Accurate identification of gas type and concentration using DNN reflecting the sensing properties of MOSFET-type gas sensor

Abstract

The principal component analysis (PCA) and deep neural network (DNN) are used to classify the gas types (reducing and oxidizing) and to identify the concentration of gases. The pMOSFET-type gas sensor is used to provide sensing data for learning. The gas sensor has 15-nm-thick ZnO as a sensing layer processed by atomic layer deposition (ALD). The sensing characteristics of NO 2 and H 2 S gases are investigated in changing temperature and concentration conditions. The same gas sensor data and temperature sensor data are analyzed by PCA and DNN algorithms. PCA provides gas type classification results without information on gas concentration. However, DNN regression model has the ability to precisely identify gas concentration and gas type in changing temperature conditions.