Piezoelectric Smart Patch Operated with Machine-Learning Algorithms for Effective Detection and Elimination of Condensation

Qian Zhang,Yong Wang,Tao Wang,Dongsheng Li,Jin Xie,Hamdi Torun,Yongqing Fu

DOI: https://doi.org/10.1021/acssensors.1c01187

IF: 8.9

2021-08-18

ACS Sensors

Abstract:Timely detection and elimination of surface condensation is crucial for diverse applications in agriculture, automotive, oil and gas industries, and respiratory monitoring. In this paper, a smart patch based on a ZnO/aluminum (∼5 μm/50 μm thick) flexible Lamb wave device has been proposed to detect, prevent, and eliminate condensation, which can be realized using both of its surfaces. The patch is operated using a machine-learning algorithm which consists of data preprocessing (feature selection and optimization) and model training by a random forest algorithm. It has been tested in six cases, and the results show good detection performance with average precision = 94.40% and average F1 score = 93.23%. The principle of accelerating evaporation is investigated to understand the elimination and prevention functions for surface condensation. Results show that both dielectric heating and acoustothermal effect have their contributions, whereas the former is found more dominant. Furthermore, the functional relationship between the evaporation rate and the input power is calibrated, showing a high linearity (R2 = 97.64%) with a slope of ∼3.6 × 10–5 1/(s·mW). With an input power of ∼0.6 W, the flexible device has been proven effective in the prevention of condensation.The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acssensors.1c01187.Effect of flow rate on the generation of condensation (Figure S1); temperature sensitivity (TCF) of the flexible device (Figure S2); S21 spectra recorded in the temperature, humidity, and condensation experiments (Figure S3); classification performance using random forest, SVM, and NBM (Figure S4); normalized response to respiration in the form of the seven features and resonant frequency (Figure S5); investigation of robustness on the frequency band selection (Figure S6); evaporation time of a single deionized water droplet (∼1 μL) when the flexible device is excited by four different frequencies (Figure S7); investigation of the influence of bending on heating performance (Table S1); calculation method of the seven features; training method of the random forest classifier (PDF)This article has not yet been cited by other publications.

chemistry, multidisciplinary, analytical,nanoscience & nanotechnology