Abstracts Economics

A Multi-Stream Quality Monitoring and Control System for Flour Mills

by John Kalitsis

Flour mills require a significant capital investment and operate at low returns, with wheat representing the highest operating cost. This thesis explores opportunities to improve milling plant and raw material use, increasing the rate of return of a milling operation. Options considered included optimising flour mill performance using experimental process optimisation techniques, mathematical computer modelling for flour stream blending, and low-cost inline near-infrared sensors to achieve continuous real-time measurement of end-product and intermediary flour streams. This thesis aims to demonstrate how these approaches, on their own or in combination, improve flour mill utilisation, efficiency and profitability. Improvements in mill efficiency were achieved by improving flour yield and blending flour streams to hit target flour type quantities and specifications. The improvements were achieved by better understanding the flour milling process by developing models using response surface methodology and mathematical optimisation techniques, such as linear programming and generalised reduced gradient, to improve the economics of flour stream blending. These benefits were enhanced by installing inline low-cost and accurate near-infrared spectrometers that provide a reliable and accurate prediction of flour constituents. There are opportunities to improve the optimisation and flour blending models through machine learning to incorporate a more significant number of mill processing variables and a more comprehensive range of wheat grists. In addition, alternative calibration techniques for the NIRS sensors may deliver improved accuracy and robustness, improving this technology's benefits to the milling industry. This research identified these opportunities and is the scope for potential future work.