Share Email Print

Proceedings Paper

Ion track based tunable device as humidity sensor: a neural network approach
Author(s): Mamta Sharma; Anuradha Sharma; Vandana Bhattacherjee
Format Member Price Non-Member Price
PDF $14.40 $18.00
cover GOOD NEWS! Your organization subscribes to the SPIE Digital Library. You may be able to download this paper for free. Check Access

Paper Abstract

Artificial Neural Network (ANN) has been applied in statistical model development, adaptive control system, pattern recognition in data mining, and decision making under uncertainty. The nonlinear dependence of any sensor output on the input physical variable has been the motivation for many researchers to attempt unconventional modeling techniques such as neural networks and other machine learning approaches. Artificial neural network (ANN) is a computational tool inspired by the network of neurons in biological nervous system. It is a network consisting of arrays of artificial neurons linked together with different weights of connection. The states of the neurons as well as the weights of connections among them evolve according to certain learning rules.. In the present work we focus on the category of sensors which respond to electrical property changes such as impedance or capacitance. Recently, sensor materials have been embedded in etched tracks due to their nanometric dimensions and high aspect ratio which give high surface area available for exposure to sensing material. Various materials can be used for this purpose to probe physical (light intensity, temperature etc.), chemical (humidity, ammonia gas, alcohol etc.) or biological (germs, hormones etc.) parameters. The present work involves the application of TEMPOS structures as humidity sensors. The sample to be studied was prepared using the polymer electrolyte (PEO/NH4ClO4) with CdS nano-particles dispersed in the polymer electrolyte. In the present research we have attempted to correlate the combined effects of voltage and frequency on impedance of humidity sensors using a neural network model and results have indicated that the mean absolute error of the ANN Model for the training data was 3.95% while for the validation data it was 4.65%. The corresponding values for the LR model were 8.28% and 8.35% respectively. It was also demonstrated the percentage improvement of the ANN Model with respect to the linear regression model. This demonstrates the suitability of neural networks to perform such modeling.

Paper Details

Date Published: 28 January 2013
PDF: 10 pages
Proc. SPIE 8760, International Conference on Communication and Electronics System Design, 87600W (28 January 2013); doi: 10.1117/12.2011313
Show Author Affiliations
Mamta Sharma, PEC Univ. of Technology (India)
Anuradha Sharma, YMCA Univ. of Science and Technology (India)
Vandana Bhattacherjee, Birla Institute of Technology (India)

Published in SPIE Proceedings Vol. 8760:
International Conference on Communication and Electronics System Design
Vijay Janyani; M. Salim; K. K. Sharma, Editor(s)

© SPIE. Terms of Use
Back to Top