Presentation of a new system to monitor and stabilize mid infrared spectral data

Size: px

Start display at page:

Download "Presentation of a new system to monitor and stabilize mid infrared spectral data"

Arnold Banks
5 years ago
Views:

1 Presentation of a new system to monitor and stabilize mid infrared spectral data Clément Grelet, Vincent Baeten, Pierre Dardenne, Juan Antonio Fernandez Garcia, Ouissam Abbas, & Frédéric Dehareng Walloon Agricultural Research Centre, Belgium

2 C-O, OH CH Lactose C-O CH C=O, N-H, C-N C=O Fat Fat + protein Fat Protein Fat, organic acids Position of the peaks Qualitative analysis Intensity of the peaks Quantitative analysis

3 But Spectra of a same milk could differ: - between different brands apparatus - between different models apparatus of the same brand - between apparatus of the same model of the same brand Moreover, even with the same instrument, the spectra could be different for the same milk. It's not stable in time! -T /humidity in the lab -Piece replacement -Maintenance operation -Use/wear

Milk component Fat, Protein, Lactose, SNF Casein Hewavitharana et al. 1997. Analyst 122:701 704. Urea Hansen, P. W. 1998.

4 Milk component Fat, Protein, Lactose, SNF Casein Hewavitharana et al Analyst 122: Urea Hansen, P. W Milchwissenschaft 53: Milk MIR spectra Classical slope/bias correction Fatty acids Soyeurt et al J. Dairy Sci. 89: Reference

Spectral Standardization Milk MIR spectra Phenotype Body Energy status Mc Parland et al. 2011 J. Dairy Sci. 94: 3651 3661 Methane Dehareng et al. 2012 Animal.

5 Spectral Standardization Milk MIR spectra Phenotype Body Energy status Mc Parland et al J. Dairy Sci. 94: Methane Dehareng et al Animal. 6 : Milk Indirect Lactoferrin Soyeurt et al J. Dairy Sci. 90: Major minerals Soyeurt et al J. Dairy Sci. 92: Blood component Blood BHB and NEFA M. Gelé et al. 2015, ICAR Coagulation, titrable acidity, ph De Marchi et al J. Dairy Sci. 92: Acetone, β-hydroxybutyrate, and citrate Grelet et al J. Dairy Sci. 99 :

7 Project : Develop MIR models predicting cow status PIECE-WISE DIRECT STANDARDIZATION (PDS)

But PDS Standardization is a solution to

these problems - between different brands

with the same instrument, the spectra could be

8 But PDS Standardization is a solution to Spectra of a same milk could differ: solve these problems - between different brands apparatus - between different models apparatus of the same brand - between apparatus of the same model of the same brand Moreover, even with the same instrument, the spectra could be different for the same milk. It's not stable in time! -T /humidity in the lab -Piece replacement -Maintenance operation -Use 100 machines 3 continents 13 countries 40 labs

9 Creation of common models, more robust Use of models on all instruments Sharing of data/models Creation of spectral database

10 Daily Monitoring and Stabilization of spectra 1. Monitoring of the stability between 2 PDS standardization 2. If needed, realization of a stabilization procedure

11 Dailycheck Monitoring of the stability in time Standardization 3 Standardization 1 Standardization 2 Strong perturbation biased predictions Creation of an alert system for the labs Month 1 Month 2

12 Dailycheck Monitoring of the stability in time Standardization 1 Just after the standardization: 10 UHT milks constituting the stability reference

13 Dailycheck Monitoring of the stability in time Based on the GH (Global H) which is the standardised Mahalanobis distance 10 UHT milks constituting the stability reference

14 Dailycheck Monitoring of the stability in time Based on the GH (Global H) which is the standardised Mahalanobis distance 10 UHT milks constituting the stability reference Variability covered by the stability reference New daily-check sample

15 Dailycheck Monitoring of the stability in time Based on the GH (Global H) which is the standardised Mahalanobis distance GH = 2 OK 10 UHT milks constituting the stability reference Variability covered by the stability reference 2 New daily-check sample

16 Dailycheck Monitoring of the stability in time Based on the GH (Global H) which is the standardised Mahalanobis distance GH = UHT milks constituting the stability reference Variability covered by the stability reference New daily-check sample

17 Dailycheck Monitoring of the stability in time Based on the GH (Global H) which is the standardised Mahalanobis distance GH = 20 NOT OK 10 UHT milks constituting the stability reference Variability covered by the stability reference New daily-check sample

18 Dailycheck Monitoring of the stability in time Standardization 1 After n samples out: Alert! Just after the standardization: 10 UHT milks constituting the stability reference

19 Dailycheck Monitoring of the stability in time

20 Dailycheck Monitoring of the stability in time

21 Dailycheck Monitoring of the stability in time Alert System Standardization Just after the standardization: 10 UHT milks constituting the stability reference

22 Dailycheck Monitoring of the stability in time Alert System Standardization Stabilization Standardization Just after the standardization: 10 UHT milks constituting the stability reference

23 Stabilization GH of raw daily check Calibration Validation Impact on GH GH GH validation= Samples GH of stabilized daily check Calibration Validation Coefficients GH GH validation= Samples

24 Stabilization SD fat validation= Fat predicted values of raw daily check Impact on Fat Predicted fat (g/100ml) Calibration Validation Samples Predicted fat (g/100ml) SD fat validation= Fat predicted values of stabilized daily check Calibration Validation Coefficients Samples

25 Conclusion: Routine standardization: Creation of common models, more robust Use of models on all instruments Sharing of data/models Creation of spectral database Alert detection: Detection of all deviations/perturbations between 2 monthly ring test based on GH and/or Parameter measurements Stabilization: Potential to use models in routine

Full spectral approach fostering the development of new innovative concepts

Full spectral approach fostering the development of new innovative concepts Gavin Scott, Silvia Orlandini & Frédéric Dehareng 15-11-216 Version 1. ICR Puerto Varas, Chile 216 Introduction 1961 Dr Goulden