Artificial Intelligence Assisted Noncontact Microwave Sensor for Multivariable Biofuel Analysis
Masoud Baghelani, Navid Hosseini, Mojgan Daneshmand
Abstract
Multivariable component analysis is one of the most challenging topics in the area of microwave resonator based sensors. In this article, a new approach is developed for introducing new independent features for analyzing the volumetric fraction of water, ethanol, and gasoline in E85 biofuel samples. The novel features are extracted based on a multiharmonics measurement of frequency and amplitude variations of the transmission response of the resonator over multiple harmonics due to nonlinearity and uniqueness of the permittivity spectrum of different materials. For the experiments, 60 samples of biofuel mixtures are prepared with randomly chosen percentages of each of the components. An artificial neural network is trained with the extracted features from 40 of the samples and tested over the remaining 20 samples. The average relative error in determining the water concentration in the biofuel samples of as low as 0.09% is achieved. The experimental results verify the capability of the sensor for selective analysis of all the components of a multivariable mixture simultaneously.