Plant Optimisation Using Process Analytical Techniques

Abstract

This thesis addresses the challenge posed by the large impact that surfactant raw material composition has on final product quality attributes in shampoo manufacturing. Best practice to date has required pausing production whilst off-line product quality testing is performed before final corrections are made to ensure product specifications are met. This work demonstrates the implementation of in-line process analytical tools, combined with adaptive predictive process control models, to the optimisation of shampoo manufacturing, with the result of reduced product quality failures during manufacturing. Mass spectrometry gas chromatography has been used to determine compositional differences between samples of sodium lauryl ether sulfate and for the laboratory quantification of unsulfated alcohols within the surfactant. In-line near infrared, mid infrared and Raman spectroscopic techniques have been explored as methods for real time acquisition of surfactant compositional information as alternatives to gas chromatography techniques which are time and resource intensive. The challenge of using spectroscopic methods for this application is the absence of pure component spectra for each compound within the surfactant raw material and the need to distinguish the R-OH feature of the unsulfated alcohols from the OH features of water within the surfactant material. Using partial least squares regression calibration techniques and data pre-processing optimisation; in-line near infrared absorption spectroscopy was proven to be the most successful method for the real-time determination of unsulfated alcohols in sodium lauryl ether sulfate against statistical measures. The model provides an RMSECV of 0.09 w/w% and was externally validated against an independent set of samples to produce an RMSEP of 0.12 w/w%. The measurement of surfactant quality alone is not enough for the successful control of a shampoo manufacturing process. The impact of this variation on product quality, was experimentally investigated and empirical models were created to relate surfactant quality, product formulation, and product critical quality attributes. These model allow for not only the prediction of critical quality attributes, but also the calculation of the required formulation to meet a pre-defined quality specification. By using such models in combination with the use of the in-line near infrared measurement technique the rate of product meeting quality specifications without the need for quality adjustment was successfully increased from 57% to 90%. This work not only reduces the compositional analysis time of the surfactant from 1 hour using gas chromatography, to 16 seconds using NIR spectroscopy, it also removes the need for operator input making it a simple and viable approach to control real-time manufacturing.

Date of Award	23 Nov 2020
Original language	English
Awarding Institution	The University of Manchester
Supervisor	Peter Martin (Main Supervisor), Philip Martin (Co Supervisor) & Barry Lennox (Co Supervisor)