Cooperative regulation and control system and method for dynamically testing pod environmental parameters

Abstract

The invention discloses a coordinated regulation and control system and method for dynamically testing pod environmental parameters, and the system comprises a multi-source sensor array which comprises a plurality of sensors used for collecting various environmental parameters in a pod in real time; the central controller is used for fusing sensor data and digital twinborn prediction information in real time through an FPGA + GPU heterogeneous architecture, dynamically generating and optimizing a regulation and control strategy based on a deep reinforcement learning algorithm, and synchronously coordinating an execution mechanism group to realize multi-physics field coupling control; the digital twinning interface is used for synchronizing the physical test environment and the bidirectional data flow of the digital twinning interface in real time, calculating and predicting the environment parameter evolution trend based on multi-physics field coupling, and feeding back closed-loop correction model parameters through reinforcement learning; and the execution mechanism group is used for adjusting the environmental parameters according to the control instruction. The system has the testing environment control functions of multi-parameter coordinated regulation and control, low-delay dynamic response and intelligent optimization capability, and the accuracy and testing efficiency of complex working condition simulation can be improved.

Description

Technical Field

[0001] The present invention relates to the field of intelligent dynamic testing technology, and in particular to a system, method and system for collaboratively controlling environmental parameters of a dynamic test pod, which is particularly suitable for collaboratively controlling dynamic environmental parameters in reliability testing of aerospace, new energy vehicles and electronic equipment. Background Art

[0002] In traditional testing environments, the environmental parameters of a pod (or test chamber), such as temperature, humidity, air pressure, vibration, oxygen content, deformation, environmental cleanliness, and organic matter concentration, are typically manually preset or independently controlled. This approach presents the following problems in actual use:

[0003] First, the ability to coordinate and control multiple parameters is insufficient: existing systems struggle to dynamically coordinate and simulate multiple environmental parameters, resulting in significant discrepancies between the test environment and actual operating conditions. For example, when an aircraft performs high-speed maneuvers, parameters such as temperature, air pressure, and vibration must change synchronously, and traditional methods cannot accurately simulate this coupling effect.

[0004] Second, the dynamic response performance is poor: there is a significant delay in adjusting environmental parameters, making it difficult to meet the requirements of high-precision dynamic testing (such as aircraft transient environment simulation and new energy vehicle battery sudden temperature change testing). This lag can lead to distorted test data and affect product reliability verification.

[0005] Third, they lack adaptive learning capabilities: Existing systems typically employ fixed control strategies and are unable to adaptively adjust parameter combinations based on real-time feedback from the test object (such as material deformation and battery thermal runaway signals). This passive control model limits test efficiency and coverage, making it difficult to meet the verification requirements of complex working conditions.

[0006] Therefore, there is an urgent need for a test environment control system with multi-parameter coordinated regulation, low-latency dynamic response and intelligent optimization capabilities to improve the accuracy of complex working condition simulation and testing efficiency. Summary of the Invention

[0007] The present invention aims to solve at least one of the technical problems existing in the prior art. To this end, the present invention proposes a system and method for collaboratively controlling environmental parameters in a dynamic test pod. This system features test environment control functions with multi-parameter collaborative control, low-latency dynamic response, and intelligent optimization capabilities, improving the accuracy and efficiency of complex working condition simulations.

[0008] A dynamic test pod environmental parameter collaborative control system according to an embodiment of the present invention includes:

[0009] A multi-source sensor array, including various sensors for real-time collection of various environmental parameters in the pod, with a sampling frequency of ≥10kHz, and the collected data synchronized by timestamps; all sensors are installed in the pod;

[0010] The central controller is used to integrate sensor data and digital twin prediction information in real time through an FPGA+GPU heterogeneous architecture, dynamically generate and optimize control strategies based on deep reinforcement learning algorithms, and synchronously coordinate the actuator group to achieve multi-physics field coupling control;

[0011] The digital twin interface is used to synchronize the bidirectional data flow between the physical test environment and the digital twin interface in real time, predict the evolution trend of environmental parameters based on multi-physics field coupling calculations, and correct model parameters through reinforcement learning feedback closed loop;

[0012] The actuator group is used to adjust the environmental parameters according to the control instructions.

[0013] According to the collaborative control system for environmental parameters of a dynamic test pod according to an embodiment of the present invention, a multi-source sensor array transmits environmental parameter data to a central controller in real time, combines it with the digital twin prediction results to generate an optimized control strategy, and sends it to the actuator via a high-speed bus. Simultaneously, it monitors environmental changes in real time to form a closed-loop control, with a full-link delay of ≤5ms. This system breaks through the limitations of traditional single-parameter independent control and achieves precise dynamic control of multiple physical quantities in complex test environments through multi-source data fusion, digital twin collaboration, and reinforcement learning optimization.

[0014] In some embodiments of the present invention, the central controller includes:

[0015] Real-time signal preprocessing module, deployed on FPGA, used to implement Kalman filtering noise reduction and feature extraction of sensor data;

[0016] A multi-parameter collaborative control module, deployed on the GPU, is used to run the DNN policy generator including the attention mechanism to dynamically generate control policies;

[0017] The digital twin comparison module is used to perform time series alignment and difference analysis between sensor measured data and simulation prediction data.

[0018] In some embodiments of the present invention, the working method of the multi-parameter coordinated control module includes:

[0019] Establish an environmental parameter coupling matrix to quantify the physical correlation between temperature, humidity, air pressure and vibration;

[0020] The mapping relationship between historical control strategies and test object responses is learned through the LSTM network;

[0021] The control parameter combination is optimized in real time based on the PPO algorithm and output to the actuator group.

[0022] In some embodiments of the present invention, the digital twin interface implementation method includes:

[0023] Based on the FMI standard package multi-physics simulation model, it supports thermal-fluid-solid coupling calculation;

[0024] A sliding window algorithm is used to align measured and simulated data streams, and the window width can be dynamically adjusted;

[0025] Identify deviation patterns and generate remediation strategies through a variance analysis engine.

[0026] In some embodiments of the present invention, the actuator group includes a temperature control module, an air pressure regulating valve and an active shock absorption platform. The temperature control module includes a variable frequency semiconductor temperature control piece connected in parallel with a liquid cooling circuit and distributed on the four walls of the pod; the air pressure regulating valve is a butterfly valve driven by a multi-stage stepper motor and installed in the air intake / exhaust channel at the top of the pod; the active shock absorption platform is a six-degree-of-freedom electromagnetic actuator integrated in the base at the bottom of the pod.

[0027] In some embodiments of the present invention, a fault tolerance module may also be included, and the fault tolerance module is used to:

[0028] Real-time detection of sensor anomalies, switching to redundant sensors when data deviates by ≥3σ;

[0029] When the actuator fails, the control quantity is redistributed based on the coupling model;

[0030] Periodically evaluate system health and trigger an alarm when the confidence level is less than 95%.

[0031] In some embodiments of the present invention, the system supports multi-level control modes.

[0032] Local fast response mode, where the FPGA directly processes emergency signals;

[0033] Global optimization mode, where GPU integrates digital twin data to generate a full parameter control plan;

[0034] Manual intervention mode, adjust the control weight parameters through the visual interface.

[0035] According to an embodiment of the present invention, a method for collaboratively controlling environmental parameters based on the above system includes:

[0036] S1, collects environmental parameters in real time through a multi-source sensor array, synchronizes timestamps and transmits them to the central controller;

[0037] S2, the digital twin interface receives the virtual model prediction data and compares it with the measured data in multiple dimensions;

[0038] S3. The central controller dynamically adjusts the parameter combination of the actuator group according to the difference analysis results;

[0039] S4. Record the test object response data and update the control strategy knowledge base.

[0040] In some embodiments of the present invention, a fault self-diagnosis process is further included. The fault self-diagnosis process is triggered when an abnormality is detected. The fault self-diagnosis process includes:

[0041] Implement triple verification of abnormal sensor data: adjacent sensor comparison, digital twin prediction value verification, and historical data trend analysis;

[0042] The actuator failure compensation strategy includes: concentrating the temperature control task on the non-faulty area and adjusting the air pressure valve to maintain thermal balance.

[0043] In some embodiments of the present invention, the PPO algorithm is used to perform strategy optimization in step S3, generating N groups of candidate actions in each iteration; the effects of the candidate actions are evaluated through rapid simulation of digital twins; and the action combination with the highest comprehensive score is selected for execution. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 is a schematic diagram of a system for collaboratively controlling environmental parameters of a dynamic test pod according to an embodiment of the present invention;

[0045] Figure 2 Schematic diagram of the process of the present invention. DETAILED DESCRIPTION

[0046] The following describes embodiments of the present invention in detail, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present invention, and are not to be construed as limiting the present invention.

[0047] Reference below Figure 1-Figure 2A dynamic test pod environmental parameter collaborative control system according to an embodiment of the present invention is described, which includes a multi-source sensor array, a central controller, a digital twin interface and an actuator group. The multi-source sensor array includes a variety of sensors for real-time collection of various environmental parameters in the pod, with a sampling frequency ≥10kHz, and the collected data are synchronized through timestamps; the central controller is used to integrate sensor data and digital twin prediction information in real time through an FPGA+GPU heterogeneous architecture, dynamically generate and optimize control strategies based on deep reinforcement learning algorithms, and synchronously coordinate the actuator group to achieve multi-physical field coupling control; the digital twin interface is used to synchronize the two-way data flow between the physical test environment and the digital twin interface in real time, predict the evolution trend of environmental parameters based on multi-physical field coupling calculations, and correct model parameters through a reinforcement learning feedback closed loop; the actuator group is used to adjust the environmental parameters according to control instructions.

[0048] According to an embodiment of the present invention, a dynamic test pod environmental parameter collaborative control system includes:

[0049] For example, the multi-source sensor array may include various high-precision sensors for detecting temperature, humidity, pressure, vibration, electromagnetic interference, and light intensity in the pod, respectively. These sensors can be networked via CAN FD / Gigabit Ethernet to form a distributed monitoring network. Specifically, the multi-source sensor array includes: temperature sensors installed on the four walls and top / bottom of the pod, such as infrared array + thermocouple redundant design, with a resolution of 0.1°C and a response time of 10ms; infrared array and thermocouple dual-mode temperature detection, with the infrared array (non-contact) and thermocouple (contact) mutually calibrated, and self-diagnosis can be triggered when the deviation is greater than 0.5°C; humidity sensors installed diagonally in the pod, such as capacitive polymer thin film sensor units with an accuracy of ±1%RH and a drift of <0.5% / year; and pressure sensors installed in the top air intake / exhaust duct, such as MEMS piezoresistive absolute pressure sensors with a range of 0-1500hPa and a linearity of 0.05%F. Force sensor; vibration sensor, installed at the bottom base of the pod, such as a three-axis MEMS accelerometer with a bandwidth of 5kHz and a range of ±50g; electromagnetic interference sensor, installed near the key electronic equipment of the cabin, such as a broadband field strength probe (10Hz-6GHz) with a dynamic range of 80dB and a sensitivity of 1V / m; light intensity sensor, installed in the lighting area of ​​the cabin, such as a photodiode array with a spectral range of 380-1100nm and an accuracy of ±5lux; deformation sensor, installed at the stress concentration point of the cabin structure, which can use an electromagnetic interference-resistant fiber Bragg grating strain gauge; gas composition sensor, installed close to the test object, such as a laser spectrum analyzer with a detection limit of 0.1ppm and a response time of <1s.

[0050] The central controller, serving as the "brain" of the entire environmental parameter control system, utilizes a heterogeneous FPGA+GPU architecture. The FPGA handles real-time signal preprocessing, while the GPU runs the DNN-RL joint optimization algorithm. This ensures end-to-end latency of ≤1ms through the Time-Sensitive Networking (TSN) protocol. The central controller's input receives raw data from the multi-source sensor array and simulated data streams from the digital twin interface. Its output sends control commands to the actuator group via a high-speed I / O interface. The real-time linkage solution for the digital twin interface requires a multi-module collaborative system to achieve data closure and policy optimization. The actuator group adjusts the operating parameters of each actuator based on the commands to achieve environmental parameter control.

[0051] It is understandable that the synchronous sampling frequency of all sensors is ≥10kHz, and the local sampling rate of the vibration sensor can reach 50kHz (to cope with transient shocks). Each sensor collects data synchronously with high precision, and can achieve μs-level time synchronization through the IEEE 1588v2 (PTP) protocol to ensure the time alignment of multi-parameter data and time sequence consistency. The multi-source sensor array transmits environmental parameter data to the central controller in real time, combines the digital twin prediction results to generate an optimized control strategy, and sends it to the actuator through the high-speed bus. At the same time, real-time monitoring of environmental changes forms a closed-loop control, and the full-link delay is ≤5ms. It breaks through the limitations of traditional single-parameter independent control and realizes precise dynamic control of multiple physical quantities in complex test environments through multi-source data fusion, digital twin collaboration and reinforcement learning optimization.

[0052] In some embodiments of the present invention, the central controller may include a real-time signal preprocessing module, a multi-parameter collaborative control module and a digital twin comparison module; the real-time signal preprocessing module is deployed on the FPGA to implement Kalman filtering denoising and feature extraction of sensor data; the multi-parameter collaborative control module is deployed on the GPU to run a DNN strategy generator including an attention mechanism to dynamically generate control strategies; the digital twin comparison module is used to perform time series alignment and difference analysis between the sensor measured data and the simulation prediction data.

[0053] Specifically, the training method of the DNN strategy generator may include: the input layer receives a 10-second historical data window and a test object state vector; the attention mechanism layer assigns a dynamic weight of 0.7-1.2 to the burst signal; and the output layer generates a control instruction set including temperature setting values ​​and valve openings.

[0054] In some embodiments of the present invention, the working method of the multi-parameter collaborative control module includes: first establishing an environmental parameter coupling matrix to quantify the physical correlation between temperature, humidity, air pressure and vibration; then learning the mapping relationship between historical control strategies and test object responses through the LSTM network; then, optimizing the control parameter combination in real time based on the PPO algorithm and outputting it to the actuator group.

[0055] For example, the environmental parameter coupling matrix is ​​to establish a multi-parameter interaction matrix to measure the linear interference between temperature changes on humidity, vibration, and air pressure. Specifically, the parameter interaction matrix is ​​constructed based on the nonlinear partial differential equation as follows:

[0056]

[0057] Where T is temperature, H is humidity, P is air pressure, and coefficients α, β, and γ are fitted through historical data.

[0058] It should be noted that the DNN-RL joint optimization algorithm running on the GPU in the central controller can include: the digital twin predicts environmental state trends over the next 5 seconds; the deep neural network (DNN) generates an initial control strategy based on the prediction results; and the reinforcement learning (RL) module corrects the strategy online, optimizing the control parameter combination by minimizing the objective function. The objective function includes: an environmental parameter tracking error term, which calculates the root mean square deviation between the measured value and the target value; an energy consumption term, which evaluates the total power consumption of the actuator action; and a system stability term, which monitors parameter overshoot and oscillation frequency.

[0059] In view of this, the DNN-RL joint optimization algorithm running on GPU is as follows:

[0060] First, a deep neural network (DNN) is trained based on historical test data to predict the optimal parameter set;

[0061] Second, a reinforcement learning (RL) mechanism is introduced to adjust the control strategy based on real-time feedback from the test object (such as battery charge and discharge efficiency and material deformation);

[0062] Third, realize online update of algorithm model.

[0063] Specifically, the deep neural network (DNN) training inputs are historical environmental parameter sequences (time windows ≥ 10 seconds) and test object feedback (such as battery voltage and deformation data). The output is the optimal parameter combination (temperature setpoint, air pressure valve opening percentage, etc.). The objective function is to minimize the weighted sum of energy consumption and error. It introduces an attention mechanism to enhance robustness to sudden interference (such as electromagnetic pulses).

[0064] The state space of reinforcement learning (RL) online optimization is the current environmental parameter value, the test object state, and the digital twin prediction value; the action space is the actuator control instruction (such as temperature control power and valve opening), and its reward function is:

[0065] R=ω1(1-Err1)+ω2(1-W P )-ω3P 超能 , where Err1 is the relative temperature error, W P is the vibration energy consumption, P超能 is the air pressure excess statistic, ω1, ω2, ω3 are the reward and punishment coefficients of the three.

[0066] The online update mechanism is to regularly upload local model parameters, aggregate multi-node data, and then send the global model for update.

[0067] In some embodiments of the present invention, the digital twin interface implementation method includes: encapsulating a multi-physics field simulation model based on the FMI standard to support thermal-fluid-solid coupling calculations; using a sliding window algorithm to align measured and simulated data streams, and the window width can be dynamically adjusted; identifying deviation patterns through a difference analysis engine and generating correction strategies.

[0068] Specifically, the digital twin in this system serves as an intelligent link between the physical world and virtual space, playing the following core roles:

[0069] 1) Dynamic Prediction Engine: Builds a high-fidelity virtual pod model based on multi-physics coupled simulation (thermodynamics / fluid mechanics / structural mechanics); outputs real-time prediction curves of environmental parameters (temperature gradient / air pressure fluctuation / vibration spectrum, etc.) for the next 5 seconds. For example, predicting the impact of airflow changes on temperature and humidity;

[0070] 2) Virtual-Real Calibration Hub: Uses a sliding window algorithm to align measured data streams with simulation timing (synchronization accuracy <100μs); identifies deviation patterns (trend offsets / mutation points / phase differences) through an LSTM network; and dynamically modifies model parameters (such as updating turbulence coefficients or material thermal conductivity).

[0071] 3) Strategy Verification Sandbox: Provides a virtual testing environment for reinforcement learning, capable of evaluating over 200 control strategy combinations per second; rapidly simulating system responses under extreme operating conditions (e.g., a transient temperature rise of 15°C); and proactively identifying potential control risks (e.g., resonance caused by electromagnetic interference).

[0072] 4) Knowledge accumulation carrier: Stores a complete digital image of historical test scenarios (parameter combination + control effect); constructs a traceable cause-and-effect relationship map (such as the correlation between excessive vibration and valve opening); supports strategy interpretation based on physical rules (displaying the simulation basis for control decisions).

[0073] For example, when a sudden change in battery temperature is detected, the digital twin simultaneously executes the following actions: Invoking an electrochemical model to predict thermal runaway trends; testing the stability of three cooling strategies in virtual space; and transmitting the optimal strategy parameters (82% liquid cooling flow + 55% TEC power) to the actuator. Analysis shows that this strategy saves 23% energy compared to traditional methods.

[0074] In some embodiments of the present invention, the actuator group includes a temperature control module, an air pressure regulating valve and an active shock absorption platform. The temperature control module includes a variable frequency semiconductor temperature control piece connected in parallel with a liquid cooling circuit and distributed on the four walls of the pod; the air pressure regulating valve is a butterfly valve driven by a multi-stage stepper motor and installed in the air intake / exhaust channel at the top of the pod; the active shock absorption platform is a six-degree-of-freedom electromagnetic actuator integrated in the base at the bottom of the pod.

[0075] It's easy to understand that the variable-frequency semiconductor temperature controller works in conjunction with the liquid cooling circuit to adjust the pod's temperature; opening and closing the air pressure regulating valve adjusts the cabin's air pressure; and controlling the active vibration damping platform adjusts vibration. Adjustments to each component are executed based on control commands sent by the central control system.

[0076] In some embodiments of the present invention, a fault-tolerant module may also be included. The fault-tolerant module can be used to: detect sensor anomalies in real time, and switch redundant sensors when the data deviates by ≥3σ; when the actuator fails, reallocate the control quantity based on the coupling model; and periodically evaluate the system health and trigger an alarm when the confidence level is <95%.

[0077] Specifically,

[0078] In some embodiments of the present invention, the system supports multi-level control modes, namely: local fast response mode, in which the FPGA directly processes emergency signals; global optimization mode, in which the GPU integrates digital twin data to generate a full-parameter control plan; and manual intervention mode, in which the control weight parameters are adjusted through a visual interface.

[0079] According to an embodiment of the present invention, a method for collaboratively controlling environmental parameters based on the above system includes:

[0080] S1, collects environmental parameters in real time through a multi-source sensor array, synchronizes timestamps and transmits them to the central controller;

[0081] S2, the digital twin interface receives the virtual model prediction data and compares it with the measured data in multiple dimensions;

[0082] S3. The central controller dynamically adjusts the parameter combination of the actuator group according to the difference analysis results;

[0083] S4. Record the test object response data and update the control strategy knowledge base.

[0084] For example, the parameter adjustment in step S3 includes: when a sudden change in the temperature gradient is detected, synchronously adjusting the air pressure valve opening and the damping coefficient of the shock-absorbing platform; when the vibration spectrum characteristics exceed the threshold, triggering the temperature control module to reduce the frequency.

[0085] In some embodiments of the present invention, a fault self-diagnosis process is further included. The fault self-diagnosis process is triggered when an abnormality is detected. The fault self-diagnosis process includes:

[0086] Implement triple verification of abnormal sensor data: adjacent sensor comparison, digital twin prediction value verification, and historical data trend analysis;

[0087] The actuator failure compensation strategy includes: concentrating the temperature control task on the non-faulty area and adjusting the air pressure valve to maintain thermal balance.

[0088] In some embodiments of the present invention, the PPO algorithm is used to perform strategy optimization in step S3, generating N groups of candidate actions in each iteration; the effects of the candidate actions are evaluated through rapid simulation of digital twins; and the action combination with the highest comprehensive score is selected for execution.

[0089] Example 1: Aerospace Testing

[0090] Scenario: Simulates sudden changes in air pressure (1MPa→0.005MPa) and temperature fluctuations (-50℃→70℃) during aircraft takeoff and landing.

[0091] Implementation steps:

[0092] a. A multi-source sensor array monitors the cabin pressure gradient change rate in real time and transmits the collected information to a central controller at 1ms intervals. This information is used to obtain the cabin pressure gradient (dP / dt) and temperature change rate (dT / dt), and then performs Kalman filtering and noise reduction processing using an FPGA.

[0093] b. The reinforcement learning model is based on the PPO (Proximal Policy Optimization) algorithm, dynamically adjusting the duty cycle (0-100%) and air pressure valve opening (0-90°) of the temperature control module to control the temperature and pressure curve to a deviation of ≤0.3% from the GJB150A-2009 standard;

[0094] c. The vibration module generates a random excitation signal based on the turbulence spectrum (provided by the digital twin), with a frequency range of 5-2000 Hz. It is synchronized with the temperature and pressure parameters in real time through a coupling model to achieve multi-dimensional real-time control of the pod's environmental parameters.

[0095] Example 2: New Energy Vehicle Battery Testing

[0096] Scenario: Simulate battery performance degradation test under cycle shock from -55℃ to 70℃.

[0097] Implementation steps:

[0098] a. Build a battery thermodynamic simulation model based on COMSOL Multiphysics and import real-time electrochemical parameters through the digital twin interface;

[0099] b. Based on the battery surface temperature feedback, the temperature control wind speed and refrigerant flow are dynamically adjusted to ensure the temperature gradient is ≤2°C / s. Specifically, the temperature control wind speed is achieved through PID + feedforward control, and the refrigerant flow is adjusted by a proportional valve driven by a stepper motor. The temperature gradient control accuracy reaches ±0.2°C / s.

[0100] c. Record the control strategy optimization process and generate a reusable test knowledge base. The knowledge base uses a graph database (Neo4j) to store control strategies and test results, supporting fast strategy matching based on similarity retrieval.

[0101] In summary, the system of the present invention achieves global optimization of environmental parameters through multi-parameter collaborative control: breaking through the limitations of independent regulation of a single parameter, and realizing global optimization of environmental parameters through coupling models; capable of real-time dynamic response: using predictive control, combined with digital twins to predict the environmental state in the next 3-5 seconds, and triggering the actuator in advance; possessing self-learning capabilities: the system can automatically generate customized control schemes according to different test scenarios (such as extreme cold, high altitude, and electromagnetic interference environments). In addition, it also has fault self-diagnosis: a built-in anomaly detection model identifies sensor drift or actuator failure, and initiates seamless switching of redundant modules. Ultimately, the coordination of various components brings the following advantages:

[0102] First, multi-parameter coordinated control reduces the traditional step-by-step testing time by approximately 40%, improving testing efficiency;

[0103] Second, the DNN+RL algorithm optimizes the power consumption of actuators, reducing energy consumption by 15%-30%.

[0104] Third, the environmental parameter control error is ≤0.5% (traditional systems are about 2%-5%), which improves accuracy;

[0105] Fourth, it supports new sensor types and control targets (such as adding a sand and dust concentration simulation module) and has strong scalability.

[0106] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to the embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the claims and their equivalents.

Claims

1. A dynamic test pod environmental parameter coordinated control system, characterized in that: include: Multi-source sensor array, including a variety of sensors for real-time collection of various environmental parameters in the pod, with a sampling frequency of ≥10kHz, and the collected data is synchronized by time stamps; The central controller is used to integrate sensor data and digital twin prediction information in real time through an FPGA+GPU heterogeneous architecture, dynamically generate and optimize control strategies based on deep reinforcement learning algorithms, and synchronously coordinate the actuator group to achieve multi-physics field coupling control; The digital twin interface is used to synchronize the bidirectional data flow between the physical test environment and the digital twin interface in real time, predict the evolution trend of environmental parameters based on multi-physics field coupling calculations, and correct model parameters through reinforcement learning feedback closed loop; The actuator group is used to adjust the environmental parameters according to the control instructions.

2. The system according to claim 1, wherein: The central controller includes: Real-time signal preprocessing module, deployed on FPGA, used to implement Kalman filtering noise reduction and feature extraction of sensor data; A multi-parameter collaborative control module, deployed on the GPU, is used to run the DNN policy generator including the attention mechanism to dynamically generate control policies; The digital twin comparison module is used to perform time series alignment and difference analysis between sensor measured data and simulation prediction data.

3. The system according to claim 2, characterized in that The working method of the multi-parameter collaborative control module includes: Establish an environmental parameter coupling matrix to quantify the physical correlation between temperature, humidity, air pressure and vibration; The mapping relationship between historical control strategies and test object responses is learned through the LSTM network; The control parameter combination is optimized in real time based on the PPO algorithm and output to the actuator group.

4. The system according to claim 1, wherein: The digital twin interface implementation method includes: Based on the FMI standard package multi-physics simulation model, it supports thermal-fluid-solid coupling calculation; A sliding window algorithm is used to align measured and simulated data streams, and the window width can be dynamically adjusted; Identify deviation patterns and generate remediation strategies through a variance analysis engine.

5. The system according to claim 1, wherein: The actuator group includes a temperature control module, an air pressure regulating valve and an active shock absorption platform. The temperature control module includes a variable frequency semiconductor temperature control plate connected in parallel with a liquid cooling circuit and distributed on the four walls of the pod; the air pressure regulating valve is a butterfly valve driven by a multi-stage stepper motor and installed in the air intake / exhaust channel at the top of the pod; the active shock absorption platform is a six-degree-of-freedom electromagnetic actuator integrated into the base at the bottom of the pod.

6. The system according to claim 1, wherein: It also includes a fault tolerance module, wherein the fault tolerance module is used to: Real-time detection of sensor anomalies, switching to redundant sensors when data deviates by ≥3σ; When the actuator fails, the control quantity is redistributed based on the coupling model; Periodically evaluate system health and trigger an alarm when the confidence level is less than 95%.

7. The system according to claim 1, wherein: The system supports multi-level control modes. Local fast response mode, where the FPGA directly processes emergency signals; Global optimization mode, where GPU integrates digital twin data to generate a full parameter control plan; Manual intervention mode, adjust the control weight parameters through the visual interface.

8. A method for collaboratively controlling environmental parameters based on the system of any one of claims 1 to 7, characterized in that: include: S1, collects environmental parameters in real time through a multi-source sensor array, synchronizes timestamps and transmits them to the central controller; S2, the digital twin interface receives the virtual model prediction data and compares it with the measured data in multiple dimensions; S3. The central controller dynamically adjusts the parameter combination of the actuator group according to the difference analysis results; S4. Record the test object response data and update the control strategy knowledge base.

9. The method according to claim 8, characterized in that The process also includes a fault self-diagnosis process, which is triggered when an abnormality is detected and includes: Implement triple verification of abnormal sensor data: adjacent sensor comparison, digital twin prediction value verification, and historical data trend analysis; The actuator failure compensation strategy includes: concentrating the temperature control task on the non-faulty area and adjusting the air pressure valve to maintain thermal balance.

10. The method according to claim 8, characterized in that In step S3, the PPO algorithm is used to perform strategy optimization, generating N groups of candidate actions in each iteration; the effects of the candidate actions are evaluated through rapid simulation of digital twins; and the action combination with the highest comprehensive score is selected for execution.

Images