High temperature dielectric properties measurement system at 915 MHz based on deep learning

摘要

Abstract A ridged waveguide and heating system based on deep learning for dynamic measurement of dielectric properties at high temperatures has been developed. Firstly, specially designed ridged waveguide at the frequency of 915 MHz was utilized. The finite difference time domain simulation was then combined with the deep learning algorithm to construct the dielectric properties. In addition, the calibration method based on a two‐point algorithm was involved in the network. The precision of the measured dielectric constant and loss tangent was improved by 4.4 and 1.9, respectively. Simulated results indicated that the deep learning algorithm had higher accuracy than back propagation neural networks and support vector machine algorithms. The reliability of the system was verified by Si 3 N 4 and alcohols at room temperature. The dielectric properties of quartz particles were also tested over a temperature range up to 900°C. The experiment results agreed well with the reference values. This system can measure a wider range of dielectric properties and achieve higher accuracy using the advanced deep learning algorithm. It is more convenient to measure large size materials in industrial applications at high temperatures.