A Mass Spectrometry-Machine Learning Approach for Detecting Volatile Organic Compound Emissions for Early Fire...

Area of Science:

• Analytical Chemistry
• Materials Science
• Computational Chemistry

Background:

• Early fire detection is crucial for safety and damage mitigation.
• Identifying specific chemical signatures of burning materials can provide early warnings.
• Traditional methods may lack the specificity or speed for certain fire scenarios.

Purpose of the Study:

• To develop and validate a novel technique for real-time, chemically specific detection of fire indicators.
• To investigate the use of mass spectrometry coupled with machine learning for identifying volatile organic compounds (VOCs) from thermal decomposition.
• To assess the accuracy of this method in distinguishing between different materials and their mixtures.

Main Methods:

• Utilized quadrupole mass spectrometry to characterize VOCs from the thermal decomposition of Mylar, Teflon, and poly(methyl methacrylate) (PMMA).
• Collected mass spectral data, identifying unique chemical signatures (patterns of peaks) for each material.
• Applied a random forest machine learning classification model to analyze spectral data sets for material identification.

Main Results:

• Unique mass spectral patterns were identified for Mylar, Teflon, and PMMA thermal decomposition.
• These chemical signatures remained detectable and consistent when materials were heated together.
• The machine learning model achieved 100% accuracy for single material spectra and 92.3% for mixed material spectra.

Conclusions:

• Mass spectrometry combined with real-time machine learning offers a novel, rapid, and accurate method for fire event detection.
• The technique successfully identifies chemically specific early indicators of fires based on VOC emissions.
• This approach shows significant promise for enhanced fire safety systems through precise chemical signature analysis.