APPLICATION OF AN ELECTRONIC NOSE BASED ON GAS SENSOR ARRAY COMBINED WITH ARTIFICIAL NEURAL NETWORK FOR DISCRIMINATION OF SKIN CRACKERS; APLIKASI HIDUNG ELEKTRONIK BERBASIS LARIK SENSOR GAS YANG DIKOMBINASIKAN DENGAN JARINGAN SYARAF TIRUAN UNTUK DISKRIMINASI KERUPUK KULIT

Wicaksono, Alfian Nur, Kuwat Triyana

2015 | Skripsi | FMIPA

An electronic nose (e-nose) system was used to classify and discriminate volatile odors produced by three types of skin crackers. They were cow skin crackers, buffalo skin crackers, and pig skin crackers. A measurement system, equipped with eight Tauguchi Gas Sensor (TGS) metal oxide semiconductor gas sensors, was used to generate a recognition pattern of volatile compounds from skin cracker samples. Maximum response signal considered as the best feature extraction method for extracting characteristic of data from sensor response. Principal component analysis (PCA) and backpropagation artificial neural network (ANN) were used to analyze the experiment data. The PCA method can classify skin crackers related to the type of animal skin being used. PCA analysis also shows that TGS826, TGS822, TGS2600 and TGS2611 contribute the most in discrimination process. The backpropagation neural network shows good correlation with PCA result. Its architecture was designed to be 8-9-3-1 network with logistic sigmoid as activation function. The correct rate in discrimination using backpropagation neural network is 90.72%. Both PCA and backpropagation results indicate that e-nose can also be used to classify and discriminate skin crackers and owns many superiority such as rapid detection, easy operation, and high accuracy comparing to other methods.

Kata Kunci : e-nose; skin cracker; PCA; backpropagation; ANN