Intra-Class Classification of Volatile Amines using Impedance Spectroscopy Platform

Amines play a significant role in everyday life, and their detection remains a crucial focus in research and development. Although conducting polymer-based gas sensors have been widely reported for amine detection using DC resistivity measurements, they often lack selectivity to distinguish between intragroup compounds. To overcome this limitation, this work reports a machine learning-enabled electrochemical impedance spectroscopy (EIS) approach using a tin oxide–polypyrrole composite sensor, forming a virtual sensor array for intraclass amine classification. The sensor's impedance response to four representative amines, ammonia, dimethylamine, trimethylamine, and trimethylamine, was investigated and classified over a wide frequency range (100 Hz to 8 MHz). Electrical parameters were extracted from the obtained impedance data from Nyquist plots using equivalent circuit models with an overall fitting error of less than 3%. It was observed that the sensor exhibits frequency-dependent impedance behaviour in response to different amines and can be classified using a unique set of frequencies as features. Principal component analysis was employed to visualize 20 distinct classes of the four different amines at five concentrations each (4x5 matrix). A support vector machine (SVM) classifier was trained and tested on the obtained dataset, with 40%:60%, respectively. The tested data was classified with an accuracy greater than 90% for all 20 classes. In addition, a wrapper approach was employed for feature selection to enhance the model efficiency. This work establishes EIS combined with machine learning as a powerful strategy for selective classification of different amines and highlights its real-time prediction capability.