Use of in-line Near Infrared Spectroscopy to Predict the Viscosity of Shampoo using Multivariate Analysis.

Objective In the personal care industry, viscosity is a critical quality attribute that influences product quality and process economics. Like many industrial liquids, personal care liquids are complex non-Newtonian liquids made up of aqueous surfactant systems whose viscosity depends on the build-up of micellar networks. Measuring the viscosity of complex liquids offline is easily done using benchtop rheometers and viscometers. The challenge lies in measuring the viscosity of personal care liquids online during manufacturing. Being able to track the viscosity of such products through their manufacturing cycle will not only allow for better process control but also more enhanced quality control. Therefore, the aim of this work was to investigate how proxy measurements using inline near-infrared (NIR) spectroscopy in transmission mode can be used to predict the viscosity of shampoo. NIR spectroscopy has not, to the best our knowledge, been used to predict the viscosity of complex surfactant systems like shampoo and could significantly affect the way quality is monitored in a manufacturing environment. Method This work focuses on viscosity changes because of differences in chloride content as salt is often used to adjust viscosity. The relationship between salt content and the viscosity of shampoo is well known following the salt curve. From an industrial perspective the region of interest for the formulation studied in this work only covers a small section of this curve. Therefore, two predictive models were developed: one covering the full range of the salt curve and another focusing on the industrially applicable region. Result Models were produced using partial least squares (PLS) where both datasets showed some predictive ability with the concentrated region of interest showing enhanced performance [root mean square error of prediction (RMSEP) - 2.32 Pa s] compared with the larger range (RMSEP - 4.44 Pa s). Conclusion This work provides a good starting point for developing robust predictive models for in situ viscosity measurements for shampoo manufacturing, where further work into different sources of variation and the extent of the modelling capability with regards to different formulations should be studied.