This dissertation introduces machine learning techniques to enhance laser absorption spectroscopy for detecting multiple gas species, particularly in challenging environments. It details methods like deep denoising autoencoders for improving signal quality in high-speed pyrolysis experiments and an unsupervised framework to handle interference from unknown species. The research also presents a blind source separation technique for reconstructing concentrations and spectral signatures without prior calibration data, and methods for recovering weakly absorbing species. These ML-enhanced approaches are experimentally validated and aim to enable real-time, interference-resilient gas detection for applications in combustion science, environmental monitoring, and industrial safety. AI
Summary written by gemini-2.5-flash-lite from 1 source. How we write summaries →
IMPACT Introduces advanced ML techniques for enhanced gas detection, potentially improving industrial safety and environmental monitoring.
RANK_REASON This is a research paper detailing novel machine learning methods applied to laser spectroscopy. [lever_c_demoted from research: ic=1 ai=1.0]