Machine Learning-Based Near-Infrared Laser Leakage Detection System for Wine Bottles

Area of Science:

• Analytical Chemistry
• Spectroscopy
• Machine Learning

Background:

• Traditional wine bottle leakage detection methods are inefficient and prone to errors.
• Existing techniques struggle with detecting subtle micro-leakages and distinguishing spectral interferences.

Purpose of the Study:

• To develop a novel, rapid, and non-destructive system for wine bottle packaging leakage detection.
• To enhance detection accuracy and reduce false positives using advanced spectroscopy and machine learning.

Main Methods:

• A near-infrared laser system utilizing tunable diode laser absorption spectroscopy at 1392 nm was employed.
• Gaseous ethanol vapor from leaks was detected, with spectral data processed using baseline correction, normalization, and down-sampling.
• A LightGBM machine learning model was optimized via a two-stage hyperparameter search and 5-fold cross-validation.

Main Results:

• The LightGBM model achieved an AUC of 0.9949 for binary classification and R² of 0.5854 for regression.
• Robustness was confirmed with 0.94 overall accuracy and 0.99 alcohol recall in anti-interference tests.
• A simulated micro-leakage test yielded 0.95 accuracy with zero false negatives, with a total detection cycle under 5 seconds.

Conclusions:

• The proposed system offers a low-cost, reliable, and efficient solution for online wine bottle leakage detection.
• It successfully addresses limitations of traditional methods, enabling industrial deployment for packaging integrity assessment.