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

The recent development of new mid infrared prediction equations such as cow related phenotypes have obliged the use of standardization procedure. On one hand it permits transfer of prediction equations between instruments of the same or/and different brands. On another hand, it stabilizes and corrects monthly the deviations resulting from hardware modification over time. In the approach proposed within the frame of the Optimir project, each spectrometer is monthly standardized against a common reference. This is done by analyzing common sets of raw milk samples on all analyzers and correcting each one wavenumber by wavenumber. This method is applied on a network of instruments in several countries and allowed to successfully develop common predictions equations for new parameters. Most of all, it permits the transfer of these models on all analyzers and through the monthly harmonization to provide routinely predictions for management of dairy farms. However, between two monthly standardizations, there is currently a lack of knowledge regarding the stability of the individual spectrometers. So far there is no quality insurance system to monitor daily the stability of the predictions for the new parameters such as cow status. A complementary system is consequently needed to insure daily accuracy and validity of predictions. The aim of this presentation is to evaluate the potential of a a new approach being developed to monitor and stabilize daily the spectral data. The daily monitoring system has been developed to provide a tool to dairy labs providing a visual representation of the spectral stability. A daily stabilization system has been developed to correct the raw spectral data when perturbation of drift are detected. The first results obtained show that this method can help lab manager to have a precise monitoring of the raw spectral data, being complementary to the other quality insurance systems focusing on final predictions. Secondly the results show that the approach has an interesting potential to be used in routine for stabilizing all the new MIR equations predicting indirect parameters.