Neural network classification and quantification of organic vapors based on fluorescence data from a fiber-optic...

Area of Science:

• Analytical Chemistry
• Computational Science

Background:

• Accurate detection and quantification of organic vapors are crucial in environmental monitoring and industrial safety.
• Existing sensor technologies may face challenges in specificity and sensitivity for complex mixtures.

Purpose of the Study:

• To develop and validate computational neural network models for classifying and quantifying nine specific organic vapors.
• To assess the performance of these models using a fluorescence-based sensor array.

Main Methods:

• Utilized a sensor array of 19 fiber optics with immobilized dyes in polymer matrices.
• Collected fluorescence intensity change data over time upon exposure to organic vapor analytes.
• Calculated descriptors from intensity-time plots to train neural network models.

Main Results:

• Achieved classification rates approaching 100% for training data.
• Correctly classified 90% of organic vapor analytes in the prediction set.
• Assigned correct relative concentrations to 97% of prediction set observations.

Conclusions:

• Computational neural networks effectively classify and quantify multiple organic vapors with high accuracy.
• The developed models demonstrate robust performance for real-world applications in vapor sensing.
• Fluorescence-based sensor arrays coupled with neural networks offer a promising approach for chemical detection.