A Prototype to Detect Alcohol Content in Local Toddy using an Electronic Nose

Abstract

For the purpose of preserving product quality and guaranteeing regulatory compliance, accurate alcohol detection in alcoholic beverages, particularly in local toddy, is crucial. The goal of this study is to produce a reliable alcohol detection device employing electronic nose (e-nose) technology customized for local toddy. The envisioned device includes a sensor array that can detect and examine volatile substances linked to alcohol in local toddy. The system enables real-time alcohol content readings by gathering and analyzing sensor data, enabling accurate quality control and monitoring. Extreme learning machines, artificial neural networks, and multiple linear regression, and multiple nonlinear regression are used to examine the sensor responses to alcohol-associated volatile chemicals. Performance evaluation of the prototype shows that the Artificial N eural N etwork (ANN) model outperforms other models, achieving a Mean Squared Error (MSE) of 1.0000 across multiple test runs, compared to MSE values of 85.627444 for MLR and MNLR models. The Mean Absolute Error (RAE) for the ANN model is as low as 0.0001 in certain runs, demonstrating its precision. These quantitative findings suggest that the ANN model is best suited for accurate alcohol detection in local toddy, offering a significant improvement over traditional methods. The results demonstrate the potential of machine learning methods for detecting alcohol in alcoholic beverages and shed light on the intricate connection between sensor data and alcohol concentrations. These algorithms provide encouraging ways to improve quality control procedures and guarantee constant product quality.