Smell index for indoor air quality system based on multilayer perceptron (MLP)

Abstract

Indoor air quality index (IAQI) is developed to help users understand the effect of air pollutants to human’s health, with respect to each band. It also gives an overall idea about the condition of air in the locations or rooms being measured. However, most of the IAQ indices presented by previous researchers in their works is calculated based on single pollutatant parameter only like carbon monoxide (CO) and sulfur dioxide (SO2). Since smell is also part of indoor air contaminants, thus it is important to calculated the IAQ indices based on an array of pollutant parameters. This study proposes a smell index (SI) that can inform the user about the perception of smell presents either the smell is “Neutral”, “Pleasant” or “Unpleasant”. In contrast with the IAQI which generates its index based on single pollutant parameter, SI is generated based on an array of pollutant parameters. It generates the smell perceptions based on all pollutants input from six gas sensors, which are parts of an indoor air quality monitoring system (IAQMS). In order to classify the perception smell, multilayer perceptron (MLP) classifier with a back propagation learning algorithm has been used. The results show that the classifier has successfully classify the perception of smell for each pollutant present in indoor environment like ambient air, human activity, presence of chemical products, presence of food and beverage, and presence of fragrance. The model for smell classification which is used to produce smell index (SI) is assigned with the following weightage: “Pleasant” - 1, “Neutral” – 0 and “Unpleasant - 2”. This SI is embedded to the IAQMS system.