Quantitative Elemental Oxides Analysis of Rock Cuttings Using Laser-Induced Breakdown Spectroscopy Coupled with...

Area of Science:

• Geochemistry
• Machine Learning
• Spectroscopy

Background:

• On-site elemental analysis is vital for operational efficiency and safety in the energy sector.
• Rapid geochemical data aids geothermal exploration, carbon storage integrity, and hydrocarbon exploration.

Purpose of the Study:

• To evaluate Bayesian optimization coupled with support vector machine for quantifying major oxide elements in rock cuttings using Laser-Induced Breakdown Spectroscopy (LIBS).
• To automatically optimize model performance and minimize manual hyperparameter tuning.

Main Methods:

• Utilized over 1000 rock cutting samples for model development.
• Employed X-ray fluorescence as the reference method for elemental concentration.
• Integrated Bayesian optimization with support vector machine for hyperparameter tuning.

Main Results:

• Achieved high model reliability and accuracy with R-squared values between 0.93 and 0.97.
• Demonstrated the model's effectiveness in handling high-dimensionality and nonlinearity in LIBS data.

Conclusions:

• The combined approach of LIBS, machine learning, and Bayesian optimization provides fast, reliable elemental analysis for energy sector applications.
• This method supports adaptive and cost-effective resource development through accurate geochemical characterization.