Battery-Free, Artificial Neural Network-Assisted Microwave Resonator Array for Ice Detection

Abstract

Ice detection has developed into an integral part of aerospace and wind turbine sectors with the aim of preventing hazards and component breakdown. This article presents an ice detection system consisting of a battery-free, chip-less wirelessly interrogated resonator array, and an artificial neural network for enhanced detection robustness. The designed array of split-ring resonators (SRRs), operating at 3.05 GHz, was a narrowband frequency-selective structure with a ground plane reflector which shielded resonance from the effect of installation material. For an incident wave on the array’s surface, the reflection coefficient changed with the electrical properties of the ice and water, consequently affecting the <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$S_{11}$ </tex-math></inline-formula> parameter of an interrogator antenna. The array had a surface area of <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$13\times6.5$ </tex-math></inline-formula> cm2 and a substrate thickness of 0.79 mm and was wirelessly interrogated from a distance of 33 cm by a standard gain horn antenna. A custom LabView program was utilized for time-based data acquisition of the antenna’s reflection coefficient [ <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$S_{11}$ </tex-math></inline-formula> (dB)], with results demonstrating a resonant frequency shift of 150 MHz when 30 <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\mu \text{L}$ </tex-math></inline-formula> ( <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\sim$ </tex-math></inline-formula> ) of ice was formed on the main split of SRRs. The artificial neural network then classified the reflection coefficients of the interrogator antenna to enhance ice/water differentiation. The neural network improved the classification of the <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$S_{11}$ </tex-math></inline-formula> (dB) raw data to 94.67% while achieving an accuracy of 93.33% for a noisy simulated data set. The proposed battery-free artificial neural network-assisted ice sensor can be implemented in various sizes (depending on the footprint requirements of installation) with applications in aerospace and wind turbine industry.