IDENTIFICATION OF FOREST FIRE SMOKE BASED ON ELECTRONIC NOSE USING ARTIFICIAL NEURAL NETWORK

Forest fires are still being carried out manually using monitoring posts, binoculars and human labor. There are still many problems related to early warning. The smoke detectors are limited and cannot distinguish between types of fire smoke that have complex gas compositions. Therefore, an intelligent instrument is required. The electronic nose (e-nose) based on the gas sensor array and pattern recognition has the ability to identify samples based on their characteristics. In this research, we identified
fire smoke types using the e-nose and artificial neural network (ANN). Peat (140 g), wood (12 g) and grass (10 g) each were burned and repeated 10 times through sniffing process. Periodic sensor response wave formed in preprocessing using the difference is used to eliminate unnecessary information to get a sharp and scalable sensor response. Feature extraction using an integral method is applied to finding unique information contained in the sensor response. Data (210 × 12) was subsequently used by ANN (12-20-1) for learning (60%) and testing (40%). The results of ANN can clearly identify each sample. The network is optimized, stable and the pattern of each sample is unique and consistent. So, the e-nose can be used for forest fire smoke detection.