[Discrimination of Crude Oil Samples Using Laser-Induced Time-Resolved Fluorescence Spectroscopy].

Area of Science:

• Spectroscopy
• Pattern Recognition
• Machine Learning

Context:

• Traditional 3D fluorescence analysis is limited to laboratory settings, hindering rapid field detection of oil spills.
• Effective field detection methods are crucial for timely response and mitigation of oil spill incidents.
• Laser-induced fluorescence (LIF) combined with pattern recognition is established for oil discrimination.

Purpose:

• To develop and validate a novel field-deployable method for crude oil identification using laser-induced time-resolved fluorescence.
• To enhance the accuracy and flexibility of oil spill detection by optimizing data analysis in time and wavelength domains.
• To apply a support vector machine (SVM) model for classifying crude oil samples based on spectral data.

Summary:

• A new method utilizes laser-induced time-resolved fluorescence data, reduced to two dimensions by selecting optimal time (54–74 ns) and wavelength (387.00–608.87 nm) ranges.
• This data reduction and selection process, forming an SVM database, significantly improved crude oil classification accuracy from 83.3% to 88.1% (time domain) and 84% to 100% (wavelength domain).
• The optimized method demonstrates increased accuracy and flexibility, addressing fluctuations common in field fluorescence lidar detection.

Impact:

• Provides a more adaptable and accurate method for on-site crude oil identification, crucial for emergency response.
• The data reduction technique offers a valuable reference for future development of oil classification systems.
• Enhances the potential for rapid and reliable oil spill assessment in diverse environmental conditions.