1. Harmonic Amplitude Pass Probability: A Lightweight Signal Descriptor for Oxygen Concentration Detection in Pharmaceutical Glass Vials.
- Author
-
Liu, Zihuai, Li, Yonggang, Luo, Qiwu, Yang, Chunhua, and Gui, Weihua
- Subjects
- *
BAYESIAN analysis , *TUNABLE lasers , *SEMICONDUCTOR lasers , *AUTOMATIC optical inspection , *LASER spectroscopy , *NAIVE Bayes classification - Abstract
In situ and accurate oxygen concentration detection in pharmaceutical glass vials is of great significance in the development of the pharmaceutical industry. This article reveals a universal data distribution rule in which each sampling point follows an approximately normal distribution with its own mean ($\mu $) and standard deviation ($\sigma $) in a wavelength modulation-based tunable diode laser absorption spectroscopy (TDLAS/WMS) system. Based on this investigation, we propose a novel concept, harmonic amplitude pass probability (HAPP), where noisy data points in the demodulated second harmonic signal can be cleverly transformed into informative probability labels. That is, the awkward interference components in the harmonic data are implicitly washed out. Innovatively, the HAPP acts as a smart bridge between the raw harmonic data and Bayesian network. The HAPP-based Bayesian classifier achieves an average detection accuracy of 97.3% when fed with actual data captured from a penicillin encapsulation line, which presents a data-mining methodology for online package leak detection from a probability statistics perspective. [ABSTRACT FROM AUTHOR]
- Published
- 2022
- Full Text
- View/download PDF