Fluid Classification\nvia the Dual Functionality of\nMoisture-Enabled Electricity Generation Enhanced by Deep Learning

Classifications of fluids using miniaturized sensors\nare of substantial\nimportance for various fields of application. Modified with functional\nnanomaterials, a moisture-enabled electricity generation (MEG) device\ncan execute a dual-purpose operation as both a self-powered framework\nand a fluid detection platform. In this study, a novel intelligent\nself-sustained sensing approach was implemented by integrating MEG\nwith deep learning in microfluidics. Following a multilayer design,\nthe MEG device including three individual units for power generation/fluid\nclassification was fabricated in this study by using nonwoven fabrics,\nhydroxylated carbon nanotubes, poly(vinyl alcohol)-mixed gels, and\nindium tin bismuth liquid alloy. A composite configuration utilizing\nhydrophobic microfluidic channels and hydrophilic porous substrates\nwas conducive to self-regulation of the on-chip flow. As a generator,\nthe MEG device was capable of maintaining a continuous and stable\npower output for at least 6 h. As a sensor, the on-chip units synchronously\nmeasured the voltage (V), current (C), and resistance (R) signals\nas functions of time, whose transitions were completed using relays.\nThese signals can serve as straightforward indicators of a fluid presence,\nsuch as the distinctive “fingerprint”. After normalization\nand Fourier transform of raw V/C/R signals, a lightweight deep learning\nmodel (wide-kernel deep convolutional neural network, WDCNN) was employed\nfor classifying pure water, kiwifruit, clementine, and lemon juices.\nIn particular, the accuracy of the sample distinction using the WDCNN\nmodel was 100% within 15 s. The proposed integration of MEG, microfluidics,\nand deep learning provides a novel paradigm for the development of\nsustainable intelligent environmental perception, as well as new prospects\nfor innovations in analytical science and smart instruments.