Computational models based on machine learning and validation for predicting ionic liquids viscosity in mixtures

Area of Science:

• Physical Chemistry
• Computational Chemistry
• Materials Science

Background:

• Ionic liquids are versatile solvents with tunable properties.
• Accurate viscosity prediction is crucial for their application in various chemical processes.
• Existing models may lack the precision required for complex ionic liquid solutions.

Purpose of the Study:

• To develop and evaluate machine learning models for predicting ionic liquid solution viscosity.
• To identify the most effective algorithms for viscosity estimation based on key parameters.
• To optimize model performance using hyper-parameter tuning.

Main Methods:

• Utilized Random Forest (RF), Gradient Boosting (GB), and XGBoost (XGB) machine learning algorithms.
• Input parameters included cation type, anion type, temperature (K), and ionic liquid concentration (mol%).
• Employed Glowworm Swarm Optimization (GSO) for hyper-parameter optimization.

Main Results:

• Random Forest (RF) achieved the highest predictive accuracy with an R² of 0.9971.
• Gradient Boosting (GB) and XGBoost (XGB) also demonstrated high performance with R² values of 0.9916 and 0.9911, respectively.
• Models showed excellent agreement with experimental data, further validated by RMSE and MAPE metrics.

Conclusions:

• Machine learning models, particularly Random Forest, are highly effective for predicting ionic liquid solution viscosity.
• The chosen input parameters and optimization techniques significantly enhance predictive capabilities.
• These models offer a reliable and efficient tool for the design and application of ionic liquids.