Description
Semen extenders (SE) are zootechnical products added to animal semen as diluent and preservative before artificial insemination procedures. A key-step of SE production is the blending phase; the definition of the process endpoint is, usually, done according to a specific blending duration, decided on experience bases.
Aim of the present work is the development of a strategy for the monitoring of the blending phase in a continuous and non-destructive way, determining a specific process endpoint for every single batch, in a quality by design (QbD) view.
In order to reach the aim, a MicroNIR PAT-U sensor (Viavi Solutions), working in the spectral range between 900 and 1700 nm, was directly clamped on the blender, enabling the acquisition of a spectrum every 4 s, through a sapphire window during the whole process. Before the implementation of the NIR sensor, the endpoint of SE blending was set at 20 min. Spectra were acquired for 20 batches of SE, used as calibration set, and the strategy was validated on other 6 independent batches.
On the collected signals, a proper data pretreatment was performed and then, standard deviation spectra were calculated applying a moving-block strategy. After a proper mean-centering, principal component analysis (PCA) was applied on the calibration batches, and the validation batches were projected into the space defined by the two lowest-order principal components. The influence plot (Hotelling’s T2 vs. Q residuals) and its statistical boundaries at a 95% confidence level were implemented as a multivariate control chart, for the monitoring of the behavior of new batches in the orthogonal space defined by PCA. The endpoint criterion was defined after acceptance of 15 consecutive spectra accepted by the multivariate control chart. Thanks to the present approach, the average time of the process was reduced to 5 min.
The whole strategy was implemented in a dedicated software, called NIRNova, developed in house with Matlab App Designer. Thanks to NIRNova, the production of SE is now monitored continuously in a non-destructive way, with a definition of the process endpoint for every single batch. Uniformity tests demonstrated that, despite the significantly time reduction, the final product is properly mixed.
