CuO Nanowires-Based Resistive Sensor for Accurate Classification of Multiple Vapors

• Published in: IEEE sensors journal Vol. 23; no. 10; pp. 10293 - 10300
• Main Authors: Kulkarni, Saraswati, Bharath, B. N., Ghosh, Ruma
• Format: Journal Article
• Language: English
• Published: New York IEEE 15.05.2023; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Acetonitrile; Algorithms; Biomarkers; Circuits; Classification; Classification algorithms; computational complexities; Crystallites; CuO-based resistive sensors; Discrete wavelet transforms; Environmental monitoring; Field emission microscopy; Filter banks; Machine learning; machine learning (ML) algorithms; Metals; Nanomaterials; Nanowires; Room temperature; selectivity; Sensors; Support vector machines; Temperature sensors; VOCs; volatile organic compound (VOC) detection; Volatile organic compounds
• ISSN: 1530-437X; 1558-1748
• DOI: 10.1109/JSEN.2023.3262877

Developing technological solutions for identifying multiple gases using a minimum number of sensors is critical for several applications including environmental monitoring, disease diagnosis, etc. This article reports the surfactant-mediated growth of CuO nanowires using a simple co-precipitation technique. X-ray diffractometer (XRD) studies reveal the monoclinic phase, and the crystallite size was calculated to be approximately 16 nm. Field emission scanning electron microscopic (FESEM) images showed the 1-D nanowires-like morphology of CuO. Using diffuse reflectance spectroscopy (DRS), the band gap of the material was calculated to be around 1.42 eV. The nanomaterial was employed for testing four concentrations (500-5000 ppm) of four volatile organic compounds (VOCs), namely, methanol, acetonitrile, acetone, and isopropanol at room temperature. Discrimination of VOCs is a highly desired property of vapor sensors, but the CuO sensor was observed to fail while differentiating between these VOCs if only the steady state response of the sensor for these VOCs were considered. Hence, an attempt to find a correlation between the VOCs and their concentrations with features like response, response times, and recovery times was taken for classification and regression using five machine learning (ML) algorithms-support vector machines (SVMs), random forest (RF), <inline-formula> <tex-math notation="LaTeX">{k} </tex-math></inline-formula>-nearest neighbors (KNNs), naïve Bayes (NB), and linear regression. An approximate estimation of the computation complexities of these ML algorithms for the dataset generated in this research was made to find the algorithm with the highest accuracy and minimum resource requirements for their circuit-level implementations.