Machine Learning to Immune Microwave Sensors from Temperature Impact

In this paper a planar microwave sensor is used for contactless material characterization at~2.7 GHz. The extraneous impact of temperature is shown to cause interleaved data (amplitude) traces for various material that leads to potential confusion for sensor readout. This issue is resolved with proper Artificial Neural Network (ANN) design with single hidden layer. Various concentrations of water in acetone (0 - 50% with 10% increments) are exposed to temperature rise of 25°C - 60°C. The output of the proposed system confirms successful discrimination of the flowing aqueous solutions up to 91% accuracy.