Font Size: 

The usage of various effective algorithm will be helpful in early fire detection and prevention. In this paper, an in-building early fire detection algorithm has been proposed using Multivariate Analysis (MVA) using Principal Component Analysis. The experiments were performed on recorded smell samples from combustion of ten different commonly available household, including candle, joss sticks, air freshener, mosquito coil, newspaper, card board, plastic materials, Styrofoam and wood. All the experiments were done in a test room with humidity and temperature sensors. Portable Electronic Nose (PEN3) from Airsense Analytics is used as the measurement device. The smell source is placed 1.5m from the PEN3 and the time-series signal is measured for two minutes. The odour metrics is fed to the PCA for analysis. It is found that the MVA is able to cluster the odour sample accordingly with its first principal component of 92.74%.