Identification of polar liquids using support vector machine based classification model

Abstract

The dispersive nature of polar liquids creates ambiguity in their identification process. It requires a long time and effort to compare the measured values with the available standard values to identify the unknown liquid. Nowadays machine learning techniques are being used widely to assist the measurement techniques and make predictions with great accuracy and less human effort. This paper proposes a support vector machine (SVM) based classification model for the identification of six polar liquids- butan-1-ol, dimethyl sulphoxide, ethanediol, ethanol, methanol and propan-1-ol for a temperature range of 10 °C–50 °C and frequency range of 0.1 GHz – 5 GHz. The model is constructed using the data from the National Physical Laboratory (NPL) report MAT 23. The identification of unknown liquid is based on complex permittivity measurement. If the measurement error in complex permittivity is less than ±6% of the standard value in NPL report, the proposed model identifies the liquids with 100% accuracy in the entire temperature and frequency range. The performance of the model is validated by testing the model with data external to the dataset used.  The findings show that the proposed model is a useful and efficient tool for identifying unknown polar liquids.