Chemometric approaches to resolving base oil mixtures.

Area of Science:

• Lubrication science
• Analytical chemistry
• Data science

Background:

• Commercial base oils are often blends from various sources to optimize cost and properties.
• Complex base oil mixtures pose formulation challenges, especially when component ratios are unknown.

Purpose of the Study:

• To apply modern data science and chemometrics to resolve component ratios in complex base oil mixtures.
• To develop an accurate method for determining the proportions of different base oil groups in blended lubricants.

Main Methods:

• Field ionization mass spectrometry (FIMS) was employed to gather chemical data from base oil mixtures.
• Data processing involved molecular formula assignment and statistical analysis within a Python environment.
• Regression techniques, including regularized linear models and automated machine learning (AutoML), were evaluated.

Main Results:

• An automated machine learning pipeline identified effective modeling strategies for the FIMS data.
• Polynomial feature generation combined with ridge cross-validation regression provided the best results.
• The methodology achieved an accuracy of ±5% in resolving the ratio of Group 2 and Group 3 base oils in blended mixtures.

Conclusions:

• Modern data science and chemometrics offer successful strategies for analyzing complex mixtures.
• This approach enhances lubricant formulation by accurately determining base oil blend ratios.