Identification of Alkali Metal Ions Using Acoustic Plate Mode Sensors and Adaptive Network-Based Fuzzy Inference Systems

Chemical sensors have been shown to be an effective means for detecting molecules in gas or liquid phases. Improvements in the abilities of chemical sensors to detect specific chemical molecules are made by either developing chemically selective sensor coatings, by using advanced pattern recognition techniques to analyze sensor data, or through a combination thereof. Past studies have shown that both conventional pattern recognition techniques based upon multivariate statistics and numerical analysis and artificial neural networks (ANN) provide effective means to analyze data produced by acoustic plate mode (APM) delay line sensors. The current study focuses research on the use of neuro-fuzzy systems, specifically the adaptive network-based fuzzy inference system (ANFIS), to process the data produced by APM sensors. The research presented in this paper shows that through proper construction and training, ANFIS networks provide an effective means to analyze the data produced by the APM sensors. The ANFIS based system performs comparably to the ANN based system while reducing the amount of data required to identify ions in single component solutions and estimate their concentrations. The possibility of using ANFIS based systems to estimate relative concentrations of ions in binary mixtures is also investigated. These preliminary investigations lay a groundwork for further research into the use of ANFIS based systems to identify alkali metal ions in binary solutions