Multi‐Channel Surface Acoustic Wave (SAW) Sensor Based on Artificial Back Propagation Neural (BPN) Network and Multivariate Linear Regression Analysis (MLR) for Organic Vapors

Abstract

AbstractA six‐channel surface acoustic wave (SAW) detection system with a 315 MHz one‐port quartz resonator and a homemade computer interface for signal acquisition and data processing was developed to detect various organic vapors. The oscillating frequency of the SAW quartz crystal decreased due to the adsorption of organic molecules on the coating materials. Polyethylene glycol, 18 crown 6 (18C6), Cr3+/cryptand‐22, stearic acid, polyvinylpyrrolidene and triphenyl phosphine coated quartz crystals were used as sensors. An artificial back propagation neural (BPN) network was used to recognize various organic gases such as hexane, 1‐hexene, 1‐hexyne, 1‐propanol, propionaldehyde, propionic acid and 1‐propylamine. It showed not only the distinction of unity of organic vapors but also mixtures of gases. The learning rate and the hidden unit of a neural network system for BPN analysis were investigated. Furthermore, the concentrations of these organic vapors were computed with about 10% error by multivariate linear regression analysis (MLR). MLR analysis with a multichannel SAW sensor was applied to determine the concentration of each component in a mixture of 1‐hexene, 1‐hexyne and propionaldehyde.