Sensor-Array Optimization Based on Time-Series Data Analytics for Sanitation-Related Malodor Detection.

Area of Science:

• Environmental Science
• Sensor Technology
• Machine Learning

Background:

• Shared sanitation facilities and onsite waste treatment require effective malodor detection systems.
• Existing technologies are often costly or lack specificity for sanitation-related odors.

Purpose of the Study:

• To develop a low-cost instrumented technology for detecting sanitation-related malodor.
• To utilize electrochemical gas sensors and machine learning for odor classification and sensor selection.

Main Methods:

• Screened 10 electrochemical gas sensors with target gases and odor samples (fecal, urine, popcorn).
• Applied machine learning techniques, including feature selection based on mutual information, for odor analysis.
• Evaluated Decision Tree (DT), K-Nearest Neighbors (KNN), and Logistic Regression (LR) classifiers for malodor detection and classification.

Main Results:

• A Decision Tree classifier with 7 features from 4 sensors achieved 88.0% balanced accuracy for malodor detection.
• A K-Nearest Neighbors classifier with 3 features from 2 sensors reached 83.3% balanced accuracy.
• A Logistic Regression classifier with 4 features from 3 sensors attained 94.0% accuracy in distinguishing urine from feces malodor.

Conclusions:

• The developed electronic nose offers a promising low-cost solution for detecting sanitation-related malodor.
• Machine learning effectively reduces sensor count and enhances odor classification accuracy.
• This technology can improve maintenance alerts for sanitation facilities and support waste treatment monitoring.