Machine learning-assisted rapid screening of oil types on microfluidic thread-based analytical devices (μTADs).

Area of Science:

• Analytical Chemistry
• Materials Science
• Machine Learning

Background:

• Conventional oil classification methods are expensive, time-consuming, and unsuitable for field analysis.
• There is a need for portable, cost-effective technologies for rapid, on-site oil classification.
• Current limitations hinder real-time decision-making in the petroleum industry.

Purpose of the Study:

• To develop a novel, low-cost, and portable oil classification system.
• To integrate microfluidic thread-based analytical devices (μTADs) with machine learning for oil analysis.
• To enable rapid, on-site classification of petroleum and non-petroleum oil samples.

Main Methods:

• Utilized capacitance data from capillary flow on microfluidic thread-based analytical devices (μTADs).
• Employed linear discriminant analysis (LDA) for dimensionality reduction and feature modeling.
• Implemented a machine learning approach for oil sample classification.

Main Results:

• Achieved 82.00% average accuracy in classifying eight petroleum oil samples.
• Successfully classified fifteen oil samples (eight petroleum, seven non-petroleum) with 72.22% average accuracy.
• Demonstrated robust performance in distinguishing oils with similar properties.

Conclusions:

• The developed system is feasible for practical field applications, enabling rapid on-site oil analysis.
• This advancement supports real-time quality monitoring and inspection in the petroleum industry.
• The technology facilitates immediate decision-making in regulatory processes and field operations.