Development Of The Instrument For Measuring Thermal Conduction Properties Of Materials Based On Full Three-Dimensional Numerical Simulation

Accurate measurement of thermal properties of materials plays a vital role in the fields of electronics,construction,aerospace and other fields based on energy and power.The common methods for measuring the thermal conductivity are the hot line method and the planar heat source method(TPS).However,because the calculation of the thermal differential equation based on the three-dimensional model of the probe is computationally intensive and it is too hard to obtain the analytical solution,neither of them consider the three-dimensional geometric effect of the probe.The hot wire method simplifies the probe to infinitely thin and the TPS method simplifies the probe to infinitely thin.Moreover,the two methods can only obtain the thermal conductivity accurately,while the measurement error of the thermal diffusivity is larger than an order of magnitude.Aiming at the above problems,this paper proposes a new method for measuring thermal properties in combination with three-dimensional numerical simulation and machine learning.The method fully considers the three-dimensional geometric effect of the probe,and the the probe can be any shape.In addition,it can effectively solve the problem of large calculations to solve the three-dimensional heat conduction differential equation without losing the calculation accuracy.Moreover,it and obtain the K and ? relatively accurately at the same time,greatly simplifying the measurement process.In order to provide training data for the machine learning model,the transient increased temperature-time database of the heating probe under different K and ? is established.And in order to provide experimental data for the machine learning model to predict the thermal properties of materials,the thermal conductivity measuring instrument is built and three standard materials,including quartz glass,SUS304 stainless steel and 45 steel,are measured to obtain the experimental measuring transient increased temperature.In this paper,the three-dimensional model of the thermal property parameter measuring instrument,including the base,the heating probe and the material to be tested,is established based on the self-designed heating probe,and its temperature field is solved based on the finite volume method,so that the database is established.The temperature signal of the metal is converted into an electrical signal based on the characteristic that the resistance of the metal changes linearly with temperature.And a circuit is designed to make the power of the heating probe constant.Based on Ohm's law and the probe resistance temperature relationship,the experimental measuring transient increased temperature of the probe can be reversed.Based on the above work,two machine learning models,neural network 1 and neural network 2,are built.The neural network 1 is trained using the average temperature rise of the probe under each K in the transient increased temperature-time database.At each K,the vector X consisting of the transient temperature rise in the database is taken as input,and ? is taken as output,then the neural network 2 can be trained.At last,taking the experimental measuring transient increased temperature as input,the K and ? of the three standard materials can be obtained.The results show that the thermal conductivity error is less than 10% and the thermal diffusion coefficient is less than 30%.