The innovation process modelling for ethanol gas sensing using artificial neural network / Muhammad Afiq Wazini Jemani ... [et al.]
Ethanol gas is one of the most common sources of pollution in the majority of installation and emission units, and it can be toxic if overexposed. A precautionary measure to control this pollutant such as a gas sensor may be designed to limit the amount of ethanol released into the air. The performa...
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Conference or Workshop Item |
| Language: | English |
| Published: |
2021
|
| Subjects: | |
| Online Access: | https://ir.uitm.edu.my/id/eprint/56821/1/56821.pdf https://ir.uitm.edu.my/id/eprint/56821/ https://ispike2021.uitm.edu.my/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Ethanol gas is one of the most common sources of pollution in the majority of installation and emission units, and it can be toxic if overexposed. A precautionary measure to control this pollutant such as a gas sensor may be designed to limit the amount of ethanol released into the air. The performance of a particular gas sensor is mainly dependent on the operating temperature. SnO2 is one of the most used sensor materials in ethanol gas sensors. Although the host material of the sensor is the same, doping with different metals often resulted in different response values at specific operating temperatures. Hence, this research successfully develops a process modelling using Artificial Neural Network (ANN) that can predict the response of different doped SnO2 towards ethanol gas at different temperatures. In this context, ANN is an artificial program that can create a linear and non-linear model without making any assumptions. Three input neurons which are time, temperature and concentration of target gas were applied with one output neuron, which is the response of the sensor. The optimal numbers of hidden layers were achieved by the trial-and-error concept. Four models were developed which involve undoped SnO2, cobalt, nickel and iron-doped SnO2. For the method involved, each of the network structures of the model was built with two hidden layers. Training rule and transfer function of Levenberg-Marquardt (trainlm) and tangent sigmoid (TanSig) were used. The mean square error (MSE) performance plots and coefficient of determination (R2) graphs were observed to evaluate the performance of the ANN model developed, where for all results obtained the value ranged from 0.0-0.1 for MSE and performance plots. The finding shows that the constructed ANN model can produce a decent recognition. This novel process modelling is highly demanding in controlling ethanol gas pollution from industrial activities. |
|---|