Bayesian optimization for the direct optimization of temperature-programmed separations in ultra-fast gas...

Area of Science:

• Analytical Chemistry
• Chemometrics
• Instrumentation

Background:

• Ultra-fast gas chromatography electronic nose (UFGC-E-nose) systems enable rapid volatile fingerprinting of complex food and beverage matrices.
• Optimizing separation in UFGC-E-nose systems is challenging due to thermal hysteresis caused by high ramp rates, invalidating traditional models.

Purpose of the Study:

• To introduce an automated optimization framework using Bayesian optimization (BO) for UFGC-E-nose systems.
• To systematically improve separation and discrimination efficiency while balancing resolution and analysis time.

Main Methods:

• Implemented a Bayesian optimization framework with a multi-objective chromatographic response function.
• Utilized a dynamic weighting strategy to shift optimization priority from resolution to throughput.
• Performed optimization within 35 experimental trials.

Main Results:

• Achieved a 28.3% increase in detected peaks and a 62.5% improvement in average resolution compared to default settings.
• Reached baseline separation (Rs≥1.5) for most volatiles, with significant reductions in retention time (75%) and peak area (60%) RSD.
• Successfully discriminated Baijiu samples with varying ratios, capturing 82.4% of total variance.

Conclusions:

• The BO framework provides a data-efficient solution for automated method development in high-speed GC systems.
• This strategy overcomes limitations of physicochemical modeling under thermal non-equilibrium conditions.
• Optimized UFGC-E-nose methods demonstrate practical utility for complex sample analysis and quality control.