A key uncertainty when delivering bauxite ore to an alumina factory is the moisture content in the shipment. A microwave analyser measures phase shift and attenuation in real time, of microwaves passing through the ore while it is being offloaded on a conveyor belt. This data is used to infer moisture content, which affects the weight of the ore, and directly impacts the price paid per tonne. Our study is informed by data provided to a European Study Group with Industry that was collected from a number of shipments of bauxite ore in 2017.

Simple linear models fail to explain the highly nonlinear dependence on bauxite depth that is seen in plots of analyser data. These simple linear models are presently used by the microwave analyser software to infer moisture content.

Motivated by this failure, we use Maxwell's equations to develop a four-layer model that allows multiple reflections at interfaces between ore and air. Our model goes some distance towards improving real-time measurement of ore moisture content using microwaves on a conveyor belt. The model should extend to any mixture of solid ore, air and water on a conveyor belt. A continuing challenge is how to share in a useful way the four-layer model solutions with a company that makes the microwave analysers.