Standard container temperature control method based on digital twinning

Abstract

The invention discloses a standard container temperature control method based on digital twinning, and the method comprises the following steps: collecting the temperature data of a specific region of the wall surface of a standard container through a plurality of temperature sensors disposed on the wall surface of the standard container; recording physical parameters of the standard container; a control system is established for the standard container and used for accurately adjusting the temperature in the standard container; establishing a control strategy library for dynamically selecting a matched control method from predefined control strategies; establishing a physical simulation model of the standard container based on the obtained physical parameters; based on a data assimilation technology, real-time sensor data and a physical simulation model are fused; based on the digital twin model, establishing a prediction model by using a physical information neural network, and establishing a data set for training the prediction model; and the physical information neural network takes the position and the temperature value of the wall sensor as input, and outputs local temperature field distribution and overall average temperature in the standard container.

Description

Technical Field

[0001] The present invention relates to the technical field of temperature control and precision measurement, and particularly relates to a temperature control method and device for a PVTt experimental standard container based on digital twin, which is applicable to the high-precision and dynamic regulation of the temperature of the standard container in a gas volume-pressure-temperature-time (PVTt) measurement system, aiming to improve the measurement accuracy, temperature control efficiency and system intelligence level. Background Art

[0002] The PVTt measurement method is a high-precision measurement technology widely used in gas flow standard devices. The core principle is to deduce the gas flow or gas quantity by monitoring the relationships of the volume, pressure and temperature of the gas in a closed standard container changing with time. In PVTt experiments, the temperature control of the standard container is crucial. Temperature fluctuations will directly affect the gas state parameters, thereby introducing measurement errors and reducing the accuracy and stability of the measurement system.

[0003] Currently, the temperature control of the standard container in the PVTt system usually relies on a single-point temperature sensor and traditional feedback control (such as PID control). However, due to the large size of the container, strong thermal inertia, and the influence of environmental changes (such as room temperature fluctuations, radiation interference, etc.), single-point measurement often cannot comprehensively reflect the temperature distribution state inside the container. In addition, the sensor itself may be affected by environmental noise, electromagnetic interference, etc., resulting in measurement deviations. Traditional control methods are prone to problems such as response lag, insufficient temperature control accuracy and increased energy consumption when facing the complex thermal dynamics inside the container.

[0004] Digital Twin technology brings new solutions to the temperature control of the standard container. By constructing a high-fidelity virtual model corresponding to the physical standard container, real-time fusing multi-source sensor data, and combining physical modeling with data-driven analysis, it is possible to realize the dynamic mapping, state prediction and optimization decision-making of the container temperature field. Compared with traditional methods, digital twin can not only capture the internal temperature changes of the container in real time, but also perform predictive simulation and adaptive control optimization in the virtual space, effectively improving the response speed and adjustment accuracy of temperature control, and significantly enhancing the overall measurement performance of the PVTt system.

[0005] Based on this, the present invention proposes a temperature control method for a standard container based on digital twin, aiming to provide high-reliability, high-precision and intelligent temperature control support for the PVTt measurement system, and ensuring the accuracy of measurement data and the stability of system operation. Summary of the Invention

[0006] The present application provides a temperature control method for a standard container based on digital twin, aiming to achieve accurate prediction and rapid regulation of the average temperature inside the standard container. The present application adopts the following technical solutions:

[0007] A standard container temperature control method based on digital twin, comprising the following steps:

[0008] Collect temperature data of a specific area on the wall surface of the standard container through several temperature sensors installed on the wall surface of the standard container, and preprocess the collected temperature data;

[0009] Record the physical parameters of the standard container, including the contour dimensions, internal structure, material properties, and external environment parameters of the standard container, for subsequent digital twin model construction;

[0010] Establish a control system for the standard container to achieve precise regulation of the internal temperature of the standard container; establish a control strategy library for dynamically selecting a matching control method from predefined control strategies;

[0011] Based on the obtained physical parameters, establish a physical simulation model of the standard container. Describe the movement of the fluid through the momentum conservation equation, describe the pressure field distribution through the Poisson equation, and use the heat transfer differential equation to characterize the evolution process of the temperature field; this physical simulation model is spatially divided into multiple discrete grid cells, and it is calculated using numerical solution methods to realize the reproduction of the internal thermodynamic behavior of the standard container;

[0012] Based on data assimilation technology, fuse real-time sensor data with the physical simulation model, continuously correct and optimize the physical model, and form a dynamically updated and real-time mapped digital twin model, thereby improving the modeling accuracy and real-time response ability;

[0013] To achieve precise prediction and rapid response of the internal temperature of the standard container, based on the digital twin model, use a physics-informed neural network (PINN) to establish a prediction model, and establish a data set for training the prediction model;

[0014] The physics-informed neural network takes the wall sensor position and its temperature value as inputs, and outputs the local temperature field distribution T pred (x, y, z) and the overall average temperature T avg,pred ; During the model training process, embed the heat conduction equation into the loss function to effectively guide the network to follow physical laws such as energy conservation while maintaining the data fitting accuracy; through forward propagation and backward propagation, iteratively optimize the network weights to achieve the prediction of the internal temperature field and average temperature of the standard container;

[0015] According to the deviation between the predicted temperature and the set target temperature and the current temperature fluctuation characteristics, dynamically select a matching control method from the predefined control strategy library to achieve stable regulation and rapid response of the temperature.

[0016] Further, the preprocessing of the collected temperature data includes data filtering, denoising, normalization, and outlier detection.

[0017] Further, the profile dimensions of the standard container include length, width, height, and wall thickness, the internal structure includes the position and dimensions of the partition and the bracket, the material properties include thermal conductivity, specific heat capacity, and density, and the external environmental parameters include water bath temperature and humidity.

[0018] Further, the data source for training the prediction model is: collecting the temperature data, time, and spatial coordinate parameters of multiple temperature sensors outside the wall of the standard container, and the average temperature inside the standard container, where the average temperature inside the standard container is obtained through simulation or sensor measurement.

[0019] Further, the loss function consists of data loss L data , physical loss L physics and target loss L avg .

[0020] The beneficial effects of this application include: The temperature control method for the standard container based on digital twin provided by this application can obtain the environmental parameter information of the environment where the standard container is located and the information of the standard container itself; based on the environmental parameter information and the standard container information, a high-precision virtual model of the standard container can be constructed through digital twin technology, and combined with machine learning algorithms to achieve accurate prediction of the average temperature inside the container; according to the predicted temperature, select the optimal control strategy and control parameters to dynamically adjust the output power of the heating tube, thereby accurately controlling the water bath temperature, improving the accuracy and response speed of temperature regulation inside the standard container. It enhances the adaptability of the standard container to complex working conditions and provides an efficient and intelligent temperature control solution for the standard container in the PVTt measurement scenario. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] To more clearly illustrate the technical solutions of the embodiments of this application, the following briefly introduces the drawings required in the embodiments. The following drawings only show some embodiments of this application and should not be regarded as limiting the scope. For those of ordinary skill in the art, other relevant drawings can be obtained based on these drawings without creative efforts.

[0022] Figure 1 Schematic diagram of the process of a temperature control method for a standard container based on digital twin in the present invention

[0023] Figure 2 Schematic diagram of the modules of a temperature control and regulation device based on digital twin in the present invention

[0024] Figure 3 Schematic diagram of the structure of a typical computer device in the present invention Specific Embodiments

[0025] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the following will clearly and completely describe the technical solutions in the embodiments of this application with reference to the accompanying drawings in the embodiments of this application. The described embodiments are some, but not all, of the embodiments of this application. Components of the embodiments of this application described and illustrated herein are generally arranged and designed in different configurations.

[0026] Please refer to Figure 1 As shown, a standard container temperature control method based on digital twin in this embodiment includes the following steps 101 to 108.

[0027] Step 101, since standard containers usually adopt integral molding technology to ensure a low leakage rate and it is generally not easy to install temperature sensors inside, in the embodiments of this application, the temperature of a specific area on the wall of the standard container is measured by several temperature sensors installed on the wall of the standard container, and the data collected by the sensors is transmitted to the control system in real time through a data acquisition system. The data is first filtered to reduce the interference of high-frequency noise on the temperature signal.

[0028] Specifically, a moving average filtering method is used for filtering operation, and its calculation formula is:[[]]

[0029]

[0030] where T(j) is the original temperature data at the j-th sampling point, and N is the size of the moving window. The temperature sequence is smoothed by moving average filtering, effectively suppressing instantaneous measurement noise and improving data stability.

[0031] Subsequently, the filtered data is normalized to eliminate the influence brought by different sensor measurement scales.

[0032] Specifically, linear normalization method is adopted for normalization, and its calculation formula is:[[]]

[0033]

[0034] where T filtered is the filtered temperature data, T min and T max are the minimum and maximum values in the sample data respectively. After normalization, the data range is limited to the interval [0,1], which improves the convergence speed and stability of the machine learning model.

[0035] Exemplarily, in this embodiment, considering the temperature gradient distribution formed vertically inside the standard container due to the gravity effect, the standard container is divided into several layers from top to bottom, and a plurality of temperature sensors are evenly arranged on the outer wall of each layer. In addition, the temperature sensors can also be arranged in a face center point distribution, a regular grid distribution, etc.

[0036] Step 102, obtain the physical parameters of the standard container through methods such as CAD drawings, 3D scanning, experimental measurement, or simulation calculation, and input the contour dimensions of the standard container (such as length, width, height, wall thickness), internal structure (such as the position and size of the partition layer and bracket), material properties (such as thermal conductivity, specific heat capacity, density), and external environment parameters (such as water bath temperature, humidity, etc.) for subsequent digital twin model construction.

[0037] Specifically, the standard container adopted in this embodiment has an integrally formed structure. The total length of the main body is 309 mm, the length of the main cylindrical section is 145 mm, and spherical transition sections with a radius of 40 mm are provided at both ends. The wall thickness is 4.35 mm in the main cylindrical section area and 6.35 mm in the connecting flange area. There are through holes with diameters of 8 mm and 6 mm on the outside of the container for installing temperature sensors or connecting other external devices. The whole standard container is made of stainless steel, with a thermal conductivity of 16.2 W / (m·K), a specific heat capacity of 500 J / (kg·K), and a density of 7800 kg / m 3 . During the experiment, the water bath temperature is set to 293 K, the environmental relative humidity is 50%, and the device operates under standard atmospheric pressure.

[0038] Step 103, after obtaining the parameters of the standard container, use methods such as finite element analysis and computational fluid dynamics to establish a numerical simulation model, which serves as the core physical basis of the digital twin system to realize the dynamic mapping of the internal temperature field and heat conduction process of the real standard container.

[0039] Specifically, the motion of the fluid is described by the momentum conservation equation, and the equation is described as:

[0040]

[0041] where ρ l is the fluid density, is the fluid velocity vector, p is the pressure, μ is the fluid dynamic viscosity, is the body force term (such as gravity, magnetic force).

[0042] Specifically, the pressure distribution of the flow field is described by the Poisson equation, and the equation is described as:

[0043]

[0044] According to the requirements of modeling accuracy, the three-dimensional model of the entire container will be divided into multiple discrete grid cells. The grid type can be selected as structured or unstructured, and combined with a local refinement strategy to improve the simulation accuracy in the thermal gradient region. On the discretized computational domain, numerical methods are used to solve the heat transfer differential equation to obtain the temperature distribution evolving with time at each position inside the container. The heat conduction control equation adopted is:

[0045]

[0046] where ρ is the material density, c p is the specific heat capacity, k is the thermal conductivity, Q is the volume heat source term, and T is the temperature.

[0047] Step 104, to improve the prediction accuracy, the system uses a method combining a digital twin model and machine learning for temperature prediction. The machine learning model can adopt algorithms such as long short-term memory network, convolutional neural network, or gradient boosting decision tree. Through the training of historical data, it learns the spatio-temporal variation law of the standard container temperature to achieve fast and efficient temperature prediction.

[0048] Exemplarily, in this embodiment, a physics-informed neural network (PINN) is used to predict the temperature inside the standard container, and Huawei MindFlow is used to construct and train the network.

[0049] First, experiments are conducted to obtain an experimental dataset, which includes the data of multiple temperature sensors outside the wall of the standard container, as well as the time, space coordinate parameters, and the average temperature inside the standard container. The average temperature inside the standard container is obtained through simulation or sensor measurement. To improve the training effect of the physics-informed neural network model, the data is standardized or normalized.

[0050] The input of the physics-informed neural network is the position and temperature of the sensors outside the wall, and the output is the predicted average temperature of the standard container. The network structure adopts a multi-layer feedforward neural network, including an input layer, a hidden layer, and an output layer. The neurons in the input layer correspond to the three-dimensional coordinates and temperature information of the sensors. The hidden layer contains multiple fully connected layers, each layer contains several neurons, and the ReLU activation function is used to capture non-linear complex relationships. The output layer simultaneously includes the local temperature field distribution T pred (x, y, z) and the overall average temperature T avg,pred .

[0051] The physics-informed neural network ensures that the prediction results conform to physical laws by embedding physical equations into the loss function. The loss function includes data loss, physical loss, and target loss.

[0052] Specifically, the data loss ensures that the wall temperature predicted by the model is consistent with the measured data. The function expression is as follows:

[0053]

[0054] where N is the temperature measurement point on the surface of the standard container, T pred (x i , y i , z i ) is the i-th predicted temperature, T meas (x i , y i , z i ) is the i-th measured temperature. Force the model to accurately match the observed data at the known measurement points to establish a direct association between the input and the output.

[0055] Specifically, the physical loss represents the residual of the physical equation, and the functional expression is as follows:

[0056]

[0057] where M is the number of sampling points set internally. After adding the physical loss, ensure that the predicted temperature and the temperature outside the wall satisfy the heat conduction equation and conform to physical laws such as energy conservation.

[0058] Specifically, the target loss represents the error between the predicted average temperature and the true average temperature, and the expression is as follows:

[0059]

[0060] where is the internal average temperature measured through simulation. The final loss function is a weighted combination of the three:

[0061] L total = αL data + βL physics + γL avg

[0062] where α, β, and γ are weight coefficients. In this embodiment, α = 1, β = 0.1, and γ = 10.

[0063] In this embodiment, the training process of the physics-informed neural network is carried out in a supervised learning manner. First, calculate the model output through forward propagation, including the local temperature distribution T pred (x, y, z) and the overall average temperature T avg,pred . Subsequently, in the backpropagation stage, combined with the automatic differentiation technique, solve the gradient of the total loss function L total with respect to the network weight parameters, and use the gradient descent algorithm to iteratively update the network parameters until the loss converges to the set threshold.

[0064] During the training process, the physical constraint term is embedded in the loss function as a strong regularization term to ensure that the network can still output physically consistent prediction results in areas lacking internal temperature measurement points. This mechanism effectively suppresses model overfitting and improves the generalization ability in unobserved areas. The finally trained PINN model, while accurately fitting the temperature data of the sensors on the outer wall of the container, the predicted internal temperature field of the container strictly satisfies the partial differential equation of heat conduction, and can stably and reliably infer the average temperature of the standard container.

[0065] Compared with traditional numerical simulation methods, the PINN method in this embodiment has obvious advantages. On the one hand, traditional CFD or finite element methods have large computational amounts and long calculation times, making it difficult to meet the real-time requirements of temperature prediction for the temperature control system of standard containers; after the training of this model, its inference speed can reach milliseconds to seconds, greatly improving the system response efficiency. On the other hand, traditional methods rely on complete initial conditions and boundary settings, and the scalability and adaptability of the model are weak; while the PINN model can achieve adaptive prediction under different boundary conditions by integrating real-time sensor data and physical laws, and has good robustness and generalization performance, especially suitable for working conditions with complex temperature control requirements in practical application scenarios such as PVTt measurement systems.

[0066] Step 105, select a suitable control strategy from the predefined control strategy library according to the comparison between the predicted temperature and the set temperature and the temperature fluctuation situation. For example, when the system has a delay effect and is affected by multiple factors, the model predictive control method can be adopted; when an accurate mathematical model cannot be established and the system characteristics change with time, the fuzzy adaptive PID control method can be adopted. Various control methods and control parameters will be simulated and compared in the digital twin model, and the evaluation indexes include response time, overshoot, and steady-state error, etc. The system will select the optimal control strategy and control parameters.

[0067] Step 106, after determining the control strategy and control parameters, the system issues specific control instructions to the actuator through the control bus to achieve precise regulation of the temperature inside the standard container. In this embodiment, when the temperature inside the container predicted by the digital twin model is higher than the set target temperature, the control system will automatically activate the compression refrigeration circuit, start the refrigerant circulation, and adjust the refrigerant flow through the solenoid valve to quickly reduce the water bath temperature; conversely, when the predicted temperature is lower than the set value, the system controls the electric heating tube to be powered on for heating and adjusts the heating power in a PWM manner to steadily raise the temperature and avoid large temperature fluctuations. In addition, to further optimize the heat exchange efficiency and temperature uniformity, the system can also dynamically adjust the fan speed according to the temperature field distribution, control the flow rate and disturbance intensity of the air around the container, and achieve rapid diffusion of local heat and balanced distribution of the overall system temperature. All execution commands and feedback signals during the operation of the control system will be transmitted back to the digital twin platform in real time for updating the control status and correcting the system model to ensure the continuous effectiveness of the control strategy and the stability of the system operation.

[0068] Step 107, after performing the temperature regulation, the system monitors the container temperature again and compares it with the set value. If the temperature still does not reach the target value, the system returns to Step 105 to readjust the control strategy and control parameters until the set temperature range is reached. If the temperature has stabilized within the target range, the system proceeds to the next step.

[0069] Step 108, when the container temperature reaches the preset value, the system enters the stage of maintaining temperature stability. In this stage, the system will continuously monitor the temperature and appropriately adjust the control parameters to cope with possible environmental changes or the influence of heat exchange inside the container. The system will also store and analyze the temperature data for a long time to optimize the control strategy, improve energy efficiency, and reduce energy consumption.

[0070] Please refer to Figure 2 as shown in Figure 2 is a schematic diagram of the modules of a digital twin-based standard container temperature regulation device for implementing this application, including modules 201 to 204.

[0071] The data measurement and transmission module 201. This module collects the wall temperature of the standard container in real time through temperature sensors, processes the raw data such as filtering, denoising, and normalization to improve the data quality, transmits information such as the standard container contour parameters and internal structure to the control system, transmits the processed data to the cloud or local server through the Internet of Things or communication protocol, and performs storage backup for subsequent module calls.

[0072] The twin model construction module 202 creates a three-dimensional geometric model based on the size, shape, and internal structure of the standard container, divides the grid structure, integrates thermodynamic equations (such as the heat conduction equation and the energy conservation equation) and material properties (such as thermal conductivity and specific heat capacity), establishes a dynamic simulation model of the container temperature distribution, and dynamically adjusts the model parameters according to real-time data to ensure the synchronization of the virtual model and the actual container state.

[0073] The temperature prediction module 203 uses a physics-informed neural network model to predict the average temperature inside the standard container. In the model, the heat conduction equation is embedded in the loss function so that the final predicted value of the average temperature strictly follows the heat transfer law.

[0074] The control module 204, based on the temperature prediction result and the target set value, selects the most suitable control method for the current working condition from a predefined control strategy library or combines control strategies, and switches the control strategy according to the change of the working condition. Then, according to the instructions of the control strategy, it controls the output power of the compressor and the heating tube to achieve temperature regulation, monitors the operating states of the compressor and the heating tube through devices such as a speed sensor and a temperature sensor, feeds back the information to dynamically adjust the temperature control parameters, and at the same time triggers a fault alarm when abnormal power output of the compressor and the heating tube is detected.

[0075] In summary, the modular design of the standard container temperature control device based on digital twin makes it more flexible and efficient, capable of flexibly meeting the temperature control requirements under different working conditions, and further improving the overall operability and efficiency.

[0076] Please refer to Figure 3 as shown in Figure 3 a schematic structural diagram of a typical computer device for implementing the specific implementation method of the data twin-based standard container temperature control method of the present application. The computer device at least includes the following parts: CPU (Central Processing Unit) 302, RAM (Random Access Memory) 303, ROM (Read Only Memory) 304, system bus 301, storage management 305, I / O management 306, and communication management 307.

[0077] Figure 3 The structure block diagram only shows the structure of a computer device that can implement various embodiments of the present invention, and is not a limitation on the practice of the present invention. In some cases, some devices in this structure can be added or reduced according to needs, and they can also be interconnected through a network to provide the operating environment applicable to various specific implementation manners of the present invention. For example, various modules and steps of the present invention can be distributedly implemented in various interconnected computer devices.

[0078] In the method for controlling the temperature of a standard container based on digital twin provided in this application, the part involving software logic can be implemented using a programmable logic device or implemented as a computer program product that causes a computer to execute the method as demonstrated. The computer program product includes a computer-readable storage medium. The computer-readable storage medium can be an internal medium installed in the computer or a removable medium detachable from the computer main body, including various media that can store program codes such as USB flash drives, external hard drives, read-only memories, random access memories, magnetic disks, or optical discs.

[0079] Although the present invention has been disclosed above with preferred embodiments, the above embodiments are not intended to limit the present invention. For any person skilled in the art, without departing from the scope of the technical solution of the present invention, many possible changes and modifications can be made to the technical solution of the present invention by using the technical content disclosed above, or modified into equivalent embodiments with equivalent changes. Therefore, any simple modification made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention still falls within the scope of protection of the technical solution of the present invention.

Claims

1. A temperature control method for standard containers based on digital twin, comprising the following steps: Collect temperature data of a specific area on the wall surface of the standard container through a number of temperature sensors installed on the wall surface of the standard container, and preprocess the collected temperature data; Record the physical parameters of the standard container, including the contour dimensions, internal structure, material properties, and external environment parameters of the standard container, for subsequent digital twin model construction; Establish a control system for the standard container to achieve precise regulation of the internal temperature of the standard container; establish a control strategy library for dynamically selecting a matching control method from predefined control strategies; Based on the obtained physical parameters, establish a physical simulation model of the standard container. Describe the movement of the fluid through the momentum conservation equation, describe the pressure field distribution through the Poisson equation, and use the heat transfer differential equation to depict the evolution process of the temperature field; this physical simulation model is spatially divided into multiple discrete grid cells, and it is calculated using numerical solution methods to reproduce the internal thermodynamic behavior of the standard container; Based on data assimilation technology, fuse real-time sensor data with the physical simulation model, continuously correct and optimize the physical model to form a dynamically updated and real-time mapped digital twin model, thereby improving the modeling accuracy and real-time response ability; To achieve precise prediction and rapid response of the internal temperature of the standard container, based on the digital twin model, use a physics-informed neural network (PINN) to establish a prediction model, and establish a data set for training the prediction model; The physical information neural network takes the positions of the wall sensors and their temperature values as inputs, and outputs the local temperature field distribution T pred (x, y, z) and the overall average temperature T avg,pred ; during the model training process, the heat conduction equation is embedded in the loss function, and the loss function consists of the data loss L data , the physical loss L physics and the target loss L avg , effectively guiding the network to follow physical laws such as energy conservation while maintaining the data fitting accuracy; through forward propagation and backward propagation, the network weights are iteratively optimized to achieve the prediction of the temperature field and the average temperature inside the standard container; According to the deviation between the predicted temperature and the set target temperature and the current temperature fluctuation characteristics, dynamically select a matching control method from the predefined control strategy library to achieve stable regulation and rapid response of the temperature.

2. The standard container temperature control method according to claim 1, characterized in that, Preprocessing the collected temperature data includes data filtering, denoising, normalization, and outlier detection.

3. The standard container temperature control method according to claim 1, wherein The contour dimensions of the standard container include length, width, height, and wall thickness, the internal structure includes the position and size of the partition layer and brackets, the material properties include thermal conductivity, specific heat capacity, and density, and the external environment parameters include water bath temperature and humidity.

4. The standard container temperature control method according to claim 1, characterized in that The data source for training the prediction model is: collect the temperature data, time and space coordinate parameters of multiple temperature sensors outside the wall of the standard container, and the average temperature inside the standard container, and the average temperature inside the standard container is obtained through simulation or sensor measurement.

5. The standard container temperature control method according to claim 1, characterized in that, The loss function consists of a data loss L data , a physical loss L physics , and a target loss L avg .

Images