Anisotropic material heat conductivity coefficient measuring device and operation method thereof

Abstract

The invention discloses an anisotropic material heat conductivity coefficient measuring device and an operation method thereof.The measuring device comprises a box body, a weight sensor, an electric heating wire, a temperature sensor, a main control chip and an upper computer, the weight sensor is arranged between an outer shell and a cylindrical barrel and used for weighing the weight of the cylindrical barrel, and the electric heating wire is arranged in the box body; the electric heating wire surrounds the outer wall of the cylindrical barrel, the temperature sensors are circumferentially arranged in the cylindrical barrel at equal intervals and diverge from the center of the cylindrical barrel to the edge of the cylindrical barrel, and the main control chip is electrically connected with the weight sensor, the electric heating wire and the temperature sensors. The main control chip is used for collecting data signals of the weight sensor and the temperature sensor, inputting the data signals into a built-in physical information neural network and calculating and measuring the anisotropic heat conductivity coefficient of the to-be-measured material, and the upper computer is electrically connected with the main control chip, obtains data transmitted by the main control chip and sends a control instruction to the main control chip; according to the invention, the heat conductivity coefficient of the material with anisotropic characteristics can be measured and analyzed more accurately.

Description

Technical Field

[0001] The invention relates to the technical field of thermal conductivity measurement, and in particular to a device for measuring thermal conductivity of anisotropic materials and an operating method thereof. Background Art

[0002] The thermal conductivity and specific heat of a material are key parameters that determine its heat transfer characteristics and are also important research topics in physics and engineering thermodynamics. When studying the thermal conductivity of a material, common measurement methods are usually divided into two categories: steady-state method and dynamic method.

[0003] The measurement principle of the steady-state method is based on the one-dimensional heat conduction equation (where λ represents thermal conductivity, s represents the cross-sectional area of ​​the material, Q represents the amount of heat transferred, and T represents temperature). The typical structure of this type of device is to clamp the sample between a heating plate above and a cooling plate below, with the three in close contact under a certain pressure. During the experiment, the sample needs to be heated to a specific temperature and kept constant until the internal temperature field stabilizes before measurement. By measuring the steady-state temperatures T1 and T2 on the upper and lower surfaces of the sample, the temperature gradient can be obtained. where h is the thickness of the sample. Subsequently, the heat dissipation rate is calculated by recording the temperature change of the sample during cooling. Combined with the cross-sectional area s of the sample, the data is substituted into the Fourier heat conduction equation to inversely solve the thermal conductivity λ.

[0004] Although the steady-state method is often used to measure the thermal conductivity of poor conductors of heat, the method has cumbersome experimental steps and the measurement results are easily affected by the accuracy of the heat flow. Since different heat transfer models or fitting methods may cause deviations in data processing, heat flow measurement is often accompanied by systematic errors. Although a variety of improvement measures have been adopted in the prior art, such as using a host computer for data processing to reduce human errors, calculating heat flow by the electrothermal power method, or using a circulating water system to ensure steady-state heat flow, these methods are still difficult to completely eliminate errors in measurements.

[0005] For wood, its thermal conductivity is extremely sensitive to moisture content. Slight changes in moisture content will significantly affect the thermal conductivity of wood. The steady-state measurement method usually requires a long heating process, which may not only change the moisture content of the wood, but also cause the formation of humidity gradients inside the sample, further interfering with the accurate determination of thermal conductivity. Since the thermal conductivity of materials such as wood is highly dependent on their inherent physical and chemical properties, existing steady-state measurement technology often cannot avoid large measurement errors when dealing with these problems. The combined effect of the above factors makes it difficult for existing technologies to accurately and reliably measure the thermal conductivity of wood and other materials that are significantly affected by chemical properties.

[0006] The starting point of the measurement principle of the dynamic method device is the transient heat conduction equation (where represents specific heat, ρ represents density). By applying a short or periodic heat pulse to a specific part of the sample, the surface temperature of the sample rises and gradually diffuses into the interior. The temperature sensor collects the data of the temperature change over time in other areas of the sample, and combined with the transient heat conduction equation, the thermal diffusivity (also known as thermal conductivity) of the sample can be calculated. If the specific heat and density of the sample material are known, its thermal conductivity can be further calculated.

[0007] The dynamic method for measuring thermal conductivity uses the transient heat conduction equation, considers the thermal diffusion characteristics of the sample, and matches the experimental data with the theoretical model through numerical fitting technology to accurately calculate the thermal conductivity. This method has the advantages of small sample size, fast and efficient measurement, and can significantly improve the experimental efficiency. In addition, by designing periodic heat pulses, multiple sets of data can be collected to further reduce measurement errors. However, the accuracy of this method depends on the sample density and the accuracy of specific heat measurement, which may affect the final measurement results. First, the geometry and boundary conditions of the sample have an important influence on the propagation of the heat pulse and the temperature distribution. Therefore, in practical applications, the sample preparation and test conditions must be very precise. Secondly, the dynamic method requires real-time acquisition of temperature data during the measurement process, which requires high-precision temperature sensors and fast data processing systems. If the sensor response time is slow or the data acquisition is not accurate enough, it will lead to deviations in the measurement results. Summary of the invention

[0008] The purpose of this section is to summarize some aspects of the embodiments of the present invention and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section and the specification abstract and invention title of this application to avoid blurring the purpose of this section, specification abstract and invention title, and such simplifications or omissions cannot be used to limit the scope of the present invention.

[0009] Therefore, an object of the present invention is to provide an anisotropic material thermal conductivity measuring device and an operating method thereof, which can more accurately measure and analyze the thermal conductivity of materials with anisotropic characteristics.

[0010] To solve the above technical problems, according to one aspect of the present invention, the present invention provides the following technical solutions:

[0011] A device for measuring thermal conductivity of anisotropic materials, comprising:

[0012] The box body comprises an outer shell and an outer shell cover detachably connected to the outer shell, wherein a cylindrical barrel with a top opening is arranged inside the outer shell, and the outer shell cover can close the top opening of the cylindrical barrel;

[0013] A weight sensor is disposed between the outer shell and the cylindrical barrel and is used to weigh the weight of the cylindrical barrel;

[0014] An electric heating wire surrounds the outer wall of the cylindrical tube;

[0015] The temperature sensors are arranged in the cylindrical tube at equal intervals and radiate from the center to the edge of the cylindrical tube;

[0016] The main control chip is electrically connected to the weight sensor, the heating wire and the temperature sensor, and is used to collect data signals from the weight sensor and the temperature sensor and input them into the built-in physical information neural network to calculate and measure the anisotropic thermal conductivity of the material to be tested;

[0017] The host computer is electrically connected to the main control chip, obtains data transmitted by the main control chip, and sends control instructions to the main control chip.

[0018] As a preferred solution of the device for measuring thermal conductivity of anisotropic materials described in the present invention, a first thermal insulation component is provided at the bottom of the inner wall of the cylindrical barrel, and a second thermal insulation component is provided at the bottom of the outer shell cover.

[0019] As a preferred solution of the device for measuring thermal conductivity of anisotropic materials described in the present invention, the first thermal insulation component and the second thermal insulation component are both made of silica nano aerogel felt material.

[0020] As a preferred solution of the anisotropic material thermal conductivity measuring device described in the present invention, the temperature sensors are arranged vertically one at a time in sequence along the radial direction from the center outward at θ=0°, 45°, ..., 270°, 315° with the axis point of the cylindrical tube as the center point.

[0021] As a preferred solution of the device for measuring thermal conductivity of anisotropic materials described in the present invention, the temperature sensor is a thermocouple.

[0022] As a preferred solution of the device for measuring thermal conductivity of anisotropic materials described in the present invention, a handle is provided on the outer shell cover, and lock buckles are provided on both sides that are buckled on the top of the side edges of the outer shell.

[0023] As a preferred solution of the device for measuring thermal conductivity of anisotropic materials described in the present invention, the physical information neural network includes the following structure:

[0024] Input layer: Set 4 input nodes, which are r , θ , z and time t ;

[0025] Output layer: There is only one output node, which means that r , θ , z and t Temperature T ;

[0026] Hidden layer: contains 3 hidden layers, each layer has 32 nodes;

[0027] Activation function: Hyperbolic tangent function is used as the activation function;

[0028] Loss function: It consists of two parts: data loss and physical loss, which guide the network to learn the thermal conductivity that conforms to the law of heat conduction. Its specific definition is as follows:

[0029] Data loss: Calculate the temperature T measured by the sensor during the experiment measured The PINN network predicts

[0030] The temperature T predicted The mean square error between:

[0031]

[0032] Where N represents the number of temperature sampling points;

[0033] Physical loss: Taking the heat conduction equation as a constraint, introducing the inversion parameter k rr , k θθ , k zz , calculate the loss of the model under the physical equation, based on the transient heat conduction equation in the cylindrical coordinate system, calculate the relationship between temperature change and thermal conductivity, and form a loss term:

[0034]

[0035] Total loss function: Combine data loss and physical loss to form a total loss function:

[0036] L total =L data +λL physics .

[0037] As a preferred solution of the anisotropic material thermal conductivity measuring device described in the present invention, the transient heat conduction equation in the cylindrical coordinate system is specifically constructed as follows:

[0038] The transient heat conduction equation without internal heat source is expressed as:

[0039] In the cylindrical coordinate system, the relationship between rectangular coordinates and cylindrical coordinates is:

[0040] x=rcosθ, y=rsinθ, z=z (2)

[0041] The gradient operator in the cylindrical coordinate system is obtained as:

[0042]

[0043] Ignoring the coupling terms in the thermal conductivity matrix, it is assumed that the thermal conductivity of wood is orthogonal, so the thermal conductivity matrix K is expressed as:

[0044]

[0045] Substituting equations (3) and (4) into equation (1) and expanding them, we obtain the transient heat conduction equation in the cylindrical coordinate system:

[0046]

[0047] As a preferred solution of the device for measuring thermal conductivity of anisotropic materials described in the present invention, the inner wall of the outer shell is provided with a threading groove for the weight sensor, the heating wire and the temperature sensor line to pass through.

[0048] An operation method of an anisotropic material thermal conductivity measuring device, the specific steps are as follows:

[0049] S1. According to the test sample preparation requirements, the wood to be tested is accurately processed into a cylinder with a diameter of 10 cm and a height of 5 cm, and a hole is vertically drilled every 1 cm in the radial direction from the center to the outside at θ=0°, 45°, ..., 270°, 315° with its axis point as the center;

[0050] S2. Start the device power supply, ensure that the host computer and the main control chip of the measuring device are stably connected through the wireless network, and the main control chip reads the initial weight and ambient temperature status information after adding the sample to be tested through the weight sensor and the temperature sensor, and uploads the data to the host computer system to automatically perform initial calibration to ensure that all sensors are in normal working condition;

[0051] S3. Set the target temperature parameters through the host computer, and the main control chip drives the heating wire to start heating. Wait for about 1-3 minutes for the temperature to reach a stable state;

[0052] S4. Quickly open the outer cover, put the sample to be tested into the device, ensure that the reserved temperature measurement hole of the sample to be tested is aligned with the temperature sensor, and gently press it into the cylindrical tube, close the outer cover, lock the lock, and seal the device;

[0053] S5. 10 seconds after the sample to be tested is completely placed, the measuring device starts to collect temperature data to reduce the interference of noise during the placement process on the results. The temperature sensor array collects temperature data every 1 second and transmits it back to the host computer through the main control chip. At the same time, the weight sensor continuously monitors the mass of the sample to be tested to ensure that the moisture content maintains the temperature;

[0054] S6. When the set time is reached or the moisture content changes by more than ±0.01%, the heating wire stops heating, the temperature sensor array stops collecting temperature, and the collected temperature data is preprocessed to filter out obvious erroneous data records. Then, the data is organized into a four-tuple (T, r, θ, z, t) so as to be input into the set physical information neural network.

[0055] S7, input the processed data into the physical information neural network for iterative training, and after each iteration, according to the total loss function L total =L data +λL physics Calculate the loss and adjust the weight parameters between each layer of the network and the thermal conductivity k to be inverted through error back propagation rr , k θθ , k zz , when the loss reaches convergence, the final result can be obtained;

[0056] S8 After the measuring device has cooled down, open the lock, take out the test sample, and clean the thermal grease on the inner wall of the cylindrical tube to prepare for the next measurement.

[0057] Compared with the prior art, the present invention has the following beneficial effects:

[0058] 1. Traditional thermal conductivity measurement methods usually require samples to be processed into thin plates and assume that heat propagates in a single direction. Therefore, multiple samples must be prepared to meet the needs of thermal conductivity measurement in different directions. This not only increases the cost of sample preparation, but may also affect the final measurement results due to slight differences in physical and chemical properties between samples, thereby having a negative impact on industrial applications. Anisotropy usually refers to the different physical properties of crystalline materials in different directions, such as expansion coefficient, optical properties, electrical properties, and mechanical properties. This property is caused by the directional arrangement of atoms in the crystal, which makes the crystal exhibit different properties in different directions. Therefore, based on anisotropy, by measuring the thermal conductivity in multiple directions at one time, the thermal conductivity of the material in different directions can be effectively reflected, significantly reducing the sample preparation requirements and related costs, and only one sample is required, eliminating the deviation caused by sample differences, ensuring the accuracy and reliability of the measurement results. This provides a more accurate evaluation of thermal conductivity for industrial applications.

[0059] 2. Traditional measurement methods process samples into thin plates or perform other forms of physical transformation, which may not only change the microstructure of the material, but also introduce mechanical stress, thereby affecting its physical and chemical properties. This destructive treatment often leads to inaccurate results in subsequent analysis or practical applications. In contrast, the present invention fully considers the phenotypic characteristics of the wood growth structure and processes the sample into a cylindrical shape to ensure its structural integrity as much as possible. This treatment method effectively maintains the original physical and chemical properties of the sample, which is particularly important when subsequent performance testing, material research and development, or quality control is required. Maintaining the integrity of the sample not only improves the reliability of the measurement results, but also provides higher credibility for material applications.

[0060] 3. Traditional heat conduction measurement methods mainly use differential method or finite element method. Δx There are limitations in the selection of : if it is too large, it cannot capture small temperature changes, and if it is too small, it may introduce noise, thus affecting the accuracy and robustness of the measurement. On the other hand, although the finite element method can handle complex geometry and boundary conditions, the complexity of its model establishment and meshing, as well as its sensitivity to input parameters, often make the stability of the results challenged. In contrast, the present invention adopts a physical information neural network (PINN) to guide learning by using the transient heat conduction equation as a constraint to ensure that the model satisfies the real physical laws. This enables it to more accurately describe the heat conduction phenomenon, effectively capture the complex heat flow and temperature distribution inside the material, and show good robustness to small noise in the measurement.

[0061] 4. Existing studies have shown that the moisture content of wood has a significant effect on its thermal conductivity. In order to reduce the interference of moisture content changes on the measurement results in principle, the present invention adopts a transient method of short-term heating. Compared with the long-term steady-state method, the transient method can significantly reduce water loss, thereby better maintaining the moisture content of the sample. To this end, the present invention integrates a high-precision weight sensor that can monitor the mass changes of the sample during the heating process in real time. This monitoring method ensures that the moisture content of the sample remains stable during the measurement process, and effectively avoids measurement errors caused by moisture fluctuations. Through this comprehensive method, we have enhanced the accuracy and reliability of thermal conductivity measurements and provided a more reliable basis for the evaluation of wood material properties.

[0062] 5. In order to accurately describe the thermal conductivity of cylindrical samples, we use the heat conduction equation based on the cylindrical coordinate system. This equation can effectively reflect the thermal conductivity of materials in different directions, and is compatible with the cylindrical structure of the measuring instrument, ensuring that the thermal conduction behavior of the sample can be accurately captured during the actual measurement process. By combining this mathematical model with the measuring instrument, we provide a solid theoretical foundation for subsequent data processing and thermal conductivity calculation, thereby improving the accuracy and reliability of the measurement. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the present invention will be described in detail below in combination with the accompanying drawings and detailed embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative labor. Among them:

[0064] Figure 1 It is a schematic diagram of the overall structure of a device for measuring thermal conductivity of anisotropic materials according to the present invention;

[0065] Figure 2 A partial structural top view of a device for measuring thermal conductivity of anisotropic materials according to the present invention;

[0066] Figure 3 This is a control principle block diagram of a device for measuring thermal conductivity of anisotropic materials according to the present invention;

[0067] Figure 4 A schematic diagram of the distribution of the drilling holes of the samples to be tested provided by the present invention;

[0068] Figure 5 This is a schematic diagram of the fitting results of the measured temperature and PINN at different distances from the sample to be measured in the embodiment provided by the present invention. DETAILED DESCRIPTION

[0069] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0070] The present invention provides a device for measuring thermal conductivity of anisotropic materials and an operating method thereof, which can more accurately measure and analyze the thermal conductivity of materials with anisotropic characteristics.

[0071] Figure 1-Figure 4 The figure shows a schematic diagram of an embodiment of the present invention for measuring the thermal conductivity of anisotropic materials. Figure 1-Figure 4 , the thermal conductivity of the anisotropic material of this embodiment is measured as follows:

[0072] 1. Sample preparation:

[0073] In order to maintain the structural integrity of the material, the sample needs to be precisely processed into a cylindrical sample with a diameter of 10 cm and a height of 5 cm, with the center of the annual ring as the reference center. This processing method fully considers the growth characteristics of the wood to ensure that the sample maintains its original physical and chemical properties during the measurement process. Through the design of the cylindrical sample, we effectively protect the microstructure of the material and avoid the mechanical stress introduced by the change of sample shape in the traditional method.

[0074] In order to measure the temperature distribution inside the sample, the sample needs to be drilled to insert the temperature sensor 400. The diameter of the drill hole is 3 mm, and the direction is vertically drilled from the upper surface of the sample to the lower surface. Take θ = 0°, 45°, ..., 270°, 315°, and drill a hole every 1 cm in the radial direction from the center to the outside. The distribution diagram is shown in the figure. Figure 4 This design ensures that the temperature sensor can accurately capture the temperature changes inside the sample, thereby providing more comprehensive data information for the measurement of thermal conductivity.

[0075] 2. Heat conduction model design and boundary control:

[0076] (1) Heat conduction equation in cylindrical coordinate system

[0077] In order to accurately describe the thermal conductivity of cylindrical samples, we innovatively proposed to transform the heat conduction equation in the rectangular coordinate system to the cylindrical coordinate system to better adapt to the geometric structure of the sample. This transformation not only improves the adaptability of the model, but also ensures that the thermal conduction behavior of the sample can be accurately captured during the actual measurement process, thereby providing more reliable theoretical support for the calculation of thermal conductivity. Through this method, we have significantly improved our understanding of the thermal conductivity of cylindrical materials and the accuracy of our measurements.

[0078] The transient heat conduction equation without internal heat source is expressed as:

[0079]

[0080] In the cylindrical coordinate system, the relationship between rectangular coordinates and cylindrical coordinates is:

[0081] x=r cosθ, y=r sinθ, z=z (2)

[0082] The gradient operator in the cylindrical coordinate system is obtained as:

[0083]

[0084] In view of the fact that the thermal conductivity of wood in radial, chordal and longitudinal directions is of great concern in practical production applications, we assume that the coupling terms in the thermal conductivity matrix can be ignored and that the thermal conductivity of wood is orthogonal. Therefore, the thermal conductivity matrix K is expressed as:

[0085]

[0086] Substituting equations (3) and (4) into equation (1) and expanding them, we obtain the heat conduction equation in the cylindrical coordinate system:

[0087]

[0088] (2) Boundary control design

[0089] We treat the upper and lower surfaces as adiabatic boundaries to effectively prevent heat loss through these surfaces. The setting of adiabatic boundaries means that there is no heat flow on these surfaces, ensuring that the heat of the sample depends entirely on its internal thermal conductivity. This design can avoid the influence of the external thermal environment on the sample, making the heat conduction behavior inside the sample more pure and controllable.

[0090] At the same time, the side surface is designed as an isothermal boundary to keep its temperature stable at a preset value. This treatment ensures that the side surface always maintains a constant temperature during the measurement process, preventing the measurement results of thermal conductivity from being disturbed by fluctuations in the external environment temperature.

[0091] 3. Structural design of measuring device:

[0092] like Figure 1-Figure 3 As shown, the measuring device includes a box 100, a weight sensor 200, a heating wire 300, a temperature sensor 400, a main control chip and a host computer.

[0093] The box 100 includes an outer shell 110 and an outer shell cover 120 detachably connected to the outer shell 110. The inner part of the outer shell 110 is provided with a cylindrical tube 110a with an opening at the top. The outer shell cover 120 can close the top opening of the cylindrical tube 110a. The inner wall bottom of the cylindrical tube 110a is provided with a first insulation component 110b. The bottom of the outer shell cover 120 is provided with a second insulation component 120a. When the sample to be tested is placed, the sealing effect of the device is achieved by locking the lock buckle 2. At the same time, the first insulation component 110b and the second insulation component 120a form an insulation boundary to prevent heat from being lost through these surfaces. Preferably, in this embodiment, the first insulation component 110b and the second insulation component 120a are made of silica nano aerogel felt, which has extremely low thermal conductivity and can significantly reduce the loss of heat from the upper and lower surfaces. The silica nano aerogel felt can effectively inhibit heat conduction and heat convection with its nanoscale pore structure. In addition, the material is lightweight and flexible, making it easy to cut and install according to specific shapes, ensuring a close fit to the sample surface and further improving thermal insulation performance. By using this material, we can better control heat flow in experiments and ensure that the temperature change inside the sample is only affected by lateral heat conduction, thereby improving the accuracy and repeatability of the measurement results and meeting the experimental needs for high-precision thermal environment control.

[0094] A weight sensor 200 is disposed between the outer shell 110 and the cylindrical barrel 110a and is used to weigh the weight of the cylindrical barrel 110a;

[0095] The heating wire 300 surrounds the outer wall of the cylindrical tube 110a. The cylindrical tube 110a is designed as a copper cylindrical structure. The inner wall of the shell is in thermal contact with the sample to be tested, and the outer wall wraps the heating wire 300. The high thermal conductivity of copper ensures that the heat can be quickly and evenly conducted to maintain a consistent temperature distribution around the sample. When the heating wire 300 is heated, the heat is first transferred to the copper outer wall, and then evenly conducted to the sample to be tested through the copper shell.

[0096] The temperature sensors 400 are arranged equidistantly in the cylindrical tube 110a, and diverge from the center of the cylindrical tube 110a to the edge. Specifically, the temperature sensors 400 are arranged vertically every 1 cm in the radial direction from the center to the outside with θ=0°, 45°, ..., 270°, 315°, and are arranged in sequence. The temperature sensors 400 are used to measure the temperature distribution inside the sample to be tested during the experiment. T-type thermocouples are selected as temperature sensors, which are known for their excellent linearity, large thermoelectric potential and high sensitivity, and can accurately capture small temperature changes inside the sample. In order to meet the needs of multi-point measurement, we customized the thermocouple. This customized design enables the thermocouple to simultaneously measure the temperature values ​​inside the sample at 10, 20, 30 and 40 mm from the surface, thereby providing a comprehensive temperature distribution profile.

[0097] The main control chip is electrically connected to the weight sensor 200, the heating wire 300 and the temperature sensor 400, and is used to collect data signals from the weight sensor 200 and the temperature sensor 400 and input them into the built-in physical information neural network to calculate and measure the anisotropic thermal conductivity of the material to be tested. The host computer is electrically connected to the main control chip, obtains the data transmitted by the main control chip, and sends control instructions to the main control chip, such as Figure 3 As shown in the figure, the main control chip is stably connected to the host computer through the WIFI module to ensure that the experimental data is transmitted to the host computer in time for monitoring. The accurate operation of the system depends on three core modules: monitoring status module, temperature monitoring module and heating control module. The monitoring status module detects the mass change of the sample during the experiment through a high-precision weight sensor 200 to ensure that the water content of the sample remains stable and avoids significant impact on the experimental results. The temperature monitoring module uses temperature sensors 400 evenly distributed in the sample to monitor temperature changes in real time; the heating control module adjusts the working state of the heating wire 300 through the heating control chip, and the PID controller continuously performs feedback adjustment to achieve precise control of the sample heating process.

[0098] 4. Design of physical information neural network:

[0099] The basic idea of ​​​​physical information neural network (PINN) is to combine physical knowledge with deep learning, especially in solving partial differential equations (PDEs) and inversion problems. PINN embeds physical constraints directly into the loss function of the neural network, so that the network can not only fit the data but also satisfy the required physical laws. To this end, we use physical information neural network (PINN) to invert the three thermal conductivities. The specific PINN network structure is designed as follows: Input layer: Set 4 input nodes, namely r, θ, z and time t in the cylindrical coordinate system;

[0100] Output layer: There is only one output node, which represents the temperature T under given r, θ, z and t;

[0101] Hidden layer: contains 3 hidden layers, each layer has 32 nodes;

[0102] Activation function: Hyperbolic tangent function is used as the activation function;

[0103] Loss function: It consists of two parts: data loss and physical loss, which guide the network to learn the thermal conductivity that conforms to the law of heat conduction. Its specific definition is as follows:

[0104] Data loss: Calculate the temperature T measured by the sensor during the experiment measuted The temperature T predicted by the PINN network predicted The mean square error between:

[0105]

[0106] Where N represents the number of temperature sampling points;

[0107] Physical loss: Taking the heat conduction equation as a constraint, introducing the inversion parameter k rr , k θθ , k zz , calculate the loss of the model under the physical equation, based on the transient heat conduction equation in the cylindrical coordinate system, calculate the relationship between temperature change and thermal conductivity, and form a loss term:

[0108]

[0109] Total loss function: Combine data loss and physical loss to form a total loss function:

[0110] L total =L data +λL physics (8)

[0111] Among them, λ is a hyperparameter used to balance the importance of the two parts of the loss. Through the above design, the physical information neural network (PINN) can efficiently learn the anisotropic thermal conductivity k based on limited discrete data. rr , k θθ , k zz , while ensuring that the solution obtained conforms to the physical laws of the heat conduction equation. This method fully combines the advantages of data-driven and physical models to improve the accuracy and robustness of the inversion.

[0112] The specific measurement steps of the above device are as follows:

[0113] S1. According to the preparation requirements of the sample H to be tested, the wood to be tested is accurately processed into a cylinder with a diameter of 10 cm and a height of 5 cm, and a hole is vertically drilled every 1 cm in the radial direction from the center to the outside at θ=0°, 45°, ..., 270°, 315° with its axis point as the center;

[0114] S2, start the device power supply, ensure that the host computer and the main control chip of the measuring device are stably connected through the wireless network, the main control chip reads the initial weight and ambient temperature status information after adding the sample to be tested through the weight sensor 200 and the temperature sensor 400, and uploads the data to the host computer system to automatically perform initial calibration to ensure that all sensors are in normal working condition;

[0115] S3. Set the target temperature parameter (such as 90°C) through the host computer, and the main control chip drives the heating wire 300 to start heating, and wait for about 1-3 minutes for the temperature to reach a stable state;

[0116] S4, quickly open the outer shell cover 120, put the sample to be tested into the device, ensure that the reserved temperature measurement hole of the sample to be tested is aligned with the temperature sensor 400, and gently press it into the cylindrical tube 110a, close the outer shell cover 120, lock the lock 120c, and achieve the sealing of the device;

[0117] S5. 10 seconds after the sample H is completely placed, the measuring device starts to collect temperature data to reduce the interference of noise during the placement process on the results. The temperature sensor 400 array collects temperature data every 1 second and transmits it back to the host computer through the main control chip. At the same time, the weight sensor 200 continuously monitors the mass of the sample to be tested to ensure that the moisture content maintains the temperature;

[0118] S6. When the set time (such as 30 minutes) is reached or the moisture content changes by more than ±0.01%, the heating wire 300 stops heating, the temperature sensor array 400 stops temperature collection, and the collected temperature data is preprocessed to filter out obvious erroneous data records. Then, the data is organized into a four-tuple (T, r, θ, z, t) so as to be input into the set physical information neural network.

[0119] S7. Input the processed data into the physical information neural network for iterative training. After each iteration, calculate the loss according to formula (6), (7), (8), and adjust the weight parameters between each layer of the network and the thermal conductivity k to be inverted through error back propagation. rr , k θθ 、j zz , when the loss reaches convergence, the final result can be obtained;

[0120] S8 After the measuring device cools down, the lock 120c is opened, the test sample is taken out, and the thermal grease on the inner wall of the cylindrical tube 110a is cleaned to prepare for the next measurement.

[0121] In order to verify the effectiveness of the present invention, the following examples are provided:

[0122] Taking the determination of the anisotropic thermal conductivity of pine wood as an example, the isothermal boundary was set at 90°C and the heating time was 20 minutes. The data obtained from the experimental measurement was input into the physical information neural network (PINN) for training. After 10,000 iterations, the model converged and the final L total is 1.225471, where L data is 0.592638, L physics The heat diffuses from the outer layer to the center. The measured temperature at different distances and the fitting results of PINN are as follows: Figure 5 The measured temperature data are averaged and displayed, and the result shows how well the temperature change at each location fits the fitting curve.

[0123] In the figure, the blue "×" represents the average value of the measured temperature. The sampling frequency of the original data is once per second, but due to the long sampling time and large amount of data, only one point is shown here every 10 seconds. The red curve is the fitting result obtained through PINN training. The corresponding inversion results are compared with the conclusions of existing research as shown in Table 1:

[0124] Table 1 Thermal conductivity results

[0125]

[0126] It can be seen that the present invention can efficiently measure the thermal conductivity in multiple directions at one time through the physical information neural network (PINN). Compared with the traditional method, it not only speeds up the measurement speed, but also makes the result more accurate, demonstrating its effectiveness in the determination of anisotropic thermal conductivity.

[0127] Although the present invention has been described above with reference to the embodiments, various modifications may be made thereto and parts thereof may be replaced by equivalents without departing from the scope of the present invention. In particular, as long as there is no structural conflict, the various features in the embodiments disclosed in the present invention may be used in combination with each other in any manner, and the fact that these combinations are not exhaustively described in this specification is only for the sake of omitting space and saving resources. Therefore, the present invention is not limited to the specific embodiments disclosed herein, but includes all technical solutions falling within the scope of the claims.

Claims

1. A device for measuring thermal conductivity of anisotropic materials, characterized in that: include: The box (100) comprises an outer shell (110) and an outer shell cover (120) detachably connected to the outer shell (110), wherein a cylindrical barrel (110a) with a top opening is arranged inside the outer shell (110), and the outer shell cover (120) can close the top opening of the cylindrical barrel (110a); A weight sensor (200) is arranged between the outer shell (110) and the cylindrical barrel (110a) and is used to weigh the weight of the cylindrical barrel (110a); An electric heating wire (300) surrounds the outer wall of the cylindrical tube (110a); The temperature sensors (400) are arranged in the cylindrical tube (110a) at equal intervals around the circumference and radiate from the center to the edge of the cylindrical tube (110a); A main control chip is electrically connected to the weight sensor (200), the heating wire (300) and the temperature sensor (400), and is used to collect data signals from the weight sensor (200) and the temperature sensor (400) and input the data signals into a built-in physical information neural network to calculate and measure the anisotropic thermal conductivity of the material to be tested; The host computer is electrically connected to the main control chip, obtains data transmitted by the main control chip, and sends control instructions to the main control chip.

2. The device for measuring thermal conductivity of anisotropic materials according to claim 1, characterized in that: A first heat insulating component (110b) is disposed at the bottom of the inner wall of the cylindrical barrel (110a), and a second heat insulating component (120a) is disposed at the bottom of the outer shell cover (120).

3. The device for measuring thermal conductivity of anisotropic materials according to claim 2, characterized in that: The first heat insulating component (110b) and the second heat insulating component (120a) are both made of silicon dioxide nano aerogel felt material.

4. The device for measuring thermal conductivity of anisotropic materials according to claim 1, characterized in that: The temperature sensors (400) are arranged vertically at intervals of 1 cm along the radial direction from the center outward at θ=0°, 45°, ..., 270°, 315°, with the axis point of the cylindrical tube (110a) being arranged in sequence.

5. The device for measuring thermal conductivity of anisotropic materials according to claim 1, characterized in that: The temperature sensor (400) is a T-type thermocouple.

6. The device for measuring thermal conductivity of anisotropic materials according to claim 1, characterized in that: The outer shell cover (120) is provided with a handle (120b), and two sides thereof are provided with lock buckles (120c) which are buckled on the top of the side edge of the outer shell (110).

7. The device for measuring thermal conductivity of anisotropic materials according to claim 1, characterized in that: The physical information neural network includes the following structure: Input layer: Set 4 input nodes, namely r, θ, z and time t in the cylindrical coordinate system; Output layer: There is only one output node, which represents the temperature T under given r, θ, z and t; Hidden layer: contains 3 hidden layers, each layer has 32 nodes; Activation function: Hyperbolic tangent function is used as the activation function; Loss function: It consists of two parts: data loss and physical loss, which guide the network to learn the thermal conductivity that conforms to the law of heat conduction. Its specific definition is as follows: Data loss: Calculate the temperature T measured by the sensor during the experiment measured The temperature T predicted by the PINN network predicted The mean square error between: Where N represents the number of temperature sampling points; Physical loss: Taking the heat conduction equation as a constraint, introducing the inversion parameter k rr , k θθ , k zz , calculate the loss of the model under the physical equation, based on the transient heat conduction equation in the cylindrical coordinate system, calculate the relationship between temperature change and thermal conductivity, and form a loss term: Total loss function: Combine data loss and physical loss to form a total loss function: THE total =L data +λL physics 。 8. The device for measuring thermal conductivity of anisotropic materials according to claim 7, characterized in that: The specific construction of the transient heat conduction equation in the cylindrical coordinate system is as follows: The transient heat conduction equation without internal heat source is expressed as: In the cylindrical coordinate system, the relationship between rectangular coordinates and cylindrical coordinates is: x=rcosθ, y=rsinθ, z=z (2) The gradient operator in the cylindrical coordinate system is obtained as: Ignoring the coupling terms in the thermal conductivity matrix, it is assumed that the thermal conductivity of wood is orthogonal, so the thermal conductivity matrix K is expressed as: Substituting equations (3) and (4) into equation (1) and expanding them, we obtain the transient heat conduction equation in the cylindrical coordinate system:

9. The device for measuring thermal conductivity of anisotropic materials according to claim 1, characterized in that: The inner wall of the outer shell (110) is provided with a threading groove (110c) for the weight sensor (200), the heating wire (300) and the temperature sensor (400) to pass through.

10. An operating method of the anisotropic material thermal conductivity measuring device according to any one of claims 1 to 9, characterized in that: The specific steps are as follows: S1. According to the preparation requirements of the sample (H), the wood to be tested is accurately processed into a cylinder with a diameter of 10 cm and a height of 5 cm. With its axis point as the center, a hole is vertically drilled every 1 cm in the radial direction from the center to the outside at θ=0°, 45°, ..., 270°, 315′; S2, start the device power supply, ensure that the host computer and the main control chip of the measuring device are stably connected via the wireless network, the main control chip reads the initial weight and ambient temperature status information after adding the sample to be tested via the weight sensor (200) and the temperature sensor (400), and uploads the data to the host computer system to automatically perform initial calibration, ensuring that all sensors are in normal working condition; S3. The target temperature parameters are set by the host computer, and the main control chip drives the heating wire (300) to start heating, and waits for about 1-3 minutes for the temperature to reach a stable state; S4. Quickly open the outer shell cover (120), place the sample to be tested into the device, ensure that the reserved temperature measurement hole of the sample to be tested is aligned with the temperature sensor (400), and gently press it into the cylindrical tube (110a), close the outer shell cover (120), and lock the lock (120c) to achieve the sealing of the device; S5. 10 seconds after the sample (H) to be tested is completely placed, the measuring device starts to collect temperature data to reduce the interference of noise during the placement process on the results. The temperature sensor (400) array collects temperature data every 1 second and transmits it back to the host computer through the main control chip. At the same time, the weight sensor (200) continuously monitors the mass of the sample to be tested to ensure that the moisture content maintains the temperature; S6. When the set time is reached or the moisture content changes by more than ±0.01%, the heating wire (300) stops heating, the temperature sensor (400) array stops collecting temperature, and the collected temperature data is preprocessed to filter out obvious erroneous data records. Then, the data is organized into a four-tuple (T, r, θ, z, t) form so as to be input into the set physical information neural network; S7, input the processed data into the physical information neural network for iterative training, and after each iteration, according to the total loss function L total =L data +λL physics Calculate the loss and adjust the weight parameters between each layer of the network and the thermal conductivity k to be inverted through error back propagation rr , k θθ , k zz , when the loss reaches convergence, the final result can be obtained; S8 After the measuring device cools down, the lock (120c) is opened, the test sample is taken out, and the thermal conductive silicone grease on the inner wall of the cylindrical tube (110a) is cleaned to prepare for the next measurement.

Images