A Novel Artificial Intelligence-Enabled Method for Electronic Nose Design Based on Olfactometry Data.

Area of Science:

• Analytical Chemistry
• Sensor Technology
• Biomimicry

Background:

• Electronic nose (e-nose) systems objectively evaluate odors using sensor arrays and data processing.
• They offer rapid, reproducible, and standardized chemical measurements for various industries.
• Applications span food safety, environmental monitoring, medical diagnostics, and industrial quality control.

Purpose of the Study:

• To compare electronic nose sensor measurements with human olfactometry panelist assessments.
• To evaluate the predictive accuracy of machine learning algorithms for e-nose data.
• To validate the performance of e-nose technology using a standardized reference odor (n-butanol) and TS EN 13725 standard.

Main Methods:

• Measurements from e-nose sensors were compared against human panelist assessments.
• Machine learning algorithms including Partial Least Squares (PLS), Support Vector Regression (SVR), and Gaussian Process Regression (GPR) were employed.
• Data modeling and predictive accuracy evaluation were performed using these algorithms.

Main Results:

• The electronic nose system demonstrated reliability and applicability in odor assessment.
• Machine learning models achieved low training Mean Absolute Percentage Error (MAPE) values: 6.53% for PLS, 10.89% for SVR, and 0.15% for GPR.
• Gaussian Process Regression (GPR) showed particularly high predictive accuracy.

Conclusions:

• The study successfully assessed the performance of electronic nose technology.
• The use of a standardized reference odor and robust modeling approach confirmed the system's effectiveness.
• Electronic nose technology, enhanced by machine learning, provides a reliable method for objective odor analysis.