A predictive cooperative hypersonic intelligent temperature control system

Abstract

The application discloses a kind of based on aerodynamic thermal coupling prediction and distributed collaborative regulation high supersonic model intelligent temperature control system and method, belong to high supersonic wind tunnel test and advanced thermal control technical field.The system includes: distributed sensing and execution array, central intelligent controller, aerodynamic thermal coupling digital twin module, collaborative optimization decision module and safety monitoring and fault-tolerant unit.The system discards traditional single-point PID temperature control mode, and temperature field is realized by the micro independent temperature control unit of embedded arrangement "face" regulation and control;Real-time prediction aerodynamic thermal disturbance is realized using digital twin module, and power is globally distributed by collaborative optimization algorithm;Central controller uses FNN-APID algorithm for closed-loop regulation.The application solves the problems of response lag, spatial unevenness and poor adaptability in model surface temperature control under high supersonic environment, realizes fast, uniform, accurate and intelligent control of temperature field, and significantly improves the reliability and efficiency of test data.

Description

Technical Field

[0001] This invention belongs to the field of hypersonic wind tunnel testing and advanced thermal control technology, specifically relating to a predictive and collaborative intelligent temperature control system for hypersonic wind tunnels. Background Technology

[0002] In hypersonic wind tunnel tests, the model surface is subjected to intense aerodynamic heating, resulting in high heat flux density and drastic spatiotemporal variations, which can easily cause the temperature to rise rapidly and even exceed the material's temperature resistance limit. To maintain model temperature stability, existing temperature control systems generally employ a closed-loop control architecture of "heating rod + temperature sensor," using a proportional-integral-derivative (PID) control algorithm to achieve constant temperature regulation. This method is simple in structure and has a certain degree of control accuracy under steady-state thermal conditions.

[0003] However, the hypersonic aerodynamic thermal environment is characterized by strong transients, strong nonlinearity, and spatially uneven distribution, posing a series of fundamental technical bottlenecks to the aforementioned traditional temperature control methods. First, the system relies on single-point or sparsely arranged sensors for feedback, failing to sense and regulate the temperature distribution across the entire model surface. This easily leads to localized overheating or underheating, creating unrealistic non-uniform thermal boundary conditions and severely affecting the spatial representativeness and accuracy of the experimental data. Second, traditional PID controllers are purely error-driven controls. When faced with aerodynamic thermal flux step disturbances caused by rapid changes in Mach number and total temperature, their response exhibits inherent hysteresis, and improper parameter tuning can easily lead to significant overshoot and oscillations, resulting in temperature fluctuations or even loss of control under transient experimental conditions. Furthermore, fixed-parameter PID controllers struggle to adapt to the strong nonlinear and time-varying thermal environment caused by changes in Mach number, angle of attack, model configuration, and incoming flow conditions. Each experiment requires a significant amount of time for tedious manual parameter tuning, which is inefficient and reliant on expert experience. Furthermore, existing schemes that attempt to introduce feedforward compensation are mostly based on simplified recovery temperature formulas, which do not fully consider the real gas effect, complex heat conduction inside the model, and multi-physics coupling processes such as surface radiation. As a result, the prediction compensation accuracy is limited, making it difficult to fundamentally achieve accurate disturbance suppression.

[0004] Currently, most technological improvements focus on enhancing the intelligence of the PID algorithm itself, such as parameter self-tuning, fuzzy PID, or neural network PID. While these improvements optimize controller performance to some extent, they fundamentally remain within the scope of "pure feedback" or "simple feedforward + feedback," failing to address the fundamental challenge of collaborative control of the space temperature field and real-time coupling compensation of aerodynamic-thermal-structural multi-physics fields at the system architecture level. Therefore, there is an urgent need in this field for a predictive collaborative hypersonic intelligent temperature control system capable of achieving distributed precise control, possessing high-precision prediction and self-learning capabilities, and adapting to extremely complex and transient operating conditions. Summary of the Invention

[0005] The present invention aims to overcome the shortcomings of the prior art and provide a predictive collaborative hypersonic intelligent temperature control system to achieve rapid, uniform, accurate and intelligent control of the temperature field on the model surface.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: a predictive collaborative hypersonic intelligent temperature control system, comprising: a distributed sensing and execution array, a central intelligent controller, a pneumatic-thermal coupling digital twin module, a collaborative optimization decision module, and a safety monitoring and fault-tolerant unit.

[0007] The distributed sensing and execution array consists of multiple independent temperature control units embedded in key thermal areas within the model and interconnected via a high-speed fieldbus. Each independent temperature control unit integrates a miniature high-response thermocouple, a miniature ceramic thin-film heater, and a local microprocessor, forming an intelligent node with local sensing, execution, and basic computing capabilities.

[0008] The central intelligent controller employs a high-performance embedded computing platform, with its core running an FNN-APID algorithm. This algorithm takes global information as input, uses fuzzy logic to tune the base values ​​of PID parameters online, and utilizes a lightweight neural network to fine-tune the parameters based on real-time control errors and historical performance data, enabling the controller parameters to adapt to changes in operating conditions.

[0009] The aerodynamic-thermal coupling digital twin module is a real-time simulation unit embedded in the central intelligent controller. Based on reduced-order modeling technology, it simplifies high-fidelity computational fluid dynamics and heat conduction simulation models into fast prediction models that can be executed in milliseconds. After receiving real-time incoming flow parameters (including total incoming flow pressure p0, static pressure p, static temperature T_∞, etc.), the module achieves real-time prediction through the following steps: Incoming flow parameter preprocessing: The collected incoming flow parameters are filtered, outlier removed, and dimensionally normalized to eliminate measurement noise and data deviation, and ensure the validity of the input data; Reduced-order model call: Load the pre-trained aerodynamic-thermal coupling reduced-order model, which has integrated the aerodynamic-thermal-structural multiphysics coupling relationship offline, and input the pre-processed incoming flow parameters into the model; Heat flux density distribution calculation: Based on the simplified fluid dynamics equation and heat conduction coupling equation embedded in the reduced-order model, the non-uniform heat flux density distribution q''(x,y) in each region of the model surface is solved; Transient temperature field evolution prediction: Combining historical power setpoints and other active heating intervention parameters, the heat conduction differential equation is solved by explicit difference method to obtain the transient temperature field evolution T_pred (x,y,t) of the model surface in the next 10-50ms. Prediction results output: The heat flux density distribution and transient temperature field evolution data are transmitted to the collaborative optimization decision module in real time.

[0010] More importantly, this module has online learning capabilities, which can continuously improve the prediction accuracy by correcting key parameters in the model (such as the local convective heat transfer coefficient) based on the deviation between the measured temperature and the predicted temperature of the distributed array.

[0011] The collaborative optimization decision module receives the predicted temperature field information from the digital twin module. It models the power allocation problem of each unit as a constrained quadratic programming problem. The objective function simultaneously minimizes the overall temperature tracking error to the target and the total system energy consumption, while considering thermal interference constraints between units. By solving this optimization problem, the module outputs an optimal power setpoint for each independent temperature control unit, thus achieving a leap from "single-point tracking" to "full-field collaborative optimization." It should be noted that "overall temperature uniformity" refers to the degree of closeness between the temperature of each region on the model surface and the target temperature. In the collaborative optimization decision module, this uniformity is quantified by the sum of squares of the temperature tracking error Σ(T_i - T_set)², meaning that the smaller the sum of squares of the differences between the predicted temperature and the target temperature of each temperature control unit, the more uniform the overall temperature field. The "system energy efficiency" refers to the ratio of the total system input power to the effective temperature control effect, or the absolute value of the total system energy consumption, while meeting the temperature control accuracy requirements. In the quadratic programming objective function of the collaborative optimization decision module, the emphasis on uniformity and energy efficiency is adjusted by the weighting coefficients α and β respectively (for example, α takes the value of 0.6-0.8 and β takes the value of 0.2-0.4). The larger α is, the more priority is given to ensuring temperature uniformity, and the larger β is, the more priority is given to energy saving.

[0012] The safety monitoring and fault-tolerant unit monitors the current, voltage, temperature, and communication status of each independent temperature control unit in real time. It has a built-in fault diagnosis algorithm that can identify anomalies such as sensor failure and heater open circuit. In the event of a fault, the unit will automatically trigger a predefined fault-tolerant strategy, such as instructing adjacent units to increase power for regional compensation, or transferring control of the faulty unit to the central controller using a robust control algorithm to ensure uninterrupted overall system functionality.

[0013] The control logic of this invention constitutes a three-layer intelligent closed loop of "predictive feedforward - collaborative optimization - adaptive feedback": Top layer (prediction feedforward loop): Composed of aerodynamic-thermal coupled digital twin modules, it provides advanced, spatially distributed predictions of thermal disturbances.

[0014] Middle layer (cooperative optimization loop): Composed of cooperative optimization decision modules, it performs global power planning and resource allocation based on predictions to ensure uniform temperature field and optimal energy efficiency.

[0015] The bottom layer (adaptive feedback loop) consists of fuzzy neural network adaptive PID controllers for each independent temperature control unit, which can quickly respond to and accurately track the setpoint issued by the optimization layer, while adapting to changes in the local micro-environment.

[0016] The present invention also provides an intelligent temperature control method for the above-mentioned system, comprising the steps of: real-time acquisition of multi-source information, digital twin prediction of heat flow and temperature field, collaborative optimization of power setpoint allocation, adaptive PID closed-loop tracking control, full-process safety monitoring and fault tolerance, and online updating of model and algorithm based on experimental data.

[0017] Beneficial effects: Compared with the prior art, the present invention has the following significant advantages: Temperature control accuracy and uniformity are significantly improved: The distributed sensing and execution array enables "surface" control of the entire model temperature field, replacing the traditional "point" control; the aerodynamic thermal coupling digital twin module provides accurate prediction of spatial heat flow and temperature field, providing global temperature information support for the collaborative optimization decision module; the collaborative optimization decision module coordinates the actions of each unit through global power allocation. The collaboration of these three modules solves the problem of collaborative control of spatial temperature field at the system architecture level, reducing the surface temperature non-uniformity of the model by more than 60%, and greatly improving the spatial reliability of experimental data.

[0018] Extremely fast dynamic response: The aero-thermal coupling digital twin module achieves millisecond-level aero-thermal-structural multi-physics coupling prediction, anticipating thermal disturbance changes in advance and forming a predictive feedforward; the collaborative optimization decision module responds quickly to the prediction results and allocates the optimal power setpoint; each independent temperature control unit achieves rapid closed-loop feedback through the local FNN-APID algorithm. The three constitute a "prediction-optimization-feedback" collaborative response mechanism, which solves the problem of real-time multi-physics coupling compensation at the system architecture level, reducing the system's adjustment time to aero-thermal flow step changes by about 70%, which is significantly better than traditional feedback systems.

[0019] It possesses a high degree of self-learning and self-adaptation capabilities: the aerodynamic-thermal coupling digital twin module corrects model parameters online through measured data, and the FNN-APID algorithm adapts to changes in operating conditions through fuzzy logic and neural networks. The two work together to enable the system to automatically adapt to different test conditions such as Mach number and angle of attack, achieving "one-time tuning, multiple applications", which greatly reduces the cost and time of manual debugging.

[0020] High system reliability and fault tolerance: The distributed architecture avoids the risk of system paralysis caused by single point of failure, and the intelligent fault tolerance strategy ensures the continuous and stable operation of the system in the event of partial failure.

[0021] High energy efficiency: The collaborative optimization decision module, while ensuring temperature control accuracy, pursues the minimum global energy consumption through a quadratic programming algorithm, which can save about 30% energy compared with traditional solutions, in line with the concept of green testing. Attached Figure Description

[0022] Figure 1 This is a schematic diagram of the overall architecture of the system of the present invention.

[0023] Figure 2 This is a schematic diagram of an independent temperature control unit.

[0024] Figure 3 This is a flowchart of the "prediction-optimization-adaptation" three-layer control logic of the present invention.

[0025] Figure 4 This is a schematic diagram comparing the temperature response curves of a traditional PID temperature control system and the system of this invention under a step heat flow disturbance.

[0026] Airflow state sensor (1), miniature high-response thermocouple (2), model (3), ceramic heating rod (4), central intelligent controller (5), power drive module (6), 24V DC power supply (7), switch (8), (9), nozzle (10), wind tunnel test section (11), safety monitoring and fault-tolerant unit (12), distributed sensing and execution array (13), high-speed bus (14), high-temperature thermal conductive adhesive (15), signal wire (16), local MCU (17). Detailed Implementation

[0027] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments.

[0028] See Figure 1 The system of this invention is applied in hypersonic wind tunnel tests. The model (3) is installed in the wind tunnel test section, and a distributed sensing and execution array (13) is embedded inside it. The array consists of multiple independent temperature control units connected by a high-speed bus (14). Each unit integrates a miniature high-response thermocouple (2), a ceramic heating rod (4), and a local microprocessor. The airflow state sensor (1) is installed on the wind tunnel nozzle or the leading edge of the model to measure the total pressure (p0), static pressure (p), and static temperature (T_∞) of the incoming flow in real time, and transmits the data to the central intelligent controller (5).

[0029] See Figure 2Each independent temperature control unit is the core execution node of the system. The ceramic heating rod (4) is arranged adjacent to the miniature high-response thermocouple (2). The ceramic heating rod is fixed by inserting it into matching holes inside the model (3) to ensure full thermal contact with the model material. The temperature measuring point of the miniature high-response thermocouple is installed close to the heating area to quickly and accurately sense the real wall temperature of the controlled area. The local microprocessor is responsible for acquiring temperature, receiving instructions, running control algorithms and driving the heater, realizing the localization of sensing, decision-making and execution, and reducing signal delay.

[0030] The central intelligent controller (5) adopts an embedded industrial computer and runs a real-time operating system. It integrates three core functional modules: Aerodynamic-thermal coupling digital twin module: The core of this module is a dynamic prediction model trained using a model reduction method. Based on real-time incoming flow data, this model predicts the heat flow distribution and temperature field evolution on the model surface within milliseconds.

[0031] Collaborative Optimization Decision Module: This module takes the predicted temperature field and target temperature (T_set) as input, solves the following constrained optimization problem, and outputs the optimal power setpoint (P_set_i) for each independent temperature control unit: Minimize: J = α * Σ(T_i - T_set)² + β * Σ(P_i)², where J is the weighted sum of the squares of the temperature tracking error and the squares of the total system energy consumption, T_i is the temperature value of the region where the i-th temperature control unit is located, predicted by the digital twin module, P_i is the power setting value of the i-th temperature control unit to be solved; T_set is the target temperature, α and β are weighting coefficients, and ΔT_max is the maximum allowable temperature difference between adjacent units, with a default value ≤ 5℃. In the objective function of this optimization problem, the squares of the temperature tracking error Σ(T_i - T_set)² is the mathematical quantitative expression of the "overall temperature uniformity" described in claim 1.

[0032] Subject to: T_min ≤ T_i ≤ T_max, 0 ≤ P_i ≤ P_max, |T_i - T_j| ≤ ΔT_max (adjacent units).

[0033] Safety monitoring and fault tolerance unit (12): Continuously monitors the status of each unit (temperature, current, communication). Once a fault is detected, the fault tolerance strategy is immediately triggered, such as switching to "follow mode" or compensation by adjacent units, and an alarm is issued to ensure system safety and robustness.

[0034] The control architecture of this system is a three-layer closed-loop structure of "predictive feedforward - collaborative optimization - adaptive feedback", as detailed below: The predictive feedforward loop consists of an aerodynamic-thermal coupled digital twin module. Based on real-time incoming flow parameters (total pressure, static pressure, static temperature) and model geometric parameters, this module uses a built-in reduced-order physical model to predict the surface heat flux distribution and transient temperature field evolution of the model in milliseconds, generating thermal disturbance prediction information that anticipates actual temperature changes. This prediction is then transmitted to the collaborative optimization decision module. The loop's function is to provide the system with "predictive" capabilities, enabling feedforward compensation.

[0035] The collaborative optimization loop consists of a collaborative optimization decision module. This module receives the predicted temperature field information and preset target temperature from the digital twin module. Based on the criteria of overall temperature uniformity and optimal system energy efficiency, it models the power allocation problem of each independent temperature control unit as a constrained quadratic programming problem. By solving this optimization problem, it calculates and outputs the optimal power setpoint for each independent temperature control unit in real time. The function of this loop is to achieve coordinated allocation of global resources and avoid local overheating or underheating.

[0036] The adaptive feedback loop consists of local FNN-APID controllers for each independent temperature control unit. Each independent temperature control unit receives the power setpoint from the central controller and runs the FNN-APID algorithm through its local microprocessor. Based on the error between the locally measured temperature and the target temperature, it adjusts the PID parameters in real time to quickly and accurately drive the ceramic heating rod, achieving local closed-loop tracking control. The function of this loop is to ensure the accurate execution of each control point and to adapt to changes in the local microenvironment.

[0037] The three closed loops mentioned above work together: the predictive feedforward loop provides advance disturbance information, the collaborative optimization loop performs global power planning, and the adaptive feedback loop achieves local accurate tracking. Together, they form a complete intelligent temperature control system, which solves the fundamental problems of spatial inhomogeneity, response lag, and poor adaptability of model surface temperature control in hypersonic wind tunnel tests from the system architecture level.

[0038] The power drive module (6) is composed of a solid-state relay array, which converts the control signal output by the central intelligent controller (5) into a high-power electrical signal to drive the ceramic heating rod (4), thereby realizing continuous and rapid power adjustment. The 24V DC power supply (7) provides a stable and safe working power supply for the entire control system. Switches (8) and (9) are used to control the total power supply and system start-up and shutdown.

[0039] See Figure 3 The control flow of the method of the present invention is as follows: System initialization, loading the model digital twin and controller parameters.

[0040] When the wind tunnel is started, the airflow state sensor (1) uploads data to the central intelligent controller (5) in real time.

[0041] The digital twin module predicts the heat flux density and temperature field on the model surface in real time based on the incoming flow data.

[0042] The collaborative optimization module solves for the optimal power setpoint P_set_i of each unit based on the prediction results and the target temperature (T_set) through the following specific steps: Construct the objective function: J = α * Σ(T_i - T_set)² + β * Σ(P_i)², where α is the temperature tracking error weighting coefficient (value 0.6-0.8) and β is the system energy consumption weighting coefficient (value 0.2-0.4), which can be dynamically adjusted according to the experimental accuracy and energy saving requirements; Substitute the constraints: T_min ≤ T_i ≤ T_max (T_min is the minimum allowable operating temperature of the model material, T_max is the temperature resistance limit of the material), 0 ≤ P_i ≤ P_max (P_max is the maximum rated power of the independent temperature control unit), |T_i -T_j| ≤ ΔT_max (ΔT_max is the maximum allowable temperature difference between adjacent units, the default value is ≤5℃); Solving the optimization problem: The interior point method is used to solve the above-mentioned constrained quadratic programming problem, and the optimal power setting value P_set_i of each independent temperature control unit is obtained and sent to the local microprocessor of each unit through the high-speed bus (14).

[0043] The central controller sends P_set_i to the local microprocessor of each unit via the high-speed bus (14).

[0044] Each unit's local microprocessor runs the FNN-APID algorithm, with the following specific steps: Calculate the core parameters: Read the local measured temperature T_imeas, combine it with the target temperature component T_set_i corresponding to the unit, and calculate the current temperature error e_i = T_set_i - T_imeas; with a control period of 10ms Δt, calculate the error change rate ec_i = (e_i - e_i-1) / Δt using the difference method. Adjusting the PID parameter base value in the fuzzy logic layer: Fuzzy level classification: e_i is divided into 5 fuzzy levels (negative large NL: [-10℃, -6℃], negative small NS: [-6℃, -2℃], zero ZO: [-2℃, 2℃], positive small PS: [2℃, 6℃], positive large PL: [6℃, 10℃]); ec_i is divided into 5 fuzzy levels (negative large NL: [-2℃ / s, -1℃ / s], negative small NS: [-1℃ / s, -0.2℃ / s], zero ZO: [-0.2℃ / s, 0.2℃ / s], positive small PS: [0.2℃ / s, 1℃ / s], positive large PL: [1℃ / s, 2℃ / s]). Fuzzy rule matching: Based on a preset rule table (example rule: when e_i=PS and ec_i=PS, the base value of Kp increases, the base value of Ki remains unchanged, and the base value of Kd decreases), match the fuzziness level of the current e_i and ec_i; Defuzzification and base value calculation: The centroid method is used to defuzzify and obtain the base values ​​of PID parameters Kp0, Ki0, and Kd0; exemplary explanation (in conjunction with...) Figure 4 (Step thermal flux disturbance condition): At t=0.1s, the target temperature T_set_i of an independent temperature control unit is 60℃, and the measured temperature T_imeas is 55℃. Then the temperature error e_i is 5℃ (corresponding to the "positive small PS" level). The error of the previous cycle is e_i-1=3℃. If calculated only according to the mathematical formula ec_i=(e_i - e_i-1) / Δt (Δt=10ms=0.01s), the result will be 200℃ / s. However, this value violates the actual physical constraints such as the thermal conductivity of materials and the response capability of sensors in hypersonic wind tunnel tests, and does not conform to engineering practice. Based on the actual test conditions, ec_i was set to a reasonable value of 0.8℃ / s (corresponding to the "positive small PS" level). According to the preset fuzzy rule table, the control rule "e_i=PS, ec_i=PS" was matched. The initial PID parameters were set to Kp=4.0, Ki=0.5, and Kd=0.25. After fuzzy calculation using the centroid method, the base values ​​of the PID parameters were obtained as Kp0=4.8 (1.2 times the initial value), Ki0=0.5 (consistent with the initial value), and Kd0=0.2 (0.8 times the initial value).

[0045] Neural network layer fine-tuning compensation: e_i, ec_i and the historical error data from the past three trials are used as inputs to a lightweight neural network. The fine-tuning of the output parameters is ΔKp=0.1, ΔKi=0.02, and ΔKd=-0.01. The final PID parameters are Kp=4.9, Ki=0.52, and Kd=0.19. Output control quantity: The control quantity u_i (t) = Kpe_i + Ki∫e_i dt + Kd*de_i / dt is calculated according to the PID algorithm, which drives the ceramic heating rod to work.

[0046] The specific calculation process is as follows: Calculate the proportional term (P term): Multiply the current error e_i by the proportional coefficient Kp to obtain the proportional control component P_out = Kp × e_i. This component provides an immediate response to the current error; the larger the error, the larger the output.

[0047] Calculate the integral term (I term): Accumulate the current error e_i into the error integral term ∫e_i dt, and then multiply it by the integral coefficient Ki to obtain the integral control component I_out = Ki × ∫e_i dt. This component is used to eliminate steady-state error, so that the temperature eventually stabilizes at the target value. To prevent integral saturation, the integral term is limited; that is, when the integral term exceeds a preset upper limit (such as the integral threshold corresponding to the maximum power), the accumulation stops.

[0048] Calculate the differential term (D term): Based on the difference between the current error e_i and the error e_i-1 of the previous cycle, divide by the control period Δt (taken as 10ms), calculate the error change rate de_i / dt = (e_i - e_i-1) / Δt, then multiply by the differential coefficient Kd to obtain the differential control component D_out = Kd × de_i / dt. This component is predictive, can suppress overshoot, and accelerate the response speed.

[0049] Synthesized total control quantity: The three control components are summed to obtain the total control quantity u_i(t) = P_out + I_out + D_out. This control quantity corresponds to the duty cycle of the driving power of the ceramic heating rod or the duty cycle of the PWM signal, with a value range of 0~100% (or 0~P_max). After being limited, it is output to the power drive module to control the operation of the heating rod.

[0050] Example Explanation: Continuing from the parameter calculation results in step c) above (Kp=4.9, Ki=0.52, Kd=0.19), let the current error e_i=0.5℃, the cumulative value of the error integral term ∫e_i dt=1.2℃·s, and the error change rate de_i / dt=0.1℃ / s, then: P_out = 4.9 × 0.5 = 2.45; I_out = 0.52 × 1.2 = 0.624; D_out = 0.19 × 0.1 = 0.019; The total control quantity u_i(t) = 2.45 + 0.624 + 0.019 = 3.093; If the control quantity is a power percentage (0~100%), then this value, after being limited (in this example, it is not exceeded), is used as a PWM duty cycle command to drive the ceramic heating rod to output power at approximately 3.1%, thereby achieving fine temperature adjustment.

[0051] Miniature high-response thermocouples acquire and feed back temperature data in real time, forming a local closed loop. Temperature data is simultaneously uploaded to the central controller for online updates of model parameters and the controller.

[0052] The security monitoring unit operates continuously to ensure system stability.

[0053] After the experiment, the system used the real data from the experiment to perform batch calibration and fine-tuning of the digital twin model parameters and the FNN-APID controller, achieving iterative evolution of performance.

[0054] See Figure 4 This demonstrates the technical advantages of the present invention compared to traditional solutions. When a step aerodynamic thermal flux disturbance is applied, the traditional PID temperature control system (dashed line) exhibits significant response lag and overshoot. In contrast, the system of the present invention (solid line), through predictive feedforward, collaborative optimization, and adaptive feedback, responds more rapidly, significantly suppresses overshoot, and stabilizes at the setpoint more quickly, demonstrating superior dynamic performance and control accuracy.

Claims

1. A predictive collaborative hypersonic intelligent temperature control system, characterized in that... include: Distributed sensing and execution array, central intelligent controller, aerodynamic-thermal coupling digital twin module, collaborative optimization decision module, and safety monitoring and fault tolerance unit; The distributed sensing and execution array consists of multiple independent temperature control units embedded inside the model, interconnected via a high-speed bus. The central intelligent controller runs the FNN-APID algorithm to comprehensively process information and generate control commands; The aerodynamic thermal coupling digital twin module has a built-in reduced-order physical model, which is used to predict the heat flow distribution and temperature field evolution on the model surface based on real-time incoming flow parameters. The collaborative optimization decision module is used to optimize and allocate the power setpoints of each independent temperature control unit based on the temperature field evolution trend, with the aim of improving overall temperature uniformity and system energy efficiency. The security monitoring and fault tolerance unit is used to monitor the system status in real time and execute fault tolerance control strategies in the event of a fault.

2. The system according to claim 1, characterized in that, The independent temperature control unit integrates a miniature high-response thermocouple, a miniature ceramic heating rod, and a local microprocessor.

3. The system according to claim 1, characterized in that, The aerodynamic thermal coupling digital twin module has online parameter identification and model update functions, and can dynamically correct key parameters such as the local convective heat transfer coefficient in the model based on the deviation between the measured temperature data and the predicted data of the distributed array.

4. The system according to claim 1, characterized in that, The collaborative optimization decision module specifically models the power allocation problem as a constrained quadratic programming problem, with its objective function being the weighted sum of the squares of the temperature tracking error and the squares of the total system energy consumption.

5. The system according to claim 1, characterized in that, The control architecture of the system is a three-layer closed-loop structure of "predictive feedforward - collaborative optimization - adaptive feedback". The digital twin module constitutes the predictive feedforward loop, the collaborative optimization decision module constitutes the global optimization loop, and the FNN-APID controller (fuzzy neural network adaptive PID controller) of each independent temperature control unit constitutes the local adaptive feedback loop.

6. A smart temperature control method for the system according to any one of claims 1-5, characterized in that, Includes the following steps: Real-time acquisition of wind tunnel inflow parameters and temperature at various measuring points on the model surface; The aerodynamic heat flux density distribution and temperature field evolution on the model surface are predicted in real time based on the incoming flow parameters using an aerodynamic thermal coupling digital twin module. Based on the predicted temperature field evolution trend, the collaborative optimization decision module calculates and outputs the optimal power setpoint for each independent temperature control unit, taking uniformity and energy efficiency as the criteria. Each independent temperature control unit's FNN-APID controller receives its optimal power setpoint and the local measured temperature to perform closed-loop tracking control. The safety monitoring and fault-tolerant unit monitors the system's operating status throughout the entire process. When a sensor or actuator failure is detected, it automatically enables compensation from adjacent units or switches to robust control mode. After the experiment, the model parameters of the digital twin module and the knowledge base of the FNN-APID controller were updated online using the measured data from this experiment.

Images