Machine Learning Analysis of Low-Frequency Impedance Spectra of Binary Mixtures of Polar and Non Polar Liquids

Abstract

The complex dielectric permittivities were determined using a precision LCR meter of the binary mixtures of a polar liquid (methyl ethyl ketone, or MEK) and a non-polar liquid (dimethyl silicone fluid, or DMSF) at 303.15 K temperature. After determining the complex impedance (Z* (ω)) using the complex permittivity portion (ɛ* (ω)) of polar and nonpolar liquids, the complex impedance data (Z* (ω)) was fitted to the Nyquist Plot. Our research described here investigated the impact of ionic impurities in the pure MEK, DMSF and in their mixed state. The established parameters provided information on the impact of concentration changes on the electrical characteristics of the binary mixtures. Our research delved into utilizing machine learning (ML) techniques, such as random forest (RF), LightGBM, and XGBoost regressors, to improve the material design and prediction modeling. The main objective was to assess the efficacy of various regression techniques in evaluating the material characteristics performance. Experiments spanning from 30% to 50% of the data set were carried out, with performance metrics, like, R² score and MAE, being utilized. Notably, the RF regressor demonstrated outstanding performance in these evaluations with an R² score of 0.9999. The simulation results indicated that the ML-based techniques offer resource and time savings, serving as effective tools for predicting material performance at intermediate frequencies.

Keywords:

• Complex impedance spectra
• bode plot analysis
• EC lab software
• machine learning
• methyl ethyl ketone
• dimethyl silicone fluid

Disclosure statement

The authors have no conflicts to disclose.

Data availability statement

The data supporting the outcomes of this study can be obtained directly from the corresponding author upon making a reasonable request.

Additional information

Funding

For their kind financial support, the authors would like to thank the Department of Science and Technology (DST), New Delhi, for the DST-FIST (Level-I) project (SR/FST/PSI-001/2006) and DST-FIST (Level-II) project (SR/FST/PSI-198/2014). Furthermore, deep gratitude is extended to the University Grants Commission (UGC), New Delhi, for its financial support of the DRS-SAP program grants [No.F.530/10/DRS/2010 (SAP-I)] and [No.F.530/17/DRS-II/2018 (SAP-I)].