System Design and SVM Identification Algorithm for the Ultrasonically Catalyzed Single-Sensor E-Nose

Abstract

In this article, a system design method and Support Vector Machine (SVM) gas identification algorithm for the ultrasonically catalyzed single-sensor E-nose is proposed and investigated for the first time. In the aspect of hardware, the E-nose system consists of an ultrasonic module, sensing module, data processing module, personal computer, and power supply module. In the aspect of software, a microcontroller in the data processing module is coded by C programming language for data acquisition and operational control, and a user interface is made on the PC by C# programming language for display and sending commands. Performance of the integrated E-nose is tested and evaluated by methanol, ethanol, acetone, and hydrogen gases, respectively. The experimental results show that the integrated E-nose can achieve up to 100% gas identification success rate in the concentration range greater than 10% LEL (LEL = lower explosive limit), with a concentration measurement error less than 10%, and the SVM algorithm can effectively improve the gas identification ability in the low-concentration range (1%–10% LEL). For 1% LEL methanol target gas, it raises the gas identification success rate from 0% to 90%.