Real-time adaptive adjustment system and method for wind tunnel boundary layer control

Abstract

The present invention discloses a real-time adaptive adjustment system and method for wind tunnel boundary layer control, relating to the technical field of wind tunnel boundary layer control. The system comprises a data acquisition module that uses multiple sensors to collect data at high frequency and transmits it through an ad hoc network. A data preprocessing module processes the data through specific filtering and dynamic normalization. A boundary layer state assessment module uses deep learning and traditional models to assess and predict boundary layer states. A control decision module integrates hierarchical reinforcement learning with a model predictive control algorithm to determine strategies. Model validation and optimization steps utilize Bayesian fusion assessment, with online model updates. Human-computer interaction and remote monitoring steps utilize virtual reality and augmented reality interaction for remote monitoring and early warning. The present invention improves wind tunnel experiment accuracy, accurately collects and processes data, and precisely controls the boundary layer. It is convenient and safe to operate, features an intuitive interactive interface and remote monitoring, and includes fault diagnosis technology, promoting development in related fields.

Description

Technical Field

[0001] This invention relates to the field of wind tunnel boundary control technology, and in particular to a real-time adaptive adjustment system and method for wind tunnel boundary layer control. Background Technology

[0002] In numerous fields such as aerospace, automotive manufacturing, and civil engineering, wind tunnel experiments are a crucial method for studying aerodynamic characteristics. The wind tunnel boundary layer, as a critical component of the wind tunnel flow field, directly impacts the accuracy and reliability of experimental results. However, current wind tunnel boundary layer control faces many challenges.

[0003] During wind tunnel operation, the internal flow field is influenced by various factors, such as the shape and size of the experimental model, the wind tunnel's own structure, and changes in external environmental conditions. These factors lead to a complex and variable boundary layer state, making it difficult to control stably. Traditional boundary layer control methods are mostly based on fixed parameters and preset rules, lacking the ability to effectively respond to real-time changes in the boundary layer. When the boundary layer becomes unstable or deviates from the ideal state, traditional methods cannot adjust in time, resulting in significant errors in experimental results and failing to meet the requirements of high-precision experiments.

[0004] From a technical implementation perspective, existing boundary layer control technologies have significant shortcomings in data acquisition, processing, and regulation execution. In terms of data acquisition, the precision and layout of sensors make it difficult to comprehensively and accurately acquire various parameters of the boundary layer, resulting in inaccurate assessments of the boundary layer state. Data processing technologies are also relatively simplistic, failing to fully extract potential information from the data and making it difficult to accurately predict boundary layer trends. In the regulation execution stage, the response speed, regulation accuracy, and coordination of the actuators are poor, hindering the rapid and precise implementation of control strategies and making effective boundary layer regulation difficult. Furthermore, the structures and experimental requirements of different wind tunnels vary considerably, and existing control technologies lack versatility, making widespread application in various wind tunnel scenarios difficult. These problems severely restrict the development of wind tunnel experimental technology and hinder in-depth research in aerodynamics in related fields, urgently requiring an innovative wind tunnel boundary layer control technology to solve these challenges. Summary of the Invention

[0005] The present invention proposes a real-time adaptive adjustment system and method for wind tunnel boundary layer control to solve the problems mentioned in the prior art.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: a real-time adaptive adjustment system for wind tunnel boundary layer control, comprising:

[0007] Data acquisition module: Pressure sensors, velocity sensors, and temperature sensors are installed inside the wind tunnel, in the boundary layer region, and in the external environment of the wind tunnel. Humidity sensors and wall shear stress sensors are also installed. The sensors collect data, and the sensor nodes adopt distributed self-organizing network technology to automatically optimize the data transmission path and transmit the data to the data processing center.

[0008] The data preprocessing module employs a filtering algorithm based on wavelet decomposition and adaptive threshold fusion to filter the acquired sensor data. The threshold is adaptively determined based on the statistical characteristics of the data to remove noise interference. Next, the filtered data is normalized using a dynamic normalization method, which dynamically adjusts the normalization interval according to the real-time distribution range of the data. The formula is as follows: Adapting to changes in boundary layer state, where and These are the minimum and maximum values ​​of data that change dynamically over time.

[0009] Boundary layer state assessment module: Introduces a deep learning-based boundary layer state prediction model. Combining preprocessed data and historical data, it predicts the future state of the boundary layer. The model is based on Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) networks, while also incorporating traditional boundary layer theoretical models to comprehensively calculate the boundary layer thickness. Displacement thickness Momentum thickness θ and boundary layer transition location Parameters; the formula for predicting the boundary layer transition location is as follows: ,in Here, Tu is the local Reynolds number, Tu is the turbulence intensity, and f is a nonlinear function obtained by training a deep learning model.

[0010] Control Decision Module: This module employs an algorithm based on a fusion of hierarchical reinforcement learning and Model Predictive Control (MPC) to determine the regulation strategy. The control task is divided into two levels: high-level decision-making and low-level execution. High-level decision-making, based on reinforcement learning, determines the regulation direction and objective according to the boundary layer state and target. Low-level execution uses the MPC algorithm, which, based on high-level decision-making, predicts state changes over a future period and optimizes the regulation action sequence according to the real-time state and dynamic model of the boundary layer. Simultaneously, an adaptive exploration-exploitation strategy is introduced, dynamically adjusting the exploration and exploitation ratio based on the boundary layer state.

[0011] Adjustment Module: The adjustment actuator is driven by a microelectromechanical system (MEMS) and uses micro-nano fabrication technology to manufacture a wall structure with continuously adjustable roughness. The shape and height of the wall microstructure are controlled by external excitation through electric and magnetic fields to achieve continuous adjustment of the wall roughness. The control signal of the actuator adopts coding modulation technology to achieve anti-interference capability of signal transmission.

[0012] Model Validation and Optimization Module: During wind tunnel experiments, the boundary layer state data before and after adjustment are compared in real time with the results of theoretical and predictive models. The Bayesian Model Averaging (BMA) method is used to fuse and evaluate the prediction results of different models. Based on the fusion results, the model weights and parameters are dynamically adjusted. The formula is as follows: ,in This is the prediction result based on data D. It is the i-th model. It is the weight of the i-th model, and ;

[0013] Human-computer interaction and remote monitoring module: Develop a human-computer interaction interface based on virtual reality and augmented reality technologies. Operators can observe the boundary layer flow state and adjustment process inside the wind tunnel through the equipment, and obtain parameters and adjustment commands in real time in the actual wind tunnel scenario. At the same time, the system supports remote monitoring and control functions, and transmits wind tunnel operation data and status information to the remote monitoring center in real time through a wireless network. Operators can remotely formulate adjustment strategies and control the actuators.

[0014] Furthermore, it also includes:

[0015] Environmental perception and compensation module: Real-time monitoring of atmospheric pressure, ambient temperature, wind speed and direction outside the wind tunnel. Based on changes in environmental parameters, fuzzy control algorithm is used to compensate and adjust the boundary layer control strategy. The fuzzy control rules are established based on expert experience and experimental data. According to different combinations of environmental parameters and boundary layer states, the action amplitude and frequency of the regulating actuator are automatically adjusted.

[0016] Furthermore, the data acquisition module employs sensor fault diagnosis and fault tolerance technology. By integrating and analyzing sensor data and detecting residuals, it can determine in real time whether a sensor has malfunctioned. When a sensor fault is detected, a state estimation method based on Kalman filtering is used to reconstruct the data of the faulty sensor, while automatically adjusting the sampling strategy and data fusion algorithm of each sensor.

[0017] Furthermore, the control decision module introduces a multi-objective optimization mechanism. In addition to optimizing the stability and uniformity of the boundary layer, it also considers the energy consumption, adjustment time, and impact on wind tunnel experimental parameters during the adjustment process. The non-dominated sorting genetic algorithm NSGA-II is used to optimize and solve the objectives, obtaining a set of Pareto optimal solutions. Operators can then select an adjustment strategy from these solutions based on actual needs.

[0018] Furthermore, the adjustment and execution module adopts a collaborative control strategy. For different types of actuators, a coupled dynamics model is established to analyze their interactions and influences. A distributed collaborative control algorithm is used to enable different actuators to work together, avoiding mutual interference and conflicts during the adjustment process.

[0019] Furthermore, the model verification and optimization module also adopts a real-time data-driven model update method; it uses online learning algorithms to update and optimize the theoretical model and prediction model based on the boundary layer data collected in real time.

[0020] Furthermore, the human-computer interaction and remote monitoring module is also equipped with early warning and decision support functions. The system automatically issues early warning information and provides response suggestions and decision-making solutions when abnormal or dangerous conditions are detected by real-time analysis and prediction of boundary layer state data. At the same time, it uses knowledge graph technology to manage and mine wind tunnel experimental data and adjustment experience to provide decision support for operators.

[0021] A method for applying the aforementioned wind tunnel boundary layer control real-time adaptive adjustment algorithm system includes:

[0022] Data acquisition steps: Sensors are installed at the wind tunnel location to collect data on pressure, velocity, temperature, humidity, and wall shear stress of the boundary layer and environment through high-frequency sampling rate, and the data is transmitted through distributed self-organizing network technology;

[0023] Data preprocessing steps: The collected data is denoised using a filtering algorithm based on wavelet decomposition and adaptive threshold fusion, and then normalized using a dynamic normalization method according to the real-time distribution range of the data.

[0024] Boundary layer state assessment steps: Using a deep learning-based prediction model combined with a traditional theoretical model, calculate the boundary layer thickness, displacement thickness, momentum thickness, and transition location parameters, and predict the future state of the boundary layer;

[0025] Control decision-making steps: A hierarchical reinforcement learning and model predictive control fusion algorithm is adopted, combined with an adaptive exploration-exploitation policy, to determine the regulation policy, while considering multi-objective optimization;

[0026] The adjustment process involves: driving the adjustment action through a microelectromechanical system, transmitting control signals using coding and modulation technology, and coordinating the operation of the actuators.

[0027] Model validation and optimization steps: Real-time comparison of data and model results before and after adjustment; use Bayesian model averaging method to fuse and evaluate different models; and use online learning algorithm to update the model.

[0028] Human-computer interaction and remote monitoring steps: Human-computer interaction is achieved through virtual reality and augmented reality technologies, supporting remote monitoring and control. The system is equipped with early warning and decision support functions.

[0029] Furthermore, it also includes:

[0030] Environmental perception and compensation steps: Real-time monitoring of the external environmental meteorological conditions of the wind tunnel, and the use of fuzzy control algorithm to compensate and adjust the boundary layer control strategy according to changes in environmental parameters.

[0031] Sensor fault diagnosis and fault tolerance steps: Integrate and analyze sensor data and detect residuals to determine sensor faults, reconstruct faulty sensor data using Kalman filtering, and adjust the sampling and fusion strategies of each sensor.

[0032] Compared with existing technologies, the beneficial effects of this invention are:

[0033] In terms of improving experimental accuracy, the dense layout of high-precision, multi-type sensors and high-frequency sampling enable comprehensive and accurate acquisition of boundary layer parameters. Advanced data processing algorithms effectively remove noise interference and dynamically normalize data, making boundary layer state assessment more accurate. Predictive models combining deep learning and traditional theory accurately calculate key parameters and predict future states, providing strong support for precise control. The high response speed and precise adjustment capability of the actuators ensure that the boundary layer can quickly approach the ideal state, greatly improving the accuracy of experimental results and providing researchers with more reliable data.

[0034] From the perspective of adaptability and flexibility, the system demonstrates excellent real-time adaptive adjustment capabilities. Whether due to changes in experimental conditions inside the wind tunnel or interference from external environmental factors, the system can promptly adjust its control strategy through the environmental perception and compensation module and adaptive control algorithm. The multi-objective optimization mechanism can flexibly select adjustment strategies according to different experimental needs, meeting diverse experimental requirements. Furthermore, the system employs advanced technologies such as distributed self-organizing networks and intelligent material-driven systems, enhancing its versatility and enabling stable operation in various types of wind tunnels.

[0035] In terms of operational convenience and safety, this patented technology also offers significant advantages. The human-computer interaction interface based on virtual reality and augmented reality allows operators to intuitively and immersively observe and control the experimental process. Remote monitoring and intelligent early warning functions enable operators to stay informed about the wind tunnel's status at any time, promptly addressing any abnormalities and ensuring experimental safety. Simultaneously, the system's fault diagnosis and fault-tolerance technologies, along with model verification and optimization mechanisms, improve system reliability and stability, reduce equipment failures and maintenance costs, enhance the overall efficiency of wind tunnel experiments, and powerfully promote the development of aerodynamics research in related fields. Attached Figure Description

[0036] Figure 1 This is a schematic block diagram of the real-time adaptive adjustment system for wind tunnel boundary layer control proposed in this invention.

[0037] Figure 2 This is a schematic block diagram of the steps of the real-time adaptive adjustment algorithm for wind tunnel boundary layer control proposed in this invention;

[0038] Figure 3 This is a schematic diagram comparing the boundary layer thickness control accuracy of different systems for real-time adaptive adjustment of wind tunnel boundary layer control proposed in this invention.

[0039] Figure 4 This is a schematic diagram illustrating the change in the accuracy of experimental results over time for different systems of the wind tunnel boundary layer control proposed in this invention. Detailed Implementation

[0040] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0041] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "clockwise," and "counterclockwise," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0042] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, features defined with "first" and "second" may explicitly or implicitly include one or more of the stated features. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified. Furthermore, the terms "installed," "connected," and "linked" should be interpreted broadly; for example, they may refer to a fixed connection, a detachable connection, or an integral connection; they may refer to a mechanical connection or an electrical connection; they may refer to a direct connection or an indirect connection through an intermediate medium; and they may refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances. The invention will now be described in further detail with reference to the accompanying drawings.

[0043] Reference Figures 1 to 4 This invention proposes a specific implementation of a real-time adaptive adjustment system and method for wind tunnel boundary layer control, including:

[0044] Data acquisition module: Pressure sensors, velocity sensors, and temperature sensors are installed inside the wind tunnel, in the boundary layer region, and in the external environment of the wind tunnel. Humidity sensors and wall shear stress sensors are also installed. The sensors collect data, and the sensor nodes adopt distributed self-organizing network technology to automatically optimize the data transmission path and transmit the data to the data processing center.

[0045] The data preprocessing module employs a filtering algorithm based on wavelet decomposition and adaptive threshold fusion to filter the acquired sensor data. The threshold is adaptively determined based on the statistical characteristics of the data to remove noise interference. Next, the filtered data is normalized using a dynamic normalization method, which dynamically adjusts the normalization interval according to the real-time distribution range of the data. The formula is as follows: Adapting to changes in boundary layer state, where and These are the minimum and maximum values ​​of data that change dynamically over time.

[0046] Boundary layer state assessment module: Introduces a deep learning-based boundary layer state prediction model. Combining preprocessed data and historical data, it predicts the future state of the boundary layer. The model is based on Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) networks, while also incorporating traditional boundary layer theoretical models to comprehensively calculate the boundary layer thickness. Displacement thickness Momentum thickness θ and boundary layer transition location Parameters; the formula for predicting the boundary layer transition location is as follows: ,in Here, Tu is the local Reynolds number, Tu is the turbulence intensity, and f is a nonlinear function obtained by training a deep learning model.

[0047] Control Decision Module: This module employs an algorithm based on a fusion of hierarchical reinforcement learning and Model Predictive Control (MPC) to determine the regulation strategy. The control task is divided into two levels: high-level decision-making and low-level execution. High-level decision-making, based on reinforcement learning, determines the regulation direction and objective according to the boundary layer state and target. Low-level execution uses the MPC algorithm, which, based on high-level decision-making, predicts state changes over a future period and optimizes the regulation action sequence according to the real-time state and dynamic model of the boundary layer. Simultaneously, an adaptive exploration-exploitation strategy is introduced, dynamically adjusting the exploration and exploitation ratio based on the boundary layer state.

[0048] Adjustment Module: The adjustment actuator is driven by a microelectromechanical system (MEMS) and uses micro-nano fabrication technology to manufacture a wall structure with continuously adjustable roughness. The shape and height of the wall microstructure are controlled by external excitation through electric and magnetic fields to achieve continuous adjustment of the wall roughness. The control signal of the actuator adopts coding modulation technology to achieve anti-interference capability of signal transmission.

[0049] Model Validation and Optimization Module: During wind tunnel experiments, the boundary layer state data before and after adjustment are compared in real time with the results of theoretical and predictive models. The Bayesian Model Averaging (BMA) method is used to fuse and evaluate the prediction results of different models. Based on the fusion results, the model weights and parameters are dynamically adjusted. The formula is as follows: ,in This is the prediction result based on data D. It is the i-th model. It is the weight of the i-th model, and ;

[0050] Human-computer interaction and remote monitoring module: Develop a human-computer interaction interface based on virtual reality and augmented reality technologies. Operators can observe the boundary layer flow state and adjustment process inside the wind tunnel through the equipment, and obtain parameters and adjustment commands in real time in the actual wind tunnel scenario. At the same time, the system supports remote monitoring and control functions, and transmits wind tunnel operation data and status information to the remote monitoring center in real time through a wireless network. Operators can remotely formulate adjustment strategies and control the actuators.

[0051] This invention also includes:

[0052] Environmental perception and compensation module: Real-time monitoring of atmospheric pressure, ambient temperature, wind speed and direction outside the wind tunnel. Based on changes in environmental parameters, fuzzy control algorithm is used to compensate and adjust the boundary layer control strategy. The fuzzy control rules are established based on expert experience and experimental data. According to different combinations of environmental parameters and boundary layer states, the action amplitude and frequency of the regulating actuator are automatically adjusted.

[0053] In this invention, the data acquisition module employs sensor fault diagnosis and fault tolerance technology. By integrating and analyzing sensor data and detecting residuals, it can determine in real time whether a sensor has malfunctioned. When a sensor fault is detected, a state estimation method based on Kalman filtering is used to reconstruct the data of the faulty sensor, while automatically adjusting the sampling strategy and data fusion algorithm of each sensor.

[0054] In this invention, the control decision module introduces a multi-objective optimization mechanism. In addition to optimizing the stability and uniformity of the boundary layer, it also considers the energy consumption, adjustment time, and impact on wind tunnel experimental parameters during the adjustment process. The non-dominated sorting genetic algorithm NSGA-II is used to optimize the objectives and obtain a set of Pareto optimal solutions. The operator selects the adjustment strategy from these solutions according to actual needs.

[0055] In this invention, the adjustment execution module adopts a collaborative control strategy. For different types of actuators, a coupled dynamics model is established to analyze their interaction and influence. A distributed collaborative control algorithm is used to enable different actuators to work together, avoiding mutual interference and conflict during the adjustment process.

[0056] In this invention, the model verification and optimization module also adopts a real-time data-driven model update method; it uses online learning algorithms to update and optimize the theoretical model and prediction model based on the boundary layer data collected in real time.

[0057] In this invention, the human-computer interaction and remote monitoring module is also equipped with early warning and decision support functions. The system automatically issues early warning information and provides response suggestions and decision-making solutions when abnormal or dangerous conditions are detected by real-time analysis and prediction of boundary layer state data. At the same time, knowledge graph technology is used to manage and mine wind tunnel experimental data and adjustment experience to provide decision support for operators.

[0058] This invention also discloses a real-time adaptive adjustment method for wind tunnel boundary layer control, comprising:

[0059] Data acquisition steps: Sensors are installed at the wind tunnel location to collect data on pressure, velocity, temperature, humidity, and wall shear stress of the boundary layer and environment through high-frequency sampling rate, and the data is transmitted through distributed self-organizing network technology;

[0060] Data preprocessing steps: The collected data is denoised using a filtering algorithm based on wavelet decomposition and adaptive threshold fusion, and then normalized using a dynamic normalization method according to the real-time distribution range of the data.

[0061] Boundary layer state assessment steps: Using a deep learning-based prediction model combined with a traditional theoretical model, calculate the boundary layer thickness, displacement thickness, momentum thickness, and transition location parameters, and predict the future state of the boundary layer;

[0062] Control decision-making steps: A hierarchical reinforcement learning and model predictive control fusion algorithm is adopted, combined with an adaptive exploration-exploitation policy, to determine the regulation policy, while considering multi-objective optimization;

[0063] The adjustment process involves: driving the adjustment action through a microelectromechanical system, transmitting control signals using coding and modulation technology, and coordinating the operation of the actuators.

[0064] Model validation and optimization steps: Real-time comparison of data and model results before and after adjustment; use Bayesian model averaging method to fuse and evaluate different models; and use online learning algorithm to update the model.

[0065] Human-computer interaction and remote monitoring steps: Human-computer interaction is achieved through virtual reality and augmented reality technologies, supporting remote monitoring and control. The system is equipped with early warning and decision support functions.

[0066] This invention also includes:

[0067] Environmental perception and compensation steps: Real-time monitoring of the external environmental meteorological conditions of the wind tunnel, and the use of fuzzy control algorithm to compensate and adjust the boundary layer control strategy according to changes in environmental parameters.

[0068] Sensor fault diagnosis and fault tolerance steps: Integrate and analyze sensor data and detect residuals to determine sensor faults, reconstruct faulty sensor data using Kalman filtering, and adjust the sampling and fusion strategies of each sensor.

[0069] Example 1

[0070] I. System Hardware Deployment

[0071] A comprehensive sensor network is constructed inside the wind tunnel, in the boundary layer region, and at relevant locations in the external environment. High-precision pressure sensors are installed at technically required intervals at different heights and along the boundary layer to measure the pressure distribution within the boundary layer. Nearby, laser Doppler velocimeters are deployed as velocity sensors, with a measurement accuracy of ±0.02 m / s, ensuring accurate acquisition of airflow velocity information. Temperature sensors (accuracy ±0.1 K), humidity sensors (accuracy ±1% RH), and wall shear stress sensors (accuracy ±0.01 Pa) are also distributed at reasonable intervals to comprehensively collect various types of data. All these sensors are connected to a data acquisition card, which collects sensor data at a frequency of 1000-5000 Hz and transmits the data to the data processing center via distributed ad hoc networking technology. The ad hoc networking technology employs the AODV (Ad-HocOn-DemandDistanceVector) routing protocol to ensure the reliability and real-time performance of data transmission.

[0072] II. System Software Flow

[0073] (I) Data Preprocessing At the data processing center, after receiving data collected by sensors, noise reduction is first performed using a filtering algorithm based on multi-scale wavelet decomposition and adaptive threshold fusion. Taking wind speed data as an example, it is decomposed into sub-signals of different frequencies using multi-scale wavelet decomposition. For the detail coefficients at each scale, based on their local standard deviation... The threshold is adaptively determined based on a preset threshold coefficient k (e.g., k=3). The detail coefficients are processed using this threshold to remove noise components, and then wavelet reconstruction is performed to obtain the filtered wind speed data.

[0074] Next, the filtered data undergoes dynamic normalization. Taking pressure data P as an example, the minimum value of the pressure data within the current time period is obtained in real time. and maximum value According to the formula Normalization is performed to map the pressure data to the [0,1] interval for subsequent analysis.

[0075] (ii) Boundary layer state assessment

[0076] This study evaluates boundary layer states by combining a deep learning-based boundary layer state prediction model with a traditional boundary layer theory model. For the deep learning model, a convolutional neural network (CNN) and a long short-term memory network (LSTM) are used to construct the prediction model. Preprocessed sensor data, such as pressure, velocity, and temperature, are arranged according to time series and spatial location and used as input to the CNN. The CNN extracts spatial features of the data through convolutional and pooling layers, and then its output is passed to the LSTM. The LSTM processes the time series information and learns the changes in the boundary layer state over time.

[0077] Simultaneously, the boundary layer thickness is calculated by combining traditional boundary layer theory models. Displacement thickness Momentum thickness θ and boundary layer transition location Key parameters, such as boundary layer thickness calculation, are used as an example. According to the formula... Among them, the kinematic viscosity coefficient The distance x from the wind tunnel inlet to the measurement point can be calculated using empirical formulas based on the current temperature and gas properties. The sensor location determines the distance x, and the incoming flow velocity is also considered. The value is calculated by averaging the data measured by the velocity sensor. The boundary layer transition location is predicted based on the formula. Among them, the local Reynolds number The turbulence intensity Tu is calculated from the velocity fluctuations measured by a velocity sensor, and f is a nonlinear function trained by a deep learning model. The prediction results of the deep learning model are combined with the calculation results of the traditional theoretical model to comprehensively evaluate the stability, turbulence intensity, and other state information of the boundary layer.

[0078] (iii) Control Decisions

[0079] A hierarchical reinforcement learning and model predictive control (MPC) fusion algorithm is employed to determine the regulation strategy. The control task is divided into two levels: high-level decision-making and low-level execution. In high-level decision-making, the state space is defined as the key parameters of the boundary layer (such as boundary layer thickness, displacement thickness, turbulence intensity, etc.) and the current state of the regulating actuator; the action space is defined as the regulation direction and approximate amplitude of the regulating actuator; the reward function is set based on factors such as the degree to which the boundary layer state approaches the ideal state and regulation energy consumption. A deep Q-network (DQN) algorithm is used for training to learn the optimal regulation strategy.

[0080] In low-level execution, based on the MPC algorithm, the state changes of the boundary layer over a future period are predicted according to the real-time state and dynamic model of the boundary layer (such as a simplified Navier-Stokes discrete model). A series of adjustment action sequences are obtained by optimizing the objective function (such as minimizing the deviation between the boundary layer state and the ideal state, and the magnitude of the adjustment actions). During the optimization process, system constraints are considered, such as the adjustment range and response speed limitations of the actuators.

[0081] Simultaneously, an adaptive exploration-exploitation strategy is introduced to dynamically adjust the ratio of exploration to exploitation based on the uncertainty of the boundary layer state (such as the error of the prediction model). When the uncertainty is high, the proportion of exploration is increased to try more different adjustment actions; when the uncertainty is low, the proportion of exploiting the learned optimal strategy is increased.

[0082] (iv) Regulation and Implementation

[0083] The regulating actuator employs a shape memory alloy (SMA)-driven microelectromechanical system (MEMS) for wall suction / blowing control and wall roughness adjustment. For the wall suction / blowing device, an SMA microactuator is connected to a valve in the suction / blowing pipeline. Upon receiving a control signal, the SMA microactuator deforms differently according to the current magnitude, thereby precisely controlling the valve opening and regulating the suction / blowing rate. Its response speed is at the millisecond level, enabling rapid reaction to changes in the boundary layer state.

[0084] For wall roughness adjustment, a continuously adjustable wall structure is fabricated using micro-nano fabrication technology. By applying an electric or magnetic field to the wall microstructure, its shape and height are altered, achieving continuous and precise adjustment of the wall roughness. Control signals are transmitted using FSK (Frequency Shift Keying) coding modulation technology to improve signal anti-interference capabilities. During adjustment, different actuators (such as the air intake / blowing device and the wall roughness adjustment device) work collaboratively through a cooperative control strategy. By establishing a coupled dynamics model, the interactions and influences between them are analyzed. A distributed cooperative control algorithm is employed to ensure that the adjustment actions of each actuator are coordinated, avoiding mutual interference and improving adjustment effectiveness and efficiency.

[0085] (v) Model Validation and Optimization

[0086] During wind tunnel experiments, boundary layer state data before and after adjustment were compared in real time with the results of theoretical and predictive models. The Bayesian model averaging (BMA) method was used to fuse and evaluate the prediction results of different models. Assume there are M different models. For each model, its weights are calculated based on historical data and current experimental data. Historical data consists of boundary layer states and related parameter records accumulated during the early stages of wind tunnel experiments. It covers boundary layer data under different operating conditions and adjustment strategies, forming the basis for training and evaluating each model. When conducting the current experiment, after real-time acquisition of boundary layer state data (such as velocity profiles and turbulence intensity), this data is combined with historical data to evaluate each model. Evaluation: The likelihood of the model is calculated by analyzing its prediction accuracy for similar boundary layer scenarios in historical data and its fit with actual measurements under current experimental data. The formula is as follows: ,in This is the prediction result based on data D. This represents the prediction result of the i-th model on data D. Based on the fusion result, the model weights and parameters are dynamically adjusted. Online learning algorithms, such as Online Support Vector Machines (Online SVM), are used to continuously update and optimize the theoretical and prediction models based on real-time collected boundary layer data. Online learning algorithms can quickly adapt to changes in the boundary layer state, improving the model's real-time performance and adaptability.

[0087] (vi) Human-computer interaction and remote monitoring

[0088] Develop a human-computer interaction interface based on virtual reality (VR) and augmented reality (AR) technologies. In the VR environment, operators can immerse themselves in observing the boundary layer flow state inside the wind tunnel, and view boundary layer details from different angles, such as velocity vector distribution and pressure cloud maps, using a controller. In the AR environment, operators can use smart terminal devices, such as tablets or smart glasses, to view key parameters superimposed on the actual wind tunnel scene in real time, such as boundary layer thickness and the current status of the regulating actuators. Simultaneously, the system supports remote monitoring and control functions, transmitting wind tunnel operating data and status information to a remote monitoring center in real time via a 5G high-speed wireless network. Operators at the remote monitoring center can use computers or mobile devices, leveraging VR or AR technology, to remotely formulate regulation strategies and control actuators, just as if operating on-site.

[0089] The system also features intelligent early warning and decision support functions. Through real-time analysis and prediction of boundary layer state data, when the system detects that the boundary layer state may exceed the safe range or an anomaly occurs, it automatically issues early warning information, such as audible alarms and pop-up prompts. Simultaneously, based on preset rules and historical experience, the system provides corresponding response suggestions and decision-making solutions to help operators react quickly. Knowledge graph technology is used to manage and mine wind tunnel experimental data and adjustment experience, providing intelligent decision support for operators. For example, when encountering specific boundary layer anomalies, the system can quickly search the knowledge graph to find similar handling experiences and successful cases, providing operators with references.

[0090] III. Data Representation and Interpretation

[0091]

[0092] The data in the table clearly shows that our system offers significant improvements over traditional systems in several aspects. The substantial increase in boundary layer thickness control precision allows the boundary layer state in wind tunnel experiments to be maintained more stably within the ideal range, reducing experimental errors caused by boundary layer instability. The accuracy of experimental results is improved by 10-15 percentage points, meaning researchers can obtain more reliable data, providing a more solid foundation for research in related fields. The system response time is shortened by 50%-67%, enabling faster responses to changes in the boundary layer state and timely adjustments to control strategies, further improving the stability and reliability of the experiment. Energy efficiency is improved by 20-30 percentage points compared to traditional systems, demonstrating the energy-saving advantages of our system and reducing the operating costs of wind tunnel experiments. The ease of operation score has improved from 2-3 points to 4-5 points, an improvement of approximately 67%-150%. The VR / AR-based human-computer interaction interface and remote monitoring functions allow operators to operate and monitor more intuitively and conveniently, greatly improving work efficiency and reducing the possibility of operational errors. These improvements combined comprehensively enhance the quality and efficiency of wind tunnel experiments, providing strong support for the development of related fields.

[0093] The above are merely preferred embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A real-time adaptive adjustment system for wind tunnel boundary layer control, characterized in that, include: Data acquisition module: Pressure sensors, velocity sensors, and temperature sensors are installed inside the wind tunnel, in the boundary layer region, and in the external environment of the wind tunnel. Humidity sensors and wall shear stress sensors are also installed. The sensors collect data, and the sensor nodes adopt distributed self-organizing network technology to automatically optimize the data transmission path and transmit the data to the data processing center. The data preprocessing module employs a filtering algorithm based on wavelet decomposition and adaptive threshold fusion to filter the acquired sensor data. The threshold is adaptively determined based on the statistical characteristics of the data to remove noise interference. Next, the filtered data is normalized using a dynamic normalization method, which dynamically adjusts the normalization interval according to the real-time distribution range of the data. The formula is as follows: Adapting to changes in boundary layer state, where and These are the minimum and maximum values ​​of data that change dynamically over time. Boundary layer state assessment module: Introduces a deep learning-based boundary layer state prediction model. Combining preprocessed data and historical data, it predicts the future state of the boundary layer. The model is based on Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) networks, while also incorporating traditional boundary layer theoretical models to comprehensively calculate the boundary layer thickness. Displacement thickness Momentum thickness and boundary layer transition location Parameters; the formula for predicting the boundary layer transition location is as follows: ,in For the local Reynolds number, For turbulence intensity, It is a nonlinear function obtained by training a deep learning model; Adjustment Module: The adjustment actuator is driven by a microelectromechanical system (MEMS) and uses micro-nano fabrication technology to manufacture a wall structure with continuously adjustable roughness. The shape and height of the wall microstructure are controlled by external excitation through electric and magnetic fields to achieve continuous adjustment of the wall roughness. The control signal of the actuator adopts coding modulation technology to achieve anti-interference capability of signal transmission. Control Decision Module: This module employs an algorithm based on a fusion of hierarchical reinforcement learning and Model Predictive Control (MPC) to determine the regulation strategy. The control task is divided into two levels: high-level decision-making and low-level execution. High-level decision-making, based on reinforcement learning, determines the regulation direction and objective according to the boundary layer state and target. Low-level execution uses the MPC algorithm, which, based on high-level decision-making, predicts state changes over a future period and optimizes the regulation action sequence according to the real-time state and dynamic model of the boundary layer. Simultaneously, an adaptive exploration-exploitation strategy is introduced, dynamically adjusting the exploration and exploitation ratio based on the boundary layer state. Model Validation and Optimization Module: During wind tunnel experiments, the boundary layer state data before and after adjustment are compared in real time with the results of theoretical and predictive models. The Bayesian Model Averaging (BMA) method is used to fuse and evaluate the prediction results of different models. Based on the fusion results, the model weights and parameters are dynamically adjusted. The formula is as follows: ,in This is the prediction result based on data D. It is the i-th model. It is the weight of the i-th model, and ; Human-computer interaction and remote monitoring module: Develop a human-computer interaction interface based on virtual reality and augmented reality technologies. Operators can observe the boundary layer flow state and adjustment process inside the wind tunnel through the equipment, and obtain parameters and adjustment commands in real time in the actual wind tunnel scenario. At the same time, the system supports remote monitoring and control functions, and transmits wind tunnel operation data and status information to the remote monitoring center in real time through a wireless network. Operators can remotely formulate adjustment strategies and control the actuators. Environmental perception and compensation module: Real-time monitoring of atmospheric pressure, ambient temperature, wind speed and direction outside the wind tunnel. Based on changes in environmental parameters, fuzzy control algorithm is used to compensate and adjust the boundary layer control strategy. The fuzzy control rules are established based on expert experience and experimental data. According to different combinations of environmental parameters and boundary layer states, the action amplitude and frequency of the regulating actuator are automatically adjusted.

2. The real-time adaptive adjustment system for wind tunnel boundary layer control according to claim 1, characterized in that, The data acquisition module employs sensor fault diagnosis and fault tolerance technology. By integrating and analyzing sensor data and detecting residuals, it can determine in real time whether a sensor has malfunctioned. When a sensor fault is detected, a state estimation method based on Kalman filtering is used to reconstruct the data of the faulty sensor, while automatically adjusting the sampling strategy and data fusion algorithm of each sensor.

3. The real-time adaptive adjustment system for wind tunnel boundary layer control according to claim 1, characterized in that, The control decision module introduces an optimization mechanism, which, in addition to optimizing the stability and uniformity of the boundary layer, also considers the energy consumption, adjustment time, and impact on wind tunnel experimental parameters during the adjustment process. The non-dominated sorting genetic algorithm NSGA-II is used to optimize the objective and obtain a set of Pareto optimal solutions. Operators can then select an adjustment strategy based on actual needs.

4. The real-time adaptive adjustment system for wind tunnel boundary layer control according to claim 1, characterized in that, The regulation and execution module adopts a collaborative control strategy. For different types of actuators, a coupled dynamics model is established to analyze their interactions and influences. A distributed collaborative control algorithm is used to enable different actuators to work together, avoiding mutual interference during the regulation process.

5. The real-time adaptive adjustment system for wind tunnel boundary layer control according to claim 1, characterized in that, The model verification and optimization module also employs a real-time data-driven model update method; it utilizes online learning algorithms to update and optimize the theoretical and predictive models based on real-time collected boundary layer data.

6. The real-time adaptive adjustment system for wind tunnel boundary layer control according to claim 1, characterized in that, The human-computer interaction and remote monitoring module is also equipped with early warning and decision support functions. The system automatically issues early warning information and provides response suggestions and decision-making solutions when abnormal or dangerous conditions are detected by real-time analysis and prediction of boundary layer state data. At the same time, it uses knowledge graph technology to manage and mine wind tunnel experimental data and adjustment experience to provide decision support for operators.

7. A method for applying the wind tunnel boundary layer control real-time adaptive adjustment system according to any one of claims 1-6, characterized in that, include: Data acquisition steps: Sensors are installed at the wind tunnel location to collect data on pressure, velocity, temperature, humidity, and wall shear stress of the boundary layer and environment through high-frequency sampling rate, and the data is transmitted through distributed self-organizing network technology; Data preprocessing steps: The collected data is denoised using a filtering algorithm based on wavelet decomposition and adaptive threshold fusion, and then normalized using a dynamic normalization method according to the real-time distribution range of the data. Boundary layer state assessment steps: Using a deep learning-based prediction model combined with a traditional theoretical model, calculate the boundary layer thickness, displacement thickness, momentum thickness, and transition location parameters, and predict the future state of the boundary layer; Control decision-making steps: A hierarchical reinforcement learning and model predictive control fusion algorithm is adopted, combined with an adaptive exploration-exploitation policy, to determine the regulation policy, while considering multi-objective optimization; The adjustment process involves: driving the adjustment action through a microelectromechanical system, transmitting control signals using coding and modulation technology, and coordinating the operation of the actuators. Model validation and optimization steps: Real-time comparison of data and model results before and after adjustment; use Bayesian model averaging method to fuse and evaluate different models; and use online learning algorithm to update the model. Human-computer interaction and remote monitoring steps: Human-computer interaction is achieved through virtual reality and augmented reality technologies, supporting remote monitoring and control. The system is equipped with early warning and decision support functions.

8. The method for real-time adaptive adjustment system of wind tunnel boundary layer control according to claim 7, characterized in that, Also includes: Environmental perception and compensation steps: Real-time monitoring of the external environmental meteorological conditions of the wind tunnel, and the use of fuzzy control algorithm to compensate and adjust the boundary layer control strategy according to changes in environmental parameters.

9. The method for real-time adaptive adjustment system of wind tunnel boundary layer control according to claim 7, characterized in that, Also includes: Sensor fault diagnosis and fault tolerance steps: Integrate and analyze sensor data and detect residuals to determine sensor faults, reconstruct faulty sensor data using Kalman filtering, and adjust the sampling and fusion strategies of each sensor.

Images