decoupling of the oxygen supply and cooling functions is achieved. The problems of response lag, too high energy consumption and control conflict caused by system coupling in traditional control are solved, and the accuracy, energy efficiency and stability of tunneling face environment regulation and control in the deep high-temperature and high-humidity environment are improved.

Description

Technical Field

[0001] This invention relates to the field of mining tunneling technology, and in particular to a functional decoupling control system and method for the thermal and humid environment of a deep mine TBM tunneling face. Background Technology

[0002] As mineral resource development extends deeper, deep mine tunneling faces severe challenges from extreme hot and humid environments. Under conditions of high ground temperature, high humidity, and high stress at depth, the working area of ​​a TBM (Tunnel Boring Machine) often experiences temperatures exceeding 35°C and relative humidity approaching saturation, severely impacting equipment reliability and the health of workers. Traditional methods of controlling heat and humidity often employ centralized ventilation and mechanical refrigeration, with oxygen supply and cooling systems operating independently or in simple linkage, lacking dynamic coordination and control over their functional coupling. In actual operation, increasing oxygen supply can lead to increased heat load at the tunneling face, exacerbating the burden on the cooling system; conversely, excessive cooling can cause excessive dehumidification, reducing oxygen diffusion efficiency and even triggering condensation, creating secondary safety hazards. Furthermore, geological conditions change frequently during deep tunneling, and boundary conditions such as heat source release, moisture dissipation from surrounding rock, and oxygen consumption by personnel exhibit strong time-varying and uncertainties, causing continuous fluctuations in oxygen supply and cooling demands. Traditional fixed-ratio or experience-based control strategies struggle to achieve both optimal energy efficiency and environmental stability.

[0003] While existing technologies have attempted to introduce sensor feedback and automatic control mechanisms into local environmental intelligent control systems, most treat oxygen supply and cooling as independent subsystems, failing to fully consider the dynamic coupling characteristics of the two in thermodynamic processes, airflow organization, and energy consumption distribution. This can easily lead to control conflicts, energy waste, and response lag.

[0004] The above content is only used to help understand the technical solution of the present invention and does not represent an admission that the above content is prior art. Summary of the Invention

[0005] The main objective of this invention is to provide a functional decoupling control system and method for the thermal and humid environment of a deep mine TBM tunneling face. This aims to solve the technical problems of insufficient dynamic coordination of the oxygen supply and cooling systems in the existing thermal and humid environment control of deep mine TBM tunneling faces, which leads to control conflicts, high energy consumption, and sluggish response.

[0006] To achieve the above objectives, this invention provides a method for functional decoupling and control of the thermal and humid environment at the TBM tunneling face in deep mines, the method comprising:

[0007] Acquire multi-source environmental parameter data of the TBM tunneling face in deep mines, and fuse the multi-source environmental parameter data to generate thermal and humidity state data and air quality data;

[0008] Based on the thermal and humidity data and the air quality data, the current thermal and humidity level and oxygen demand of the tunnel face are identified, and environmental level results and oxygen supply demand signals are generated.

[0009] Based on the environmental level results and the oxygen demand signal, a coupled thermodynamic model of oxygen supply-cooling-environment is constructed, and a dynamic prediction method is used to predict the degree of functional coupling in the future time period, generating prediction results.

[0010] Based on the prediction results, an adaptive decoupling controller is used to dynamically adjust the energy consumption distribution between the oxygen supply system and the cooling system to generate control parameters.

[0011] Based on the aforementioned control parameters, control signals for the oxygen supply equipment and cooling equipment are output to decouple the oxygen supply and cooling functions.

[0012] Optionally, the generation of thermal and humidity state data and air quality data includes:

[0013] Acquire data from temperature sensor, humidity sensor, oxygen concentration sensor, and dust concentration sensor to form a raw set of environmental parameters;

[0014] The temperature sensor data is the real-time temperature value of the tunnel face, the humidity sensor data is the percentage of water vapor content in the air, the oxygen concentration sensor data is the oxygen volume percentage, and the dust concentration sensor data is the mass concentration of suspended particulate matter.

[0015] The original set of environmental parameters is preprocessed with noise reduction and drift correction to generate preprocessed environmental data;

[0016] The preprocessed environmental data is input into a multimodal fusion algorithm, which outputs the thermal and humidity state data and the air quality data.

[0017] Optionally, the step of identifying the current thermal and humidity level and oxygen demand at the tunnel face, and generating environmental level results and oxygen supply demand signals, includes:

[0018] Dry-bulb temperature and relative humidity features are extracted from the thermal and humidity state data to generate a thermal and humidity feature vector.

[0019] Oxygen partial pressure features and dust concentration features are extracted from the air quality data to generate an air quality feature vector.

[0020] The thermal and humidity feature vector and the air quality feature vector are input into the environmental assessment model, and the environmental level result and the oxygen supply demand signal are output.

[0021] The evaluation process of the environmental assessment model can be represented by the following function:

[0022] (E_level, O_signal) = F_evaluator ([V_thermo, V_air])

[0023] Where (E_level, O_signal) represents the output environmental level result and oxygen demand signal, F_evaluator represents the trained environmental assessment model function, V_thermo represents the thermal and humidity feature vector, V_air represents the air quality feature vector, and [V_thermo, V_air] represents the concatenation operation of the two feature vectors.

[0024] Optionally, the construction of the oxygen supply-cooling-environment coupled thermodynamic model includes:

[0025] Based on the environmental level results and the oxygen supply demand signal, a set of thermodynamic parameters are matched and selected from the parameter library;

[0026] Based on the aforementioned thermodynamic parameters, a coupled equation is established, comprising sensible heat exchange, latent heat exchange, and oxygen diffusion terms. This coupled equation is expressed as:

[0027] ρc∂T / ∂t=k∇²T+Q_geothermal-Q_cooling+Q_oxidation

[0028] Where ρ is air density, c is specific heat capacity of air, T is temperature field distribution at the tunnel face, k is thermal conductivity coefficient, Q_geothermal is the power released by the geothermal source, Q_cooling is the heat exchange of the cooling system, and Q_oxidation is the oxidation heat release power of the equipment.

[0029] By solving the coupled equations using the finite element method, temperature field and oxygen concentration field distribution data are generated for subsequent prediction.

[0030] Optionally, the step of using a dynamic prediction method to predict the degree of functional coupling in future time periods includes:

[0031] Within each prediction step, the interaction coefficient between oxygen supply flow rate and cooling load is calculated based on the coupled thermodynamic model to generate a coupling coefficient matrix.

[0032] Based on the coupling coefficient matrix, the degree of functional coupling is quantitatively evaluated using a coupling cost function, the specific evaluation formula of which is as follows:

[0033] C=Σ[γ·(Oa-Os)²+δ·(Ta-Ts)²+ε·(Qo·Qc)]

[0034] Where C is the total functional coupling quantification value in the future prediction time domain, Oa and Os represent the actual and set oxygen concentrations respectively, Ta and Ts represent the actual and set temperatures respectively, Qo and Qc are the energy consumption of the oxygen supply system and the cooling system respectively, and γ, δ and ε are preset weighting coefficients used to balance the importance of different indicators.

[0035] The quantitatively evaluated degree of functional coupling is then updated to the prediction results.

[0036] Optionally, the step of dynamically adjusting the energy consumption distribution between the oxygen supply system and the cooling system through an adaptive decoupling controller includes:

[0037] The prediction results are input into the expert system controller to generate preliminary control parameters;

[0038] The preliminary control parameters are optimized using a model predictive control algorithm to generate adjusted control parameters.

[0039] Based on the adjusted control parameters, the energy consumption distribution ratio between the oxygen supply system and the cooling system is calculated, and the control parameters are generated.

[0040] Optionally, after generating the control parameters, the method further includes:

[0041] When an abrupt change in the environmental grade result of heat and humidity is detected, a buffer regulation mechanism is activated to smooth the change in the regulation parameter;

[0042] The response speed of the cooling equipment control signal is adjusted by a feedforward compensation algorithm to counteract the coupling with pressure fluctuations generated by the oxygen supply system.

[0043] The control parameters are optimized online based on real-time energy efficiency ratio and environmental comfort indicators.

[0044] Optionally, the output oxygen supply equipment control signal and cooling equipment control signal include:

[0045] The aforementioned control parameters are combined with environmental target indicators and converted into target operating parameters for oxygen supply fans and refrigeration units;

[0046] The oxygen supply equipment control signal is generated based on the target operating parameters using a PID closed-loop controller.

[0047] The cooling equipment control signal is generated based on the target operating parameters using a fuzzy controller.

[0048] Optionally, after generating control parameters by dynamically adjusting the energy consumption distribution between the oxygen supply system and the cooling system through an adaptive decoupling controller, the process further includes an online learning and optimization step.

[0049] Collect historical operating data, including environmental status data, the aforementioned control parameters, and system energy efficiency indicators;

[0050] The decision-making strategy of the adaptive decoupling controller is adjusted based on the historical operating data using reinforcement learning algorithms.

[0051] By using a transfer learning model, the control parameters under new conditions are optimized based on historical data from similar operating conditions, thereby improving the environmental adaptability of the adaptive decoupling controller.

[0052] Furthermore, to achieve the above objectives, the present invention also provides a functional decoupling control system for the thermal and humid environment of a deep mine TBM tunneling face, the system comprising:

[0053] The data fusion module is used to acquire multi-source environmental parameter data of the TBM tunneling face in deep mines, and to fuse the multi-source environmental parameter data to generate thermal and humidity state data and air quality data.

[0054] The status recognition module is used to identify the current thermal and humidity level and oxygen demand of the tunnel face based on the thermal and humidity status data and the air quality data, and generate environmental level results and oxygen supply demand signals.

[0055] The coupling prediction module is used to construct an oxygen supply-cooling-environment coupling thermodynamic model based on the environmental level results and the oxygen supply demand signal, and to predict the degree of functional coupling in the future time period using a dynamic prediction method, thereby generating prediction results.

[0056] The decoupling control module is used to dynamically adjust the energy consumption distribution between the oxygen supply system and the cooling system based on the prediction results, and generate control parameters through an adaptive decoupling controller.

[0057] The instruction output module is used to output control signals for the oxygen supply equipment and the cooling equipment according to the control parameters, so as to decouple the oxygen supply and cooling functions.

[0058] This invention provides a method for functional decoupling and control of the thermal and humid environment at the tunneling face of a deep mine TBM. The method constructs a coupled thermodynamic model of oxygen supply, cooling, and environment, and predicts the degree of functional coupling. This achieves precise decoupling of the previously mutually influencing oxygen supply and cooling functions, avoiding control conflicts caused by traditional independent control or simple linkage, such as increased oxygen supply exacerbating heat load or excessive cooling affecting oxygen diffusion. This significantly improves the stability and efficiency of overall environmental control. The adaptive decoupling controller can dynamically adjust the energy consumption allocation of the oxygen supply and cooling systems based on prediction results. Combining environmental level results and oxygen demand signals, it achieves optimal energy consumption while meeting thermal and humid control and oxygen demand, reducing the total system energy consumption. Through the fusion processing of multi-source environmental parameters and thermal and humidity levels, Accurate identification of oxygen demand enables more precise output of control signals, effectively improving the thermal and humid environment and air quality at the tunneling face, providing a more comfortable and safe working environment for workers, and reducing safety hazards and health risks caused by harsh environments. Dynamic prediction methods and adaptive control strategies enable the system to better cope with the strong time-varying and uncertain boundary conditions such as geological conditions, heat source release, and personnel oxygen consumption during deep mine TBM tunneling, improving the system's adaptability and robustness to complex and changing environments. This method integrates technologies such as multi-source data fusion, intelligent identification, model prediction, and adaptive control, promoting the transformation of environmental control at deep mine tunneling faces from experience-based and extensive to intelligent and refined management, and providing strong support for achieving efficient, safe, and green deep resource mining. Attached Figure Description

[0059] Figure 1 This is a flowchart illustrating an embodiment of the functional decoupling and control method for the thermal and humid environment of a deep mine TBM tunneling face according to the present invention.

[0060] Figure 2 This is a structural block diagram of an embodiment of the functional decoupling and control system for the thermal and humid environment of a deep mine TBM tunneling face according to the present invention.

[0061] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0062] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0063] Reference Figure 1 , Figure 1 This is a flowchart illustrating an embodiment of the functional decoupling and control method for the thermal and humid environment of a deep mine TBM tunneling face according to the present invention.

[0064] In one embodiment, the method for functional decoupling and controlling the thermal and humid environment of the deep mine TBM tunneling face includes:

[0065] Step S100: Obtain multi-source environmental parameter data of the deep mine TBM tunneling face, and perform fusion processing on the multi-source environmental parameter data to generate thermal and humidity state data and air quality data.

[0066] The multi-source environmental parameter data can be a set of raw measurements reflecting the thermal and humidity conditions and air quality at the tunnel face, collected from multiple sensors or monitoring devices. This data can provide basic input for subsequent data fusion and environmental condition identification. In this embodiment, the multi-source environmental parameter data can be collected in real time by sensors deployed in the TBM tunnel face area, including sensors for temperature and humidity, oxygen concentration, wind speed, surrounding rock temperature, and personnel location. The thermal and humidity state data can be structured information, after fusion processing, characterizing the current comprehensive temperature and humidity state of the tunnel face. This data can be used to quantify the current thermal and humidity environmental level and support thermal and humidity level identification. Furthermore, the thermal and humidity state data can be obtained by extracting elements such as temperature, humidity, and surrounding rock moisture dissipation rate from the multi-source environmental parameter data and then normalizing and integrating them. The air quality data can be structured information, after fusion processing, characterizing the oxygen concentration and diffusion conditions at the tunnel face. This data can be used to assess oxygen supply effectiveness and diffusion efficiency, supporting oxygen demand identification. In a specific embodiment, the air quality data can be obtained by extracting elements such as oxygen concentration, wind speed, and airflow organization morphology from the multi-source environmental parameter data and then performing correlation modeling.

[0067] Acquiring multi-source environmental parameter data at the TBM excavation face in deep mines can be achieved by real-time collection of raw data such as temperature and humidity, oxygen concentration, wind speed, surrounding rock temperature, and personnel location in the excavation face area through a distributed sensor network. Furthermore, this operation can be implemented using the aforementioned sensor deployment methods, thereby establishing a comprehensive and real-time environmental perception foundation. Fusion processing of multi-source environmental parameter data can involve time synchronization, noise filtering, feature extraction, and semantic integration of heterogeneous data from different sensors. Further, multi-source environmental parameter data fusion can be achieved by employing Kalman filtering for multi-source data state estimation or using deep neural networks for end-to-end feature fusion, thereby improving the accuracy and robustness of environmental state characterization. Generating thermal and humidity state data and air quality data can be achieved by organizing the fused data into structured feature vectors according to the thermal and humidity and air quality dimensions, respectively. This operation further provides a clear input channel for subsequent hierarchical identification.

[0068] Step S200: Based on thermal and humidity data and air quality data, identify the current thermal and humidity level and oxygen demand at the tunnel face, and generate environmental level results and oxygen supply demand signals.

[0069] The thermal humidity level can be a discrete category of environmental thermal humidity severity based on thermal humidity state data, which can be used to provide a grading basis for cooling system regulation. In this embodiment, the thermal humidity level can be obtained by mapping the thermal humidity state data through a preset threshold range or an intelligent classification algorithm. Oxygen demand can be the oxygen supply level required to maintain safe operation, determined based on air quality data and personnel activity status, and can be used to provide a demand benchmark for oxygen supply system regulation. In a specific embodiment, oxygen demand can be dynamically calculated by combining oxygen concentration, number of personnel, labor intensity, and diffusion efficiency. The environmental level result can be a structured output of the thermal humidity level, used for subsequent model construction and control decisions, and can be used as one of the input variables of the coupled thermodynamic model to characterize thermal humidity boundary conditions. The oxygen supply demand signal can be an executable signal form of oxygen demand, used to drive the adjustment of oxygen supply equipment, and can be used as one of the input variables of the coupled thermodynamic model to characterize oxygen supply boundary conditions.

[0070] Identifying the current thermal and humidity levels and oxygen demand at the tunnel face can be done based on thermal and humidity data and air quality data, using threshold judgment or classification models to determine the levels and demand levels. Furthermore, this operation can be achieved by using fuzzy C-means clustering for level classification or support vector machines for demand classification, thereby transforming continuous environmental conditions into actionable discrete control criteria. Generating environmental level results and oxygen demand signals can be achieved by converting the identified thermal and humidity levels and oxygen demands into a standardized signal format. This operation further enables the effective transmission of environmental perception results to the control system.

[0071] Step S300: Based on the environmental level results and oxygen demand signals, construct an oxygen supply-cooling-environment coupled thermodynamic model, and use a dynamic prediction method to predict the degree of functional coupling in the future time period, generating prediction results.

[0072] The oxygen supply-cooling-environment coupled thermodynamic model can be a dynamic physical model describing the heat and mass transfer and energy among the oxygen supply system, cooling system, and tunnel face environment. It can be used to reveal the coupling mechanism between oxygen supply and cooling functions, providing a theoretical basis for decoupling control. In this embodiment, the oxygen supply-cooling-environment coupled thermodynamic model can be obtained by identifying model parameters based on mass conservation, energy conservation, and gas diffusion equations, combined with measured data. Furthermore, the oxygen supply-cooling-environment coupled thermodynamic model can receive environmental level results and oxygen demand signals as input, and output the dynamic relationship used to predict the degree of functional coupling; its output affects the energy consumption allocation strategy of the adaptive decoupling controller. The degree of functional coupling can be a quantitative indicator of the mutual interference or enhancement of oxygen supply and cooling functions under specific environmental conditions, which can be used to characterize the level of control conflict risk and guide the formulation of decoupling control strategies. In an exemplary embodiment, the degree of functional coupling can be obtained through coupled thermodynamic model simulation or online identification to determine the intensity of the impact of oxygen supply airflow changes on cooling load and the effect of refrigeration and dehumidification on oxygen diffusion efficiency. The prediction results can be numerical or trend estimates of the degree of functional coupling over a future time period, and can be used to provide a forward-looking decision-making basis for adaptive decoupling controllers. In one specific embodiment, the prediction results can be generated using time series forecasting, state-space extrapolation, or machine learning methods based on a coupled thermodynamic model and the current state.

[0073] Constructing a coupled thermodynamic model of oxygen supply, cooling, and the environment can be achieved by establishing differential equations or data-driven models describing these three components based on physical laws and measured data. Furthermore, this operation can be implemented by constructing a partial differential equation model using the finite volume method or a black-box data model using long short-term memory networks, thereby revealing the intrinsic coupling mechanism between oxygen supply and cooling functions. Predicting the degree of functional coupling in future time periods using dynamic prediction methods can be achieved by extrapolating the functional coupling strength at several future time steps using the coupled thermodynamic model and the current state. Furthermore, this operation can be implemented through rolling prediction based on a state-space model or time-series prediction using recurrent neural networks, thereby endowing the system with forward-looking control capabilities. Generating prediction results can be achieved by organizing the output of the dynamic prediction method into a structured data format. Furthermore, this operation provides the controller with resolvable prediction information.

[0074] Step S400: Based on the prediction results, the energy consumption distribution between the oxygen supply system and the cooling system is dynamically adjusted by an adaptive decoupling controller to generate control parameters.

[0075] The adaptive decoupling controller can be an intelligent control unit that dynamically adjusts the control strategies of the oxygen supply and cooling systems based on prediction results to weaken the effects of functional coupling. It can be used to achieve coordinated operation of oxygen supply and cooling functions, avoid control conflicts, and optimize energy consumption. In this embodiment, the adaptive decoupling controller can obtain control parameters in real time by employing algorithms such as model predictive control, fuzzy logic, or reinforcement learning. Furthermore, the adaptive decoupling controller can receive prediction results as input and output control parameters; its operation depends on the dynamic relationship provided by the coupled thermodynamic model. Energy consumption allocation can be the energy consumption ratio or weight configuration between the oxygen supply system and the cooling system under the premise of meeting environmental and oxygen supply needs. This can be used to minimize the total energy consumption of the system while ensuring environmental stability. In a specific embodiment, energy consumption allocation can be obtained by the adaptive decoupling controller solving for the optimal allocation scheme based on prediction results and constraints. Control parameters can be a set of specific control command parameters used to drive the operation of the oxygen supply and cooling equipment, which can be used to transform the decoupling control strategy into executable equipment actions.

[0076] Dynamically adjusting the energy consumption distribution between the oxygen supply and cooling systems using an adaptive decoupling controller can be achieved by solving for the optimal energy consumption distribution scheme based on prediction results, while satisfying environmental and oxygen supply constraints. Furthermore, this operation can be implemented by using model predictive control to solve a rolling optimization problem or by using a Q-learning algorithm to adjust the distribution strategy online, thereby achieving a balance between optimal energy efficiency and stable control. Generating control parameters can involve mapping the energy consumption distribution results to specific equipment control parameters. This operation further facilitates the transformation from strategy to execution.

[0077] Step S500: Based on the control parameters, output the control signals for the oxygen supply equipment and the cooling equipment to decouple the oxygen supply and cooling functions.

[0078] The oxygen supply equipment control signal can be an electrical signal or command sequence acting on the oxygen supply equipment (such as a fan or oxygen generator), which can be used to adjust the oxygen supply and air delivery characteristics to meet oxygen demand and avoid exacerbating heat load. The cooling equipment control signal can be an electrical signal or command sequence acting on the cooling equipment (such as a refrigeration unit or spray system), which can be used to adjust the cooling capacity and dehumidification intensity to meet heat and humidity control and avoid inhibiting oxygen diffusion. Outputting the oxygen supply and cooling equipment control signals can be done by converting the control parameters into electrical signals or communication commands that the equipment can recognize. Furthermore, this operation enables the driving physical equipment to perform decoupled control actions. Achieving decoupling of the oxygen supply and cooling functions can be achieved through independent but coordinated control signals, ensuring that the oxygen supply and cooling systems do not interfere with each other functionally and can even promote each other. Furthermore, this operation eliminates conflicts in traditional control methods and improves the overall system performance.

[0079] Taking deep, high-stress TBM tunneling operations as an example, the functional decoupling control method for the thermal and humid environment of the deep mine TBM tunneling face in this embodiment can be as follows: In a deep metal mine TBM tunneling face, the surrounding rock temperature reaches 42℃, the relative humidity is 98%, and there are 6 workers. The system acquires multi-source environmental parameter data through sensors such as temperature, humidity, O2, and wind speed, and after fusion, generates thermal and humid state data (showing severe local heat accumulation) and air quality data (showing obstructed oxygen diffusion). The thermal and humid level is identified as severe, and the oxygen demand is identified as a high-load emergency level. The coupled thermodynamic model predicts that if the oxygen supply air volume is increased in the traditional way, the cooling load will increase by 15%; if excessive cooling and dehumidification are carried out, the oxygen diffusion efficiency will decrease by 20%. The adaptive decoupling controller dynamically allocates energy consumption accordingly, appropriately reducing the air supply speed but increasing the local oxygen concentration, while using intermittent spray cooling to avoid excessive dehumidification. Ultimately, it outputs independent control signals for oxygen supply equipment (adjusting the variable frequency fan and oxygen enrichment module) and cooling equipment (controlling the spray flow and the start / stop of the refrigeration unit), achieving a stable thermal and humid environment of 32℃ / 85%RH and maintaining an oxygen concentration of over 19.5%, while reducing the total energy consumption of the system compared to traditional linkage control.

[0080] In one embodiment, generating thermal and humidity state data and air quality data includes:

[0081] Acquire data from temperature sensor, humidity sensor, oxygen concentration sensor, and dust concentration sensor to form a raw set of environmental parameters;

[0082] Temperature sensor data is the real-time temperature value of the tunnel face; humidity sensor data is the percentage of water vapor content in the air; oxygen concentration sensor data is the oxygen volume percentage; dust concentration sensor data is the mass concentration of suspended particulate matter.

[0083] The original set of environmental parameters is preprocessed with noise reduction and drift correction to generate preprocessed environmental data.

[0084] The preprocessed environmental data is input into a multimodal fusion algorithm, which outputs thermal and humidity data and air quality data.

[0085] The temperature sensor data can be the raw measurement of the ambient temperature at the tunnel face, collected by a temperature sensor. This data can characterize the thermal load level at the tunnel face and serve as the basic input for identifying the thermal and humidity conditions. In an exemplary embodiment, the temperature sensor data can sense the ambient temperature in real time and output an electrical signal using a thermocouple, resistance temperature detector (RTD), or infrared thermometer. The humidity sensor data can be the raw measurement of the water vapor content in the air, collected by a humidity sensor and expressed as a percentage of relative humidity. This data can reflect the degree of environmental humidity and assess dehumidification needs and condensation risks. Furthermore, the humidity sensor data can detect the air moisture content using capacitive, resistive, or optical humidity-sensitive elements. The oxygen concentration sensor data can be the raw measurement of the volume percentage of oxygen in the environment, collected by an oxygen sensor. This data can directly reflect the oxygen sufficiency in the work area and support oxygen demand assessment. In this embodiment, the oxygen concentration sensor data can detect the oxygen partial pressure using electrochemical, paramagnetic, or optical principles and convert it into a concentration signal. The dust concentration sensor data can be the raw measurement of the mass concentration of suspended particulate matter in the air, collected by a dust sensor. This data can characterize the degree of air pollution and assist in assessing ventilation effectiveness and health risks. Furthermore, dust concentration sensor data can detect particulate matter concentration through light scattering, beta-ray absorption, or electrostatic induction principles.

[0086] The real-time temperature value at the tunnel face can be a specific physical quantity represented by temperature sensor data, i.e., the ambient temperature of the tunnel face at the current moment. It can be used to quantify the intensity of the thermal environment and participate in thermal and humidity state modeling. The percentage of water vapor content in the air can be a specific physical quantity represented by humidity sensor data, i.e., relative humidity. It can be used to assess the humidity-heat coupling effect and the necessity of dehumidification. The oxygen volume percentage can be a specific physical quantity represented by oxygen concentration sensor data, i.e., the proportion of oxygen in a unit volume of air. It can be used to determine whether the safety breathing standards for personnel are met. The suspended particulate matter mass concentration can be a specific physical quantity represented by dust concentration sensor data, i.e., the mass of particulate matter in a unit volume of air. It can be used to assess air cleanliness and ventilation and dust removal effects. The raw environmental parameter set can be an unprocessed multi-source dataset composed of sensor data of four types: temperature, humidity, oxygen concentration, and dust concentration. It can be used as a unified input source for subsequent preprocessing and fusion. In this embodiment, the raw environmental parameter set can be formed by aligning the outputs of each sensor by timestamp and packaging them into a structured data frame.

[0087] Acquiring data from temperature, humidity, oxygen, and dust sensors can be achieved by synchronously reading raw measurements from four types of dedicated sensors deployed at the tunnel face. Furthermore, this operation can be implemented through a synchronous acquisition mechanism, thereby constructing a sensing foundation covering four-dimensional environmental elements: heat, humidity, air, and dust. Forming the raw environmental parameter set can be achieved by integrating the data from the four types of sensors into a structured dataset using a unified timestamp and format. Further, this operation can be implemented through data alignment and encapsulation protocols, thereby establishing a unified input interface for multi-source heterogeneous data. Noise reduction and drift correction preprocessing can be data processing operations that denoise the raw environmental parameter set and compensate for sensor zero-point / sensitivity drift. This can improve the accuracy and timeliness of sensor data and avoid misjudgments due to data distortion. In this embodiment, noise reduction and drift correction preprocessing can employ wavelet transform and sliding median filtering to eliminate random noise, combined with periodically calibrated data or adaptive algorithms to correct systematic drift. For example, noise reduction and drift correction preprocessing may include, but is not limited to, one or more of the following: noise reduction correction based on statistical models, drift compensation based on physical models, and joint correction based on machine learning.

[0088] Preprocessing the original environmental parameter set by denoising and drift correction can involve applying signal processing and calibration algorithms to eliminate noise interference and sensor drift errors. Furthermore, this operation can be achieved by using wavelet threshold denoising combined with sliding window midpoint filtering and adaptive drift compensation using online calibration reference values, thereby significantly improving the reliability of the original data and adapting to the high-interference conditions of deep mines. The preprocessed environmental data can be high-quality multi-source environmental parameter data after denoising and drift correction, which can be used as reliable input for multimodal fusion algorithms, ensuring the confidence level of the fusion results. In an exemplary embodiment, the preprocessed environmental data is generated from the original environmental parameter set through preprocessing operations. Generating the preprocessed environmental data can involve reorganizing the preprocessed modal data into a standardized format. Further, this operation can be achieved through data resampling and format normalization, thereby providing high-quality input for multimodal fusion.

[0089] Multimodal fusion algorithms are computational methods that semantically integrate preprocessed environmental data from different physical modes (heat, humidity, air, dust). They can be used to generate structured, high-fidelity thermal and humidity state data and air quality data, preserving the characteristics of each mode and enhancing overall perception capabilities. In this embodiment, the multimodal fusion algorithm can mine intermodal relationships through methods such as feature mapping, attention mechanisms, or graph neural networks to generate a unified representation. Furthermore, the multimodal fusion algorithm receives preprocessed environmental data as input and outputs thermal and humidity state data and air quality data; its fusion logic directly affects the accuracy of subsequent level recognition. Inputting preprocessed environmental data into the multimodal fusion algorithm can involve sending structured preprocessed data into the fusion model for feature extraction and association modeling. Further, this operation can be implemented through a model inference interface call, thereby initiating a semantic-level environmental state reconstruction process. Outputting thermal and humidity state data and air quality data can be achieved by the multimodal fusion algorithm generating structured outputs focusing on thermal-humidity coupling characteristics and air quality characteristics, respectively. Furthermore, this operation can be achieved through a decoupled output layer design, thereby providing a highly confident and clearly decoupled state representation to support subsequent intelligent decision-making.

[0090] Taking the environmental perception of a deep tunneling face with high dust and high humidity as an example, the functional decoupling control method for the thermal and humid environment of a deep mine TBM tunneling face in this embodiment can be described as follows: In a deep copper mine TBM tunneling face, after operational disturbance, the dust concentration suddenly rises to 8 mg / m³, while water seepage from the surrounding rock causes the humidity to reach 96%, and the oxygen concentration drops to 18.8%. The system simultaneously acquires data from four types of sensors: temperature 41.2℃, humidity 96%, oxygen 18.8%, and dust 8.1 mg / m³, forming a set of original environmental parameters. Due to the temporary drift of the sensors caused by operational disturbance, the system uses wavelet denoising and drift correction based on historical averages to generate preprocessed environmental data. Subsequently, a multimodal fusion algorithm identifies the coupling effect of high humidity and high dust jointly inhibiting oxygen diffusion through an attention mechanism, outputting thermal and humid state data (marked as severe damp heat accumulation) and air quality data (marked as inefficient oxygen supply risk). This result is accurately transmitted to the subsequent thermal and humid level identification module, avoiding control lag or malfunction caused by traditional single-parameter judgment.

[0091] In one embodiment, the current thermal and humidity level and oxygen demand at the tunnel face are identified, and environmental level results and oxygen supply demand signals are generated, including:

[0092] Dry-bulb temperature and relative humidity features are extracted from thermal and humid state data to generate thermal and humid feature vectors.

[0093] The dry-bulb temperature characteristic can be a physical quantity reflecting the actual air temperature extracted from thermal and humidity state data. It can be used as a key input for thermal and humidity environment assessment to measure the heat load level. The relative humidity characteristic can be a dimensionless ratio characteristic reflecting the degree of water vapor saturation in the air extracted from thermal and humidity state data. It can be used to assess the impact of heat-humidity coupling on the human body's heat dissipation capacity and the risk of condensation. For example, the relative humidity characteristic can include, but is not limited to, local relative humidity characteristics, time-cumulative relative humidity characteristics, and disturbance-corrected relative humidity characteristics.

[0094] Extracting dry-bulb temperature and relative humidity features from thermal and humidity state data can be achieved by performing feature analysis on the data to separate the numerical or statistical quantities corresponding to dry-bulb temperature and relative humidity. Further, this operation can be implemented by extracting the mean of time-series features through a sliding window or by extracting multi-scale features using wavelet transform, thereby transforming the original thermal and humidity state into computable features with clear physical meaning. Generating a thermal and humidity feature vector can be achieved by combining the extracted dry-bulb temperature and relative humidity features into a vector according to a preset format. In a specific embodiment, the thermal and humidity feature vector can be a structured vector composed of dry-bulb temperature and relative humidity features, representing the comprehensive thermal and humidity state of the tunnel face, and can be used to provide standardized and computable thermal and humidity input for environmental assessment models. Further, the thermal and humidity feature vector can include, but is not limited to, static thermal and humidity feature vectors, dynamic differential thermal and humidity feature vectors, and multi-point fused thermal and humidity feature vectors.

[0095] Oxygen partial pressure and dust concentration features are extracted from air quality data to generate an air quality feature vector.

[0096] Among them, oxygen partial pressure characteristics can be extracted from air quality data to reflect the effective oxygen content per unit volume, and can be used to directly correlate human breathing efficiency with oxygen demand intensity. Dust concentration characteristics can be extracted from air quality data to reflect the content of suspended particulate matter in the air, and can be used to correct oxygen diffusion efficiency assessments and avoid misjudging oxygen sufficiency in high-dust environments.

[0097] Extracting oxygen partial pressure and dust concentration features from air quality data can be achieved by performing component analysis on the air quality data to separate the indicators corresponding to oxygen partial pressure and dust concentration. Furthermore, this operation can be implemented by calculating oxygen partial pressure from oxygen concentration and pressure based on the gas equation of state, or by inverting dust concentration and extracting features through the principle of optical scattering, thereby quantifying key air quality elements related to respiratory safety. Generating an air quality feature vector can be achieved by combining the extracted oxygen partial pressure and dust concentration features into a vector according to a preset format. In a specific embodiment, the air quality feature vector can be a structured vector composed of oxygen partial pressure and dust concentration features, representing the respiratory safety status of the tunnel face, and can be used to provide standardized and computable air quality input for environmental assessment models. Furthermore, the air quality feature vector can include, but is not limited to, a basic air quality feature vector, a health risk-weighted air quality feature vector, and a dynamically corrected air quality feature vector.

[0098] Input the thermal and humidity feature vector and the air quality feature vector into the environmental assessment model, and output the environmental level results and oxygen demand signal.

[0099] The evaluation process of the environmental assessment model can be represented by the following function:

[0100] (E_level, O_signal) = F_evaluator ([V_thermo, V_air])

[0101] Where (E_level, O_signal) represents the output environmental level result and oxygen demand signal, F_evaluator represents the trained environmental assessment model function, V_thermo represents the thermal and humidity feature vector, V_air represents the air quality feature vector, and [V_thermo, V_air] represents the concatenation operation of the two feature vectors.

[0102] The environmental assessment model can be a trained intelligent model used to map heat, humidity, and air quality characteristics into environmental rating results and oxygen demand signals. In one specific embodiment, the environmental assessment model can be trained using supervised learning methods based on historical environmental data and expert annotations to obtain a nonlinear mapping function, which can be used to achieve end-to-end intelligent identification of multidimensional environmental parameters into control decision variables. Furthermore, the environmental assessment model can include, but is not limited to, neural network-based environmental assessment models, decision tree ensemble-based environmental assessment models, and fuzzy reasoning-based environmental assessment models.

[0103] F_evaluator can be a mathematical function representation of the environmental assessment model, reflecting the mapping relationship from input feature vectors to output control signals. It can be used to formally describe the environmental assessment process, facilitating real-time invocation within the control system. [V_thermo, V_air] can be the result of concatenating the thermal humidity feature vector and the air quality feature vector, constituting the joint input of the environmental assessment model. In an exemplary embodiment, [V_thermo, V_air] can be concatenated horizontally or vertically along channels or dimensions, which can be used to preserve the independent semantics of the two types of features while establishing a joint representation, supporting the model's learning of cross-domain coupling relationships. Inputting the thermal humidity feature vector and the air quality feature vector into the environmental assessment model can be achieved by concatenating the two feature vectors and then feeding them into the deployed F_evaluator model for forward computation. Furthermore, this operation can be achieved by connecting V_thermo and V_air along the feature dimension to form a joint input vector, thereby initiating the intelligent assessment process and realizing multi-source feature fusion and discrimination.

[0104] The output environmental level results and oxygen demand signal can be obtained by reading the values ​​of E_level and O_signal from the output layer of the environmental assessment model, respectively. E_level can be the thermal and humidity environmental level result output by the environmental assessment model, a discrete categorical variable that can be used as a grading basis for cooling system regulation, replacing traditional single temperature and humidity threshold judgments. For example, E_level can include, but is not limited to, comfort level E_level, warning level E_level, and danger level E_level. O_signal can be the oxygen demand signal output by the environmental assessment model, a continuous or graded numerical variable that can be used as a quantitative command for oxygen supply system regulation, comprehensively considering oxygen partial pressure and dust interference.

[0105] Taking the environmental identification of a deep, high-humidity, and high-dust TBM tunneling face as an example, the functional decoupling and control method for the thermal and humid environment of a deep mine TBM tunneling face in this embodiment can be as follows: In a deep copper mine TBM tunneling face, the system extracts the dry-bulb temperature of 36.2℃ (moving average feature) and relative humidity of 97% (local peak feature) from the thermal and humidity state data to form a thermal and humidity feature vector; and extracts the oxygen partial pressure of 18.1kPa (after pressure correction) and PM10 concentration of 4.8mg / m³ (time cumulative feature) from the air quality data to form an air quality feature vector. The two are then concatenated and input into the environmental assessment model F_evaluator trained based on a deep neural network. The model comprehensively judges that: high humidity inhibits sweat evaporation, and high dust hinders oxygen diffusion; although the oxygen concentration has not reached the warning value, the effective oxygen supply is insufficient. Therefore, the output E_level is "severe thermal and humidity level", and O_signal is "high-load emergency oxygen supply demand". This result is more in line with actual physiological needs than the traditional method of judging "sufficient oxygen supply" based solely on an oxygen concentration >19%. It avoids the risk of insufficient oxygen supply due to dust interference and provides accurate input for the subsequent decoupling controller.

[0106] In one embodiment, a coupled thermodynamic model of oxygen supply-cooling-environment is constructed, including:

[0107] Based on the environmental level results and oxygen supply demand signals, a set of thermodynamic parameters is matched and selected from the parameter library;

[0108] The parameter library can be a structured database storing thermodynamic parameters corresponding to different combinations of environmental levels and oxygen demand. It can be used to support the model in dynamically calling adaptive parameters based on the current environmental level results and oxygen demand signals, improving model adaptability. In this embodiment, the parameter library can be pre-built and continuously updated using historical operating data, laboratory calibration, or numerical simulation. Thermodynamic parameters can be a set of physical quantities describing air properties, heat exchange efficiency, and diffusion characteristics under specific operating conditions. They can be used as the basic input for constructing coupled equations, ensuring the model is consistent with the current tunnel face state. Furthermore, thermodynamic parameters can be selected from the parameter library based on matching environmental level results and oxygen demand signals. Matching and selecting a set of thermodynamic parameters from the parameter library based on environmental level results and oxygen demand signals can be achieved by using the current environmental level results and oxygen demand signals as index keys to retrieve the most matching set of thermodynamic parameters from the parameter library. Further, this operation can be achieved by using a nearest neighbor matching strategy to select parameter combinations, or by generating intermediate operating condition parameters in a multi-dimensional parameter space based on interpolation methods. This enables the coupled thermodynamic model to have operating condition adaptive capabilities, improving modeling accuracy.

[0109] Based on thermodynamic parameters, a coupled equation is established that includes sensible heat exchange terms, latent heat exchange terms, and oxygen diffusion terms. The coupled equation is expressed as:

[0110] ρc∂T / ∂t=k∇²T+Q_geothermal-Q_cooling+Q_oxidation

[0111] Where ρ is air density, c is specific heat capacity of air, T is temperature field distribution at the tunnel face, k is thermal conductivity coefficient, Q_geothermal is the power released by the geothermal source, Q_cooling is the heat exchange of the cooling system, and Q_oxidation is the oxidation heat release power of the equipment.

[0112] The coupled equations include: Sensible heat transfer term, which characterizes the effect of heat transfer on the temperature field under no-phase-change conditions, reflecting dry-bulb temperature changes caused by geothermal activity, equipment oxidation, and cooling systems; Latent heat transfer term, which characterizes the absorption or release of heat during water vapor phase change (evaporation / condensation), depicting the effects of surrounding rock moisture dissipation, spray evaporation, or refrigeration condensation on heat balance in high-humidity environments; and Oxygen diffusion term, a mathematical expression describing the concentration transport and diffusion of oxygen in a non-uniform airflow field, quantifying the effects of oxygen supply airflow, humidity, and temperature on oxygen distribution uniformity and breathing zone concentration. The coupled equations integrate sensible heat, latent heat, and oxygen diffusion processes into partial differential governing equations, implicitly including oxygen concentration transport equations. They provide a unified description of the physical coupling mechanism between oxygen supply, cooling, and the environment, offering a theoretical basis for prediction. Air density (ρ), the mass of air per unit volume, is affected by temperature, pressure, and humidity and can be used as a physical property parameter in the coupled equations, influencing heat capacity and flow characteristics calculations. The specific heat capacity of air (c) can be the amount of heat required to raise the temperature of a unit mass of air by a unit. It can be used in sensible heat calculations to determine the temperature response sensitivity.

[0113] The temperature field distribution (T) at the tunnel face can be represented as a continuous function of the temperature at various points within the TBM tunnel face region. It can be used as a solution variable in the coupled equations to reflect the spatial heterogeneity of the thermal environment. The thermal conductivity coefficient (k) can be a physical property parameter characterizing the thermal conductivity of air or surrounding rock. It can be used to determine the thermal diffusion rate and control the intensity of the ∇²T term in the equations. The geothermal power released (Q_geothermal) can be the geothermal power released from the surrounding rock into the tunnel face space. It can be used as the primary heat source term in the coupled equations, driving the temperature rise. The heat exchange of the cooling system (Q_cooling) can be the total heat removed from the tunnel face by the cooling equipment. It can be used as a controllable negative heat source term to offset geothermal and oxidation heat release. The equipment oxidation heat release power (Q_oxidation) can be the additional heat power generated by metal oxidation or electrical losses during TBM equipment operation. It can be used as a secondary but not negligible internal heat source term. Based on thermodynamic parameters, a coupled equation comprising sensible heat exchange, latent heat exchange, and oxygen diffusion terms can be established. This can be achieved by substituting the matched thermodynamic parameters into a partial differential equation framework based on conservation laws, forming a complete coupled equation. Furthermore, this operation can be accomplished by deriving the conservation form equations using the control volume method or by introducing the Boussinesq approximation to simplify the buoyancy term, thereby achieving a unified mathematical description of the physical processes involving oxygen supply, cooling, and the environment.

[0114] By solving the coupled equations using the finite element method, temperature field and oxygen concentration field distribution data are generated for subsequent prediction.

[0115] The finite element method (FEM) can be a method of discretizing coupled partial differential equations and numerically solving them within the computational domain. This can be used to obtain high spatial resolution physical field distributions, supporting accurate predictions. In this embodiment, the FEM can employ mesh generation, shape function interpolation, and iterative algorithms to solve for the temperature and concentration fields. The temperature field distribution data can be the numerical temperature distribution results of the tunnel face space output by the FEM, which can be used to assess the thermal environment state and characterize the spatial heterogeneity of the thermal environment. The oxygen concentration field distribution data can be the numerical oxygen concentration distribution results of the tunnel face space output by the FEM, which can be used to assess oxygen supply effectiveness and diffusion bottlenecks, supporting oxygen demand correction. Solving the coupled equations using the FEM can involve meshing the computational domain of the tunnel face and applying the finite element method to numerically solve for the temperature and oxygen concentration fields. Furthermore, this operation can be implemented using multiphysics simulation software for multiphysics coupling solutions, or by developing a custom solver based on the open-source FEniCS platform, thereby obtaining spatiotemporally continuous and physically consistent field distribution data to support high-precision predictions. Generating temperature and oxygen concentration field distribution data for subsequent prediction can be achieved by organizing the finite element solution results into a structured spatiotemporal data format for the prediction module to call, thereby providing high-fidelity state input for predicting the degree of functional coupling.

[0116] Taking the modeling of a deep, high-humidity, and high-temperature TBM tunneling face as an example, the functional decoupling control method for the thermal and humid environment of a deep mine TBM tunneling face in this embodiment can be implemented in a deep mine tunneling face where the environmental level is severe heat and humidity, and the oxygen demand signal indicates a high load. The system matches thermodynamic parameters under high humidity conditions from the parameter library, including moist air density, effective specific heat capacity, surrounding rock thermal conductivity, and oxygen diffusion coefficient. Based on this, a coupling equation is constructed, in which the sensible heat exchange term includes geothermal heat released from the surrounding rock at 42℃ (Q_geothermal) and oxidation heat release from the TBM hydraulic system (Q_oxidation), the latent heat exchange term considers continuous moisture dissipation from the surrounding rock and spray evaporation, and the oxygen diffusion term takes into account the inhibitory effect of high humidity on molecular diffusion. Through finite element analysis, the temperature field (showing local hot spots reaching 38℃) and oxygen concentration field (showing a low-oxygen zone of 18.7% on the return air side) in the 3m area in front of the tunneling face are obtained. This data was used to predict that while increasing the oxygen supply airflow could raise the concentration in low-oxygen areas, it would also cause the hot spot temperature to rise to 40°C due to increased Q_oxidation. Conversely, increasing cooling might alter the diffusion coefficient due to excessive dehumidification, thus reducing oxygen delivery efficiency. Based on this, the adaptive decoupling controller selects a strategy of moderate oxygenation combined with intermittent cooling to achieve a balance between safety and energy efficiency.

[0117] In one embodiment, a dynamic prediction method is used to predict the degree of functional coupling in future time periods, including:

[0118] Within each prediction step, the interaction coefficient between oxygen supply flow rate and cooling load is calculated based on the coupled thermodynamic model, and a coupling coefficient matrix is ​​generated.

[0119] The dynamic prediction method can be a prediction strategy that is executed on a rolling time series basis and uses a model to extrapolate future states in multiple steps. It can be used to support the updating and evaluation of the degree of functional coupling within each prediction step. The prediction step can be the basic time unit used to discretize the time axis during the dynamic prediction process, and can be used to define the time granularity of each coupling coefficient calculation and cost function evaluation. Oxygen supply flow rate can be the volume of oxygen-containing gas delivered to the work area per unit time, and can be used to directly affect the oxygen concentration distribution and local heat load level. Cooling load can be the total amount of heat removed to maintain the target temperature, and can be used to reflect the intensity of the current thermal and humid environment's demand on the refrigeration system. The mutual influence coefficient can be a sensitivity parameter that quantifies the impact of changes in oxygen supply flow rate on cooling load, or changes in cooling load on oxygen supply efficiency, and can be used to characterize the physical coupling strength between the oxygen supply and cooling subsystems. In this embodiment, the mutual influence coefficient can be obtained by identifying the input-output partial derivatives or disturbance response through a coupled thermodynamic model. The coupling coefficient matrix can be a structured matrix composed of multiple mutual influence coefficients, used to describe the coupling relationship between multiple variables, and can be used to provide structured input to the coupling cost function, supporting the systematic evaluation of functional coupling. In this embodiment, the coupling coefficient matrix can be obtained by organizing the mutual influence coefficients under different operating conditions into a matrix form according to the time or state dimension within each prediction step.

[0120] Within each prediction step, the interaction coefficient between oxygen supply flow and cooling load is calculated based on a coupled thermodynamic model. This can be achieved by calling the coupled thermodynamic model and numerically differentiating or simulating the rate of change in cooling load caused by oxygen supply flow disturbances (or vice versa). Furthermore, this operation can be implemented by calculating local sensitivity coefficients using the finite difference method and extracting Jacobian matrix elements through model linearization, thus obtaining a quantitative characterization of the dynamic interaction between oxygen supply and cooling functions. Generating the coupling coefficient matrix can be achieved by arranging the interaction coefficients calculated within each prediction step into a matrix structure according to time or state dimensions. Further, this operation can be implemented by constructing a block diagonal coupling matrix to preserve temporal independence and using sliding window aggregation to generate a low-dimensional feature matrix, thus providing a structured data foundation for coupling assessment under multi-step prediction.

[0121] Based on the coupling coefficient matrix, the degree of functional coupling is quantitatively evaluated using a coupling cost function, the specific evaluation formula of which is as follows:

[0122] C=Σ[γ·(Oa-Os)²+δ·(Ta-Ts)²+ε·(Qo·Qc)]

[0123] Where C is the total functional coupling quantification value in the future prediction time domain, Oa and Os represent the actual and set oxygen concentrations respectively, Ta and Ts represent the actual and set temperatures respectively, Qo and Qc are the energy consumption of the oxygen supply system and the cooling system respectively, and γ, δ and ε are preset weighting coefficients used to balance the importance of different indicators.

[0124] The coupling cost function can be a mathematical expression used to comprehensively evaluate the degree of functional coupling, including environmental deviation terms and energy consumption coupling terms. It can be used to transform multidimensional coupling effects into optimizable scalar indicators to guide decoupling control decisions. In this embodiment, the coupling cost function can be obtained by integrating oxygen concentration error, temperature error, and the product of oxygen supply and cooling energy consumption based on control target design. The functional coupling quantification value can be the cumulative output value of the coupling cost function in the prediction time domain, characterizing the overall coupling severity. It can be used as a core component of the prediction results to measure the risk of control conflict. The actual oxygen concentration (Oa) can be the oxygen volume fraction measured in real time in the breathing zone of the work area, which can be used to reflect the current oxygen supply effectiveness and serve as the tracking error term in the cost function. The set oxygen concentration (Os) can be a target oxygen concentration value preset according to safety regulations and operational requirements, which can be used as a benchmark reference for oxygen control and to calculate the deviation term. For example, the set oxygen concentration can be set using a safety threshold, a comfort optimization, or a dynamically adjusted set concentration.

[0125] Actual temperature (Ta) can be the measured air temperature of the work area, reflecting the current thermal environment and used for temperature tracking performance evaluation. Further, actual temperature can include dry-bulb temperature, wet-bulb temperature, black-bulb temperature, etc. Setpoint temperature (Ts) can be a target temperature value preset according to thermal comfort and equipment operating requirements, used as a reference for cooling control and for calculating temperature deviation. In an exemplary embodiment, the setpoint temperature can be a constant setpoint temperature, a segmented setpoint temperature, an adaptive setpoint temperature, etc. Oxygen supply system energy consumption (Qo) can be the energy consumption of oxygen supply equipment (such as fans, oxygen enrichment devices) within a unit forecast period, used in the calculation of energy consumption coupling terms, reflecting the resource cost of oxygen supply operation. Exemplarily, oxygen supply system energy consumption can include electrical energy consumption, compression power consumption, auxiliary system energy consumption, etc. Cooling system energy consumption (Qc) can be the energy consumption of cooling equipment (such as refrigeration units, spray systems) within a unit forecast period, used in the calculation of energy consumption coupling terms, reflecting the resource cost of cooling operation. Furthermore, the energy consumption of the cooling system may include refrigerant cycle energy consumption, water pump power consumption, dehumidification and reheat energy consumption, etc. The preset weighting coefficients (γ, δ, ε) can be configurable parameters used to adjust the relative importance of each term in the coupling cost function, and can be used to achieve a flexible trade-off between oxygen safety, thermal comfort, and system energy efficiency. In a specific embodiment, the preset weighting coefficients may be fixed weighting coefficients, operating condition adaptive weighting coefficients, multi-objective Pareto weighting coefficients, etc.

[0126] Based on the coupling coefficient matrix, the degree of functional coupling can be quantitatively evaluated through a coupling cost function. This can be achieved by mapping the coupling coefficient matrix to environmental states and energy consumption variables, and then substituting them into a formula to calculate the total coupling cost. Furthermore, this operation can be implemented by calculating the rolling cost online in real time and estimating the expected cost using Monte Carlo sampling, thereby transforming the physical coupling relationship into an optimizable scalar performance index.

[0127] The functional coupling degree after quantitative evaluation is updated to the prediction results.

[0128] Updating the quantitative assessment of functional coupling to the prediction results can be achieved by storing or transmitting the calculated quantitative value of functional coupling as a key field of the prediction results, thereby ensuring that the subsequent controller receives complete prediction information containing coupling risk information.

[0129] For example, in the scenario of regulating the thermal and humid environment of a deep, high-humidity, and high-heat TBM operating area, the functional decoupling regulation method for the thermal and humid environment of the deep mine TBM tunneling face in this embodiment can be as follows: In a certain deep mine operating area, the system performs a dynamic prediction every 30 seconds. Within the current prediction step, the coupled thermodynamic model calculates that: for every 10% increase in oxygen supply flow, the cooling load increases by 8% (mutual influence coefficient is 0.8); at the same time, cooling causes the air dew point to decrease, reducing oxygen diffusion efficiency, with a reverse influence coefficient of -0.5. These coefficients constitute a 2×2 coupling coefficient matrix. Substituting into the coupling cost function, where γ = 1.2 (emphasizing oxygen safety), δ = 1.0 (standard thermal comfort), and ε = 0.6 (suppressing energy consumption coupling), C = 42.3 is calculated. This value is higher than the threshold of 35, indicating a significant coupling conflict. After the prediction results are updated, the adaptive decoupling controller reduces the simple increase of airflow and oxygen supply, and instead adopts a strategy of local oxygen enrichment + intermittent cooling to avoid a surge in Qo·Qc terms. Ultimately, while ensuring Oa≥19.5% and Ta≤33℃, the total energy consumption of the system is reduced.

[0130] In one embodiment, the energy consumption distribution between the oxygen supply system and the cooling system is dynamically adjusted by an adaptive decoupling controller, including:

[0131] The prediction results are input into the expert system controller to generate preliminary control parameters.

[0132] The expert system controller can be an inference-based control module built on domain expert experience rules, used to map input states to preliminary control strategies. In this embodiment, the expert system controller can be used to quickly generate preliminary control parameters that conform to engineering experience, ensuring the rationality and interpretability of the initial control. Furthermore, the expert system controller can match and derive prediction results through a knowledge base (containing IF-THEN rules) and the inference engine. The preliminary control parameters can be an unoptimized set of initial control parameters generated by the expert system controller based on the prediction results. For example, the preliminary control parameters can be used as the initial feasible solution for the model predictive control algorithm, shortening the optimization convergence time. Inputting the prediction results into the expert system controller to generate preliminary control parameters can be achieved by matching the prediction results of functional coupling with the expert rule base, and then outputting the preliminary control parameters through the inference engine. Further, this operation can be implemented by using forward chain inference to generate parameters or by using fuzzy rule matching to generate parameters, thereby quickly obtaining an engineering-feasible initial control strategy and improving the system response speed.

[0133] A model predictive control algorithm is used to perform multi-objective optimization on the initial control parameters to generate the adjusted control parameters.

[0134] Model predictive control (MDC) can be an advanced control algorithm that solves multi-objective optimization problems in the rolling time domain based on a system dynamic model. In this embodiment, MDC can be used to refine the initial control parameters, taking into account multiple objectives such as thermal and humidity comfort, oxygen supply, and minimum energy consumption. Furthermore, MDC can use a coupled thermodynamic model to predict the future behavior of the system and minimize the objective function under constraints. Multi-objective optimization can be a mathematical process of seeking Pareto optimal or weighted optimal solutions under the condition of multiple conflicting or synergistic objective functions. In a specific embodiment, multi-objective optimization can be used to coordinate the contradictory relationship between thermal and humidity control accuracy, oxygen supply compliance rate, and total system energy consumption.

[0135] The adjusted control parameters can be the final set of control parameters optimized by the model predictive control algorithm. In this embodiment, the adjusted control parameters can be used as the basis for calculating the energy consumption allocation ratio, achieving optimal energy efficiency operation under functional decoupling. The model predictive control algorithm is used to perform multi-objective optimization on the initial control parameters to generate the adjusted control parameters. This can be achieved by using the initial control parameters as initial values ​​and solving an optimization problem including heat and humidity, oxygen, and energy consumption objectives under the constraints of a coupled thermodynamic model. Furthermore, this operation can be implemented by using sequential quadratic programming to solve a nonlinear MPC problem or by employing distributed optimization to process the high-dimensional parameter space, enabling optimal energy consumption while ensuring environmental safety and overcoming the limitations of pure rule-based control.

[0136] Based on the adjusted control parameters, the energy consumption distribution ratio between the oxygen supply system and the cooling system is calculated, and the control parameters are generated.

[0137] The energy consumption allocation ratio can be the relative weight or allocation coefficient of the oxygen supply system and the cooling system in the total energy consumption. In this embodiment, the energy consumption allocation ratio can be used to quantify the energy coordination relationship between the oxygen supply and cooling functions, supporting the execution of decoupled control. Furthermore, the energy consumption allocation ratio can be calculated based on the power demand of each subsystem in the adjusted control parameters. Based on the adjusted control parameters, the energy consumption allocation ratio between the oxygen supply system and the cooling system is calculated, and control parameters are generated. This can be achieved by parsing the power commands of the oxygen supply and cooling equipment in the adjusted control parameters, calculating their proportion of the total energy consumption, and standardizing the output, thereby transforming the optimization result into an executable energy consumption coordination signal to support the implementation of decoupled control.

[0138] Taking a sudden heat source event during deep, high-humidity, and high-temperature tunnel excavation as an example, the functional decoupling control method for the thermal and humid environment of the deep mine TBM excavation face in this embodiment can be as follows: During the excavation of a deep roadway, a fault zone is encountered, and the heat dissipation of the surrounding rock increases sharply. The prediction results show that the functional coupling degree will increase sharply within the next 10 minutes. The system inputs the prediction results into the expert system controller, and based on the "high temperature and high humidity + high coupling" rule base, it quickly generates preliminary control parameters: increase the oxygen supply air volume by 10% and start full-power cooling. Subsequently, the model predictive control algorithm uses these parameters as initial values, comprehensively considers the personnel's oxygen demand (≥19.5%), the upper limit of temperature (≤33℃), the upper limit of humidity (≤85%RH), and the minimum total energy consumption, and optimizes to obtain the adjusted control parameters: only increase the oxygen concentration without increasing the air volume, and adopt intermittent spray mode for cooling. Based on this, the energy consumption ratio of the oxygen supply system is calculated to be 45% and that of the cooling system is 55%. The final control parameters are generated and the equipment control signal is output, which avoids the additional heat load caused by the increase in air volume and prevents excessive dehumidification from affecting oxygen diffusion, achieving both safety and energy efficiency.

[0139] In one embodiment, after generating the control parameters, the process further includes:

[0140] When an abrupt change in the environmental grade result of heat and humidity is detected, a buffer regulation mechanism is activated to smooth the change in regulation parameters.

[0141] The response speed of the cooling equipment control signal is adjusted by a feedforward compensation algorithm to counteract the coupling with pressure fluctuations generated by the oxygen supply system.

[0142] Based on real-time energy efficiency ratio and environmental comfort indicators, online optimization and control parameters are performed.

[0143] The environmental level result of the sudden change in thermal humidity level can be an environmental state identification output of a significant jump in thermal humidity level within a short period of time. This can be used as a criterion to trigger a buffer control mechanism to cope with sudden working conditions such as water inrush in the surrounding rock and a surge in rock breaking heat. In this embodiment, the environmental level result of the sudden change in thermal humidity level can be one or more of the following: a step change in humidity caused by water seepage in the surrounding rock, a temperature change caused by a sudden increase in frictional heat of the tunnel boring machine cutterhead, or a combined thermal and humidity shock after ventilation interruption and recovery. The buffer control mechanism can be a dynamic adjustment module that applies a smooth transition strategy to the control parameters when a sudden change in environmental level is detected, in order to suppress drastic changes in the control signal. This can be used to avoid start-stop shocks or environmental parameter oscillations caused by step commands, thereby improving the stability of system operation. For example, the buffer control mechanism can smooth the original control parameters by introducing a first-order inertial element, a rate limiter, or a sliding mode filter. The change in the control parameters can be the dynamic adjustment trajectory of the control parameters in the time dimension, which can be used to reflect the response characteristics of the control system to environmental disturbances. Its smoothness directly affects the equipment lifespan and environmental stability.

[0144] Identifying abrupt changes in environmental temperature and humidity levels can be achieved by monitoring a continuous sequence of environmental temperature and humidity levels. A change is defined as a jump in temperature and humidity levels exceeding a preset threshold between adjacent time points. Furthermore, identifying abrupt changes in environmental temperature and humidity levels can be achieved by jointly setting a threshold for the difference in temperature and humidity levels and a time window, thereby triggering a buffer control mechanism to prevent the control strategy from overreacting to sudden disturbances. Activating the buffer control mechanism to smooth changes in control parameters can be achieved by inserting a smoothing filter module before the output of the control parameters to limit their rate of change or amplitude. Further, activating the buffer control mechanism to smooth changes in control parameters can be achieved by using a first-order low-pass filter to smooth the control parameters or by setting a maximum rate of change limiter to constrain the parameter adjustment step size, thereby suppressing step jumps in the control signal and avoiding mechanical shocks to equipment and environmental vibrations.

[0145] Feedforward compensation algorithms can be control algorithms that generate compensation amounts in advance based on oxygen supply system disturbance signals to adjust the response characteristics of cooling equipment. They can be used to actively counteract the coupling interference of airflow pressure fluctuations caused by the oxygen supply system on the cooling equipment, reducing response lag. In one specific embodiment, the feedforward compensation algorithm can calculate the feedforward compensation amount in real time by establishing a transfer function model between changes in oxygen supply volume and pressure disturbances in the cooling system. For example, feedforward compensation algorithms can include model-based feedforward compensation algorithms, data-driven feedforward compensation algorithms, or adaptive feedforward compensation algorithms. The response speed of the cooling equipment control signal can be the dynamic following rate of the cooling equipment control signal to changes in input commands. This can be used to determine the cooling system's ability to track changes in heat load, affecting the decoupling control accuracy. Furthermore, the response speed of the cooling equipment control signal can include fast response mode, standard response mode, or hysteresis response mode. Pressure fluctuation coupling generated by the oxygen supply system can be a dynamic coupling phenomenon where sudden changes in oxygen supply volume cause transient changes in airflow pressure within the roadway, thereby interfering with the normal operation of the cooling equipment. This can constitute another type of physical coupling path between the oxygen supply and cooling functions, besides heat and mass exchange, requiring specialized decoupling processing. In one exemplary embodiment, the pressure fluctuation coupling generated by the oxygen supply system may include pressure pulse coupling caused by the start and stop of the fan, steady-state offset coupling caused by the operation of the air volume regulating valve, or eddy current disturbance coupling caused by the misalignment of multiple fans.

[0146] Adjusting the response speed of the cooling equipment control signal through a feedforward compensation algorithm can be achieved by calculating the required cooling response compensation amount in real time based on the rate of change of the oxygen supply demand signal and superimposing it onto the original control signal. Furthermore, adjusting the response speed of the cooling equipment control signal through the feedforward compensation algorithm can be achieved by generating a feedforward compensation term based on the differential signal of the oxygen supply airflow or by using a neural network to map oxygen supply disturbances to the cooling compensation amount. This can proactively offset the interference of oxygen supply system pressure fluctuations on the cooling system, improving the accuracy of dynamic decoupling. Offsetting the coupling with pressure fluctuations generated by the oxygen supply system can be achieved by applying the feedforward compensation amount to the cooling equipment control signal, enabling its output to actively counteract airflow pressure disturbances. Furthermore, offsetting the coupling with pressure fluctuations generated by the oxygen supply system can be achieved by dynamically correcting the cooling spray flow rate or fan speed with the feedforward compensation amount. This can overcome the lag limitations of traditional feedback control and achieve dynamic decoupling between the oxygen supply and cooling subsystems.

[0147] Real-time energy efficiency ratio (EER) can be the ratio of the effective environmental control effect (such as cooling or oxygen supply) output by the system per unit time to the energy consumed. It can be used as an energy efficiency constraint index for online optimization and control parameters, guiding the system to converge towards a high-efficiency operating point. In a specific embodiment, the real-time EER can be calculated by real-time monitoring of cooling power, fan power consumption, temperature drop, and oxygen concentration increase. For example, the real-time EER may include cooling EER, oxygen supply EER, or comprehensive environmental control EER. Environmental comfort index can be a multi-dimensional evaluation parameter that quantifies the thermal and humidity comfort of workers, integrating factors such as temperature, humidity, wind speed, and oxygen concentration. It can be used as a human-centered performance constraint for online optimization and control parameters, ensuring work safety and health. Furthermore, the environmental comfort index can be calculated in real time based on the PMV-PPD model or ASHRAE standard derived formula, combined with measured environmental parameters. In an exemplary embodiment, the environmental comfort index may include thermal comfort index, air quality comfort index, or comprehensive physiological load index.

[0148] Online optimization control parameters can be dynamically adjusted based on real-time energy efficiency ratio and environmental comfort indicators during system operation. They can be used to achieve a multi-objective balance between energy efficiency and comfort, improving the system's adaptability under disturbance conditions. For example, online optimization control parameters can include energy efficiency-priority control parameters, comfort-priority control parameters, or balanced control parameters. Based on real-time energy efficiency ratio and environmental comfort indicators, online optimization control parameters can be constructed by constructing an optimization problem that includes constraints on maximizing energy efficiency ratio and achieving comfort standards, and solving for the optimal control parameters in real time. Furthermore, based on real-time energy efficiency ratio and environmental comfort indicators, online optimization control parameters can be achieved by using rolling time-domain optimization to update control parameters online or by using a multi-objective particle swarm optimization algorithm to search for Pareto optimal solutions, thus ensuring both optimal energy efficiency and personnel comfort and safety even under abrupt changes in boundary conditions.

[0149] For example, in the scenario where a tunnel boring machine (TBM) traverses a high-temperature fault zone, the functional decoupling control method for the thermal and humid environment of the deep mine TBM tunneling face in this embodiment can be as follows: when the TBM tunnels to a deep fault zone, the surrounding rock suddenly releases a large amount of hot water vapor, causing the relative humidity of the tunneling face to rise from 85% to 99% within 30 seconds, and the thermal and humid level to jump from moderate to severe. Upon recognizing the sudden change, the system immediately activated a buffer control mechanism, changing the planned step increase in cooling power to a linear increase within 5 minutes to avoid compressor overload. Simultaneously, the feedforward compensation algorithm detected that the oxygen supply fan automatically increased its speed by 15% due to the rise in humidity, and accordingly increased the cooling spray flow rate in advance to offset the resulting local negative pressure disturbance. During this process, the system continuously calculated the real-time energy efficiency ratio (currently 2.1kW / ton) and environmental comfort index (PMV = +2.8), and through online optimization, fine-tuned the supply air temperature to 18℃ and increased the oxygen concentration to 20.5%. Ultimately, while maintaining the thermal comfort of personnel (PMV decreased to +1.2), the energy efficiency ratio was stabilized above 2.3, avoiding condensation and frequent equipment start-ups and shutdowns.

[0150] In one embodiment, the output of oxygen supply equipment control signals and cooling equipment control signals includes:

[0151] The control parameters are combined with environmental target indicators and converted into target operating parameters for oxygen supply fans and refrigeration units;

[0152] The oxygen supply equipment control signal is generated based on the target operating parameters through a PID closed-loop controller.

[0153] The cooling equipment control signal is generated based on the target operating parameters using a fuzzy controller.

[0154] The environmental target indicators can be a set of setpoints characterizing the desired state of heat, humidity, and oxygen supply control at the tunnel face, including target temperature, humidity, and oxygen concentration, serving as a benchmark for converting control parameters into target operating parameters for the equipment. In this embodiment, the environmental target indicators are preset by the upper-level control strategy or safety specifications and dynamically adjusted in conjunction with the current operating conditions. The oxygen supply fan can be a ventilation device used to deliver fresh air or oxygen-enriched gas to the TBM tunnel face, enabling oxygen supply and local airflow organization control. In an exemplary embodiment, the oxygen supply fan can be a variable frequency axial flow fan, a centrifugal fan, or a local booster fan. The refrigeration unit can be a cooling equipment system used to reduce the air temperature and control humidity at the tunnel face, undertaking the main functions of heat load reduction and dehumidification. The target operating parameters can be the specific operating setpoints required by the oxygen supply fan and refrigeration unit to achieve the environmental targets, serving as the setting input for the lower-level controller to guide precise equipment operation. In a specific embodiment, the target operating parameters are generated by converting control parameters into environmental target indicators through a mapping function or rule base. For example, target operating parameters may include fan speed setpoint, cooling power setpoint, and supply air temperature and humidity setpoint.

[0155] The control parameters are combined with environmental target indicators to convert them into target operating parameters for the oxygen supply fan and refrigeration unit. This can be achieved by mapping the energy consumption allocation ratio in the control parameters to the setpoints in the environmental target indicators, or by optimizing the calculation to generate specific equipment operating setpoints. Furthermore, this operation can be achieved by using a lookup table combined with interpolation to generate target operating parameters, or by solving for the optimal operating point that satisfies multiple objective constraints through quadratic programming, thus enabling precise conversion of high-level control strategies to low-level equipment commands. The PID closed-loop controller can be a linear control unit based on a proportional-integral-derivative algorithm to achieve error feedback regulation. It can be used to implement fast and stable tracking control of the oxygen supply fan, ensuring the accuracy of the oxygen supply response. In this embodiment, the PID closed-loop controller compares the target operating parameters with the actual feedback values ​​of the equipment in real time, calculating the control output to minimize the deviation. The PID closed-loop controller generates control signals for the oxygen supply equipment based on the target operating parameters. This can be done by using the target operating parameters as setpoints, collecting actual operating feedback from the oxygen supply fan, and calculating the output control signal using the PID algorithm. Furthermore, this operation can be achieved by using a digital PID algorithm to output a PWM signal to drive the frequency converter, or by introducing feedforward compensation to enhance the ability to suppress disturbances, thereby ensuring that the oxygen supply fan can quickly and stably track the target air volume or pressure and maintain the reliability of oxygen supply.

[0156] A fuzzy controller is an intelligent control unit that processes uncertain or nonlinear inputs based on fuzzy logic rules. It can be used to address nonlinear, time-varying, and boundary fuzzy issues in the refrigeration process, improving the robustness of the cooling system. In one exemplary embodiment, the fuzzy controller fuzzifies the deviation and rate of change between the target operating parameters and the actual state, and then defuzzifies and outputs the control quantity after rule-based reasoning. The fuzzy controller generates a cooling equipment control signal based on the target operating parameters. This can be achieved by using the deviation and rate of change between the target operating parameters and the actual state of the refrigeration unit as input, and then outputting the control signal through fuzzy rule reasoning. Furthermore, this operation can be implemented by adjusting the compressor's start-up and shutdown frequency based on a fuzzy rule base built from expert experience, or by dynamically updating the membership function and rule weights using an online learning mechanism. This effectively handles nonlinearity, hysteresis, and uncertainty in the cooling process, avoiding secondary risks such as overcooling and condensation.

[0157] For example, in the scenario of coordinated control of cooling and oxygen supply at a deep, high-humidity TBM tunneling face, the functional decoupling control method for the thermal and humid environment of a deep mining TBM tunneling face in this embodiment can be as follows: In a deep tunnel excavation, the upper-level decoupling controller outputs control parameters: oxygen supply energy consumption accounts for 40%, and cooling energy consumption accounts for 60%. Combined with environmental target indicators (temperature ≤32℃, humidity ≤85%, O2 ≥19.5%), these are converted into target operating parameters: oxygen supply fan speed 1450rpm, and refrigeration unit cooling power 35kW. The PID closed-loop controller uses 1450rpm as the target and adjusts the inverter output in real time to keep the fan speed stable within ±10rpm error; the fuzzy controller dynamically adjusts the compressor frequency and spray valve opening based on the deviation and trend between the current return air temperature and the target temperature drop corresponding to 35kW, avoiding excessive dehumidification and condensation caused by sudden water inrush from the surrounding rock. The coordinated execution of both ensures oxygen supply stability and improves the environmental adaptability of the cooling system.

[0158] In one embodiment, after dynamically adjusting the energy consumption distribution between the oxygen supply system and the cooling system through an adaptive decoupling controller to generate control parameters, an online learning and optimization step is also included:

[0159] Collect historical operational data, including environmental status data, control parameters, and system energy efficiency indicators;

[0160] By using reinforcement learning algorithms, the decision-making strategy of the adaptive decoupling controller is adjusted based on historical operating data;

[0161] By using a transfer learning model, control parameters are optimized under new conditions based on historical data from similar operating conditions, thereby improving the environmental adaptability of the adaptive decoupling controller.

[0162] Historical operational data can be a multi-dimensional time-series data set recorded by the system during past operations, including environmental states, control actions, and performance feedback. This data can be used to provide training samples for reinforcement learning and transfer learning, supporting controller strategy optimization. In this embodiment, historical operational data can be continuously recorded by the data acquisition module, showing the environmental response, equipment parameters, and energy efficiency indicators after control execution, and stored in a local or cloud database. Environmental state data can be a set of real-time or historical state variables reflecting the heat, humidity, and air quality conditions at the tunnel face. This data can be used as a core component of historical operational data to evaluate control effectiveness and learn the state-action mapping relationship. System energy efficiency indicators can be quantitative evaluation parameters measuring the comprehensive energy utilization efficiency of the oxygen supply and cooling system. This data can be used as a reward signal for reinforcement learning, guiding the controller to optimize towards higher energy efficiency. In an exemplary embodiment, system energy efficiency indicators can be calculated based on the real-time power of the oxygen supply and cooling equipment, the degree of environmental improvement, and the task completion rate.

[0163] Collecting historical operational data, including environmental state data, control parameters, and system energy efficiency indicators, can be achieved by synchronously recording environmental responses, control commands, and energy efficiency feedback after each control cycle and storing them in a historical database. This operation builds the data foundation required for closed-loop learning, supporting subsequent strategy optimization. Reinforcement learning algorithms, which are machine learning methods that interact with the environment and maximize cumulative rewards through trial and error, can be used to achieve online autonomous optimization of the decision-making strategy of the adaptive decoupling controller. In one specific embodiment, the reinforcement learning algorithm can use the environmental state as input, control parameters as actions, and system energy efficiency indicators as rewards to iteratively update the policy network or Q-value function. The decision-making strategy can be a mapping rule or function by which the adaptive decoupling controller selects the optimal control parameters based on the current state. This can be used to determine the coordinated control behavior of the oxygen supply and cooling systems, directly affecting system energy efficiency and environmental stability. In this embodiment, the decision-making strategy can be continuously updated by the reinforcement learning algorithm using historical operational data to update its parameters or structure. Adjusting the decision-making strategy of the adaptive decoupling controller based on historical operational data using reinforcement learning algorithms can be achieved by using historical operational data as a training set and iteratively optimizing the controller's internal strategy through a policy gradient or value function update mechanism. Furthermore, this operation can be achieved by using offline batch reinforcement learning to update the policy network and using an online experience replay mechanism for incremental learning, thereby enabling the controller to continuously approach the optimal control strategy for energy efficiency and environmental stability during long-term operation.

[0164] Transfer learning models are machine learning frameworks that apply knowledge learned in a source domain (similar operating conditions) to a target domain (new environment) to accelerate controller convergence in a new environment and avoid cold-start performance degradation. For example, transfer learning models can transfer effective control experience from historical operating conditions to new scenarios through feature alignment, parameter sharing, or knowledge distillation. Furthermore, transfer learning models can employ feature representation-based transfer models, parameter fine-tuning-based transfer models, and meta-learning-based transfer models. Historical data on similar operating conditions can be subsets of historical operational data comparable to the current new environment in dimensions such as geological conditions, heat source characteristics, or personnel configuration. This data can be used as source domain knowledge for transfer learning to initialize or correct control parameters in the new environment. In one specific embodiment, historical data on similar operating conditions can be obtained by filtering from historical operational data through operating condition clustering, similarity metrics, or label matching.

[0165] Control parameters in a new environment can be initial or optimized control parameters generated for a tunneling environment that has not been fully experienced or is appearing for the first time. These parameters can be used to ensure that the system quickly enters a state of efficient and stable operation under new conditions. By using a transfer learning model, control parameters in a new environment can be optimized based on historical data from similar operating conditions. This can involve identifying historical operating conditions similar to the current new environment, extracting their effective control patterns, and transferring them to the current controller's initialization or calibration phase. Furthermore, this operation can be achieved by using domain-adaptive methods to align the feature distributions of old and new operating conditions and by fine-tuning parameters based on a policy network of similar operating conditions. This can significantly shorten the controller's learning convergence time under new conditions and improve initial control performance. Improving the environmental adaptability of the adaptive decoupled controller can be achieved through the synergistic effect of reinforcement learning and transfer learning, enhancing the controller's response to unknown or dynamically changing operating conditions. This allows the control strategy to transform from static rules to an autonomously evolving intelligent agent, improving system robustness.

[0166] For example, in the scenario where a tunnel boring machine (TBM) traverses a high-temperature fault zone, the functional decoupling control method for the thermal and humid environment of the deep mine TBM excavation face in this embodiment can be as follows: A deep mine TBM excavation face suddenly enters an unknown high-temperature fault, the surrounding rock temperature rises sharply to 48°C, humidity becomes saturated, and the number of personnel temporarily increases to 8. Due to the lack of prior data for this working condition, the initial control performance of the system decreases. At this time, the online learning module is immediately activated: First, it retrieves three similar high-temperature and high-humidity working conditions from the historical database (such as previous crossing of another fault, peak summer operation, etc.), and uses a transfer learning model to transfer the effective control parameters of these working conditions (such as low wind speed and high concentration oxygen supply + intermittent forced cooling) to the current controller, generating initial control parameters for the new environment; At the same time, the system begins to collect environmental status data, control parameters, and energy efficiency indicators for this operation, forming new historical operating data; The reinforcement learning algorithm uses this data to continuously optimize the decision-making strategy, improving the system's energy efficiency index by 12% within 48 hours, stabilizing the thermal and humid environment at 33°C / 88%RH, and maintaining the oxygen concentration at 20.1%. The entire process requires no human intervention, enabling a rapid transition from "cold start" to "efficient operation".

[0167] In addition, refer to Figure 2 To achieve the above objectives, the present invention also provides a functional decoupling control system for the thermal and humid environment of a deep mine TBM tunneling face, the system comprising:

[0168] The data fusion module 10 is used to acquire multi-source environmental parameter data of the TBM tunneling face in deep mines, and to fuse the multi-source environmental parameter data to generate thermal and humidity state data and air quality data.

[0169] The status recognition module 20 is used to identify the current thermal and humidity level and oxygen demand of the tunnel face based on the thermal and humidity status data and the air quality data, and generate environmental level results and oxygen supply demand signals.

[0170] The coupling prediction module 30 is used to construct an oxygen supply-cooling-environment coupling thermodynamic model based on the environmental level results and the oxygen supply demand signal, and to predict the degree of functional coupling in the future time period using a dynamic prediction method, and generate prediction results.

[0171] The decoupling control module 40 is used to dynamically adjust the energy consumption distribution between the oxygen supply system and the cooling system based on the prediction results, and generate control parameters through an adaptive decoupling controller.

[0172] The instruction output module 50 is used to output control signals for the oxygen supply equipment and control signals for the cooling equipment according to the control parameters, so as to decouple the oxygen supply and cooling functions.

[0173] Other embodiments or specific implementations of the functional decoupling control system for the thermal and humid environment of the deep mine TBM tunneling face described in this invention can be referred to the above-mentioned method embodiments, and will not be repeated here.

Claims

1. A method for functional decoupling and control of the thermal and humid environment at the TBM tunneling face in deep mines, characterized in that, The method includes: Acquire multi-source environmental parameter data of the TBM tunneling face in deep mines, and fuse the multi-source environmental parameter data to generate thermal and humidity state data and air quality data; Based on the thermal and humidity data and the air quality data, the current thermal and humidity level and oxygen demand of the tunnel face are identified, and environmental level results and oxygen supply demand signals are generated. Based on the environmental level results and the oxygen demand signal, a coupled thermodynamic model of oxygen supply-cooling-environment is constructed, and a dynamic prediction method is used to predict the degree of functional coupling in the future time period, generating prediction results. Based on the prediction results, an adaptive decoupling controller is used to dynamically adjust the energy consumption distribution between the oxygen supply system and the cooling system to generate control parameters. Based on the aforementioned control parameters, control signals for the oxygen supply equipment and cooling equipment are output to decouple the oxygen supply and cooling functions.

2. The method for functional decoupling and control of the thermal and humid environment at the deep mine TBM tunneling face as described in claim 1, characterized in that, The generation of thermal and humidity state data and air quality data includes: Acquire data from temperature sensor, humidity sensor, oxygen concentration sensor, and dust concentration sensor to form a raw set of environmental parameters; The temperature sensor data is the real-time temperature value of the tunnel face, the humidity sensor data is the percentage of water vapor content in the air, the oxygen concentration sensor data is the oxygen volume percentage, and the dust concentration sensor data is the mass concentration of suspended particulate matter. The original set of environmental parameters is preprocessed with noise reduction and drift correction to generate preprocessed environmental data; The preprocessed environmental data is input into a multimodal fusion algorithm, which outputs the thermal and humidity state data and the air quality data.

3. The method for functional decoupling and control of the thermal and humid environment at the deep mine TBM tunneling face as described in claim 1, characterized in that, The process of identifying the current thermal and humidity levels and oxygen requirements at the tunnel face, and generating environmental level results and oxygen supply demand signals, includes: Dry-bulb temperature and relative humidity features are extracted from the thermal and humidity state data to generate a thermal and humidity feature vector. Oxygen partial pressure features and dust concentration features are extracted from the air quality data to generate an air quality feature vector. The thermal and humidity feature vector and the air quality feature vector are input into the environmental assessment model, and the environmental level result and the oxygen supply demand signal are output. The evaluation process of the environmental assessment model can be represented by the following function: (E_level, O_signal) = F_evaluator ([V_thermo, V_air]) Where (E_level, O_signal) represents the output environmental level result and oxygen demand signal, F_evaluator represents the trained environmental assessment model function, V_thermo represents the thermal and humidity feature vector, V_air represents the air quality feature vector, and [V_thermo, V_air] represents the concatenation operation of the two feature vectors.

4. The method for functional decoupling and control of the thermal and humid environment at the deep mine TBM tunneling face as described in claim 1, characterized in that, The construction of the oxygen supply-cooling-environment coupled thermodynamic model includes: Based on the environmental level results and the oxygen supply demand signal, a set of thermodynamic parameters are matched and selected from the parameter library; Based on the aforementioned thermodynamic parameters, a coupled equation is established, comprising sensible heat exchange, latent heat exchange, and oxygen diffusion terms. This coupled equation is expressed as: ρc∂T / ∂t=k∇²T+Q_geothermal-Q_cooling+Q_oxidation Where ρ is air density, c is specific heat capacity of air, T is temperature field distribution at the tunnel face, k is thermal conductivity coefficient, Q_geothermal is the power released by the geothermal source, Q_cooling is the heat exchange of the cooling system, and Q_oxidation is the oxidation heat release power of the equipment. By solving the coupled equations using the finite element method, temperature field and oxygen concentration field distribution data are generated for subsequent prediction.

5. The method for functional decoupling and control of the thermal and humid environment at the deep mine TBM tunneling face as described in claim 1, characterized in that, The method of predicting the degree of functional coupling in future time periods using dynamic prediction includes: Within each prediction step, the interaction coefficient between oxygen supply flow rate and cooling load is calculated based on the coupled thermodynamic model to generate a coupling coefficient matrix. Based on the coupling coefficient matrix, the degree of functional coupling is quantitatively evaluated using a coupling cost function, the specific evaluation formula of which is as follows: C=Σ[γ·(Oa-Os)²+δ·(Ta-Ts)²+ε·(Qo·Qc)] Where C is the total functional coupling quantification value in the future prediction time domain, Oa and Os represent the actual and set oxygen concentrations respectively, Ta and Ts represent the actual and set temperatures respectively, Qo and Qc are the energy consumption of the oxygen supply system and the cooling system respectively, and γ, δ and ε are preset weighting coefficients used to balance the importance of different indicators. The quantitatively evaluated degree of functional coupling is then updated to the prediction results.

6. The method for functional decoupling and control of the thermal and humid environment at the deep mine TBM tunneling face as described in claim 1, characterized in that, The method of dynamically adjusting the energy consumption distribution between the oxygen supply system and the cooling system through an adaptive decoupling controller includes: The prediction results are input into the expert system controller to generate preliminary control parameters; The preliminary control parameters are optimized using a model predictive control algorithm to generate adjusted control parameters. Based on the adjusted control parameters, the energy consumption distribution ratio between the oxygen supply system and the cooling system is calculated, and the control parameters are generated.

7. The method for functional decoupling and control of the thermal and humid environment at the deep mine TBM tunneling face as described in claim 1, characterized in that, After generating the control parameters, the process also includes: When an abrupt change in the environmental grade result of heat and humidity is detected, a buffer regulation mechanism is activated to smooth the change in the regulation parameter; The response speed of the cooling equipment control signal is adjusted by a feedforward compensation algorithm to counteract the coupling with pressure fluctuations generated by the oxygen supply system. The control parameters are optimized online based on real-time energy efficiency ratio and environmental comfort indicators.

8. The method for functional decoupling and control of the thermal and humid environment at the deep mine TBM tunneling face as described in claim 1, characterized in that, The output oxygen supply equipment control signal and cooling equipment control signal include: The aforementioned control parameters are combined with environmental target indicators and converted into target operating parameters for oxygen supply fans and refrigeration units; The oxygen supply equipment control signal is generated based on the target operating parameters using a PID closed-loop controller. The cooling equipment control signal is generated based on the target operating parameters using a fuzzy controller.

9. The method for functional decoupling and control of the thermal and humid environment at the deep mine TBM tunneling face as described in claim 1, characterized in that, After generating control parameters by dynamically adjusting the energy consumption distribution between the oxygen supply system and the cooling system through an adaptive decoupling controller, the process also includes online learning and optimization steps. Collect historical operating data, including environmental status data, the aforementioned control parameters, and system energy efficiency indicators; The decision-making strategy of the adaptive decoupling controller is adjusted based on the historical operating data using reinforcement learning algorithms. By using a transfer learning model, the control parameters under new conditions are optimized based on historical data from similar operating conditions, thereby improving the environmental adaptability of the adaptive decoupling controller.

10. A functional decoupling control system for the thermal and humid environment of a deep mine TBM tunneling face, characterized in that, The system includes: The data fusion module is used to acquire multi-source environmental parameter data of the TBM tunneling face in deep mines, and to fuse the multi-source environmental parameter data to generate thermal and humidity state data and air quality data. The status recognition module is used to identify the current thermal and humidity level and oxygen demand of the tunnel face based on the thermal and humidity status data and the air quality data, and generate environmental level results and oxygen supply demand signals. The coupling prediction module is used to construct an oxygen supply-cooling-environment coupling thermodynamic model based on the environmental level results and the oxygen supply demand signal, and to predict the degree of functional coupling in the future time period using a dynamic prediction method, thereby generating prediction results. The decoupling control module is used to dynamically adjust the energy consumption distribution between the oxygen supply system and the cooling system based on the prediction results, and generate control parameters through an adaptive decoupling controller. The instruction output module is used to output control signals for the oxygen supply equipment and the cooling equipment according to the control parameters, so as to decouple the oxygen supply and cooling functions.

Images