Microwave moisture sensor with nonlinear data processing using artificial neural networks

The application of resonators as sensors for industrial microwave moisture measurement is advantageous, because resonators show a high sensitivity and require only scalar measurement systems. The theoretical background of the measurement procedure often assumes single tuned behaviour of the resonator and only weak distortion of the resonance through the measured material. These assumptions are often not fulfilled in practice and measurements fail. In this paper, artificial neural networks (ANNs) are proposed as a solution. With the example of a fringing field resonator, having odd resonance characteristics, and by using nonlinear data processing, it is shown that density independent moisture measurement is possible with excellent accuracy with the proposed approach. Although the example is on moisture measurement, it is believed, that the approach is applicable to other dielectric measurements based on complex detuning of resonators