Gas Detection and Classification Using Neural Network Based Gas Sensors

Abstract

Alcoholic beverages, apart from being haram, also cause loss of consciousness. The influence of alcohol while driving is very dangerous and can result in an accident. For this reason, it is necessary to detect the alcohol content in beverages so that their halal status is known and to avoid the dangers of consuming alcohol. This research is to detect the aroma of alcohol using the MQ-3 gas sensor, which consists of an aroma sensor in general with an Artificial Neuron Network (ANN), such as the number of neurons, layers, and epoch. Most of the learning schemes require testing to optimize the model structure. For this experiment, ANN is used as a liquid classification in grouping alcoholic and non-alcoholic liquids. The MQ-3 gas sensor successfully reads liquid vapor in alcohol with levels of 30%, 50%, 70%, and other water-based liquids. An artificial neural network with 2 hidden layers, 10 neurons, and 1000 iterations with the sigmoid activation function can approach a regression score of 1.1545 and sq error score of 0.5781.