Performance of chemical vapor identification using SAW gas sensor based on ANN
Ciencia y Engenharia/ Science and Engineering Journal, cilt.12, sa.1, ss.53-59, 2003 (Scopus)
- Yayın Türü: Makale / Tam Makale
- Cilt numarası: 12 Sayı: 1
- Basım Tarihi: 2003
- Dergi Adı: Ciencia y Engenharia/ Science and Engineering Journal
- Derginin Tarandığı İndeksler: Scopus
- Sayfa Sayıları: ss.53-59
- Anahtar Kelimeler: Back-propagation neural networks, Chemical vapor detection, SAW gas sensor
- Boğaziçi Üniversitesi Adresli: Hayır
Özet
Surface Acoustic Wave (SAW) gas sensor and Artificial Neural Networks (ANN) may be used to identify a particular chemical vapor from a mixture of chemical vapors. In this study, we explore the benefits and drawbacks of using the back-propagation neural network for the identification of chemical vapors. Some of the widely used heuristics of back-propagation neural network (BP) algorithm are utilized for chemical vapor environments, and their impacts on the performance measures are experimented. The neural network training is done both for noisy measurements and noise free measurements. The noisy vapor patterns are achieved by corrupting the original measurements with Additive White Gaussian Noise, N[0,σ2]. Then, the performance of the chemical vapor identification is simulated for both noise free and noisy environments using both the noise-trained network and noise-free trained network. It is shown that the noise-trained network outperforms the noise-free trained network. It is also shown that both for the noise free and low signal-to-noise ratio (SNR) noise range, the noise-trained network works with the identification rate over 99 percentile. Unfortunately, the performance of chemical vapor identification using SAWNET gas detector may not be trusty for fatal gas applications. Finally, to improve the performance of the above gas detector, we also introduce possible gas detector models using a multi-SAW sensor based on ANN.