Headspace Oxygen Concentration Measurement for Pharmaceutical Glass Bottles in Open-Path Optical Environment Using TDLAS/WMS

Abstract

Accurate measurement of oxygen concentration for pharmaceutical glass bottles is of great significance to ensure the asepsis of medicine and stability of ingredients. With merits of high sensitivity, low cost, non-contact, and real time response, the wavelength-modulation-based tunable diode laser absorption spectroscopy (TDLAS/WMS) technology shows great potential to achieve in-site oxygen concentration detection by the single-line spectrum measurement. This paper focuses on headspace oxygen concentration measurement in open-path optical environment, which is extremely challenging owing to the short light path length and random ambient noises. First, a signal reconstruction method is established based on discrete-time wavelet packet transform (DWPT), where random noise suppression is implicitly achieved. Then, oxygen concentration is inversed among the data between two adjacent valley values of the demodulated 2 nd harmonic signal by multiple linear regression (MLR), and the linear discriminant analysis (LDA) is imported to enhance the information sparsity of 2 nd harmonic signal matrix and to address the multi-collinearity problem. Simulation results prove our proposed method achieved considerable detection accuracy with average absolute error of 0.05%. This paper also designed a TDLAS/WMS prototype, the experimental results aiming at glass bottles with different oxygen concentration of 0%, 5%, 10% and 21% in open path optical environment indicate our method has achieved an encouraging average absolute error of 0.54%, and can survive well when the normalized SNR is within 0.85 to 1. These results promise that the proposed methodology can be widely applied in in-site AOI instrumentation of headspace oxygen concentration measurement for glass vials.

EWT-ASG: Empirical Wavelet Transform With Adaptive Savitzky–Golay Filtering for TDLAS

“…3), where our proposed EWT-ASG algorithm will be carried out and verified. It is worth to mention that the parameters and configurations of this prototype are the same as our previous work [1] and the temperature range in the workshop is 25°C to 32°C.The raw 2nd harmonics of 0% oxygen concentration for pharmaceutical glass bottle are obtained, taking one of the raw 2nd harmonics as an example for subsequent analysis and discussion, then a 2nd harmonic f(k) is shown in Fig. 4 (black dashed line).…”

“…4 to show the performance of our proposed SWT-ASG algorithm. To be consistent with the Reference [1], which used discrete-time wavelet packet transform (DWPT) with different mother wavelet functions to generate the signal reconstructor, we choose the same three typical wavelet functions of Coiflets, Symlets and Meyer as mother wavelet function, where the amplitude drop is in accord with the result in [1] when using Meyer. As the waveform of Meyer has a considerable discrepancy with the ideal 2nd harmonic, the result of the convolution of the raw 2nd signal with the infinite impulse response filter can not fit the symmetric 2nd harmonic well.…”

“…Coiflets, Symlets) have small amplitude attenuation with the property describing the raw principal trend well. It is worth mentioning that the reconstructed 2nd harmonic by DWPT cannot reach the quality of the filtered signal shown in [1], for the reason that DWPT can hardly extract the main signal from the more complex noises existing in the industry field, such as the constant pulse interference Fig. 6.…”

“…In order to ensure the stability of pharmaceutical ingredients, the measurement of oxygen concentration in pharmaceutical glass vials has become an urgent problem to be solved. With the merits of single-line spectrum analysis, noncontact detection and real-time response in the laboratory [2], [3], [4], [5], the wavelength-modulation-based tunable diode laser absorption spectroscopy (TDLAS/WMS) [1] performs preeminently on the oxygen concentration detection.…”

“…Therefore, many de-noising methods have been proposed successively. The signal averaging [12], Wiener filter [13], Kalman filter [14] and Wavelet Transform filter [1] have been applied for signal processing in the TDLAS system and improve the precise of the detection. The above filter methods target linear but non-stationary data and non-linear but stationary and deterministic systems.…”