|
|
說明 | 1 online resource (88 pages) |
文字 | text |
無媒介 | computer |
成冊 | online resource |
附註 | Source: Masters Abstracts International, Volume: 58-01 |
| Adviser: Greg J. Evans |
| Thesis (M.A.S.)--University of Toronto (Canada), 2018 |
| Includes bibliographical references |
| Low-cost gas sensors have been proposed in place of conventional expensive instruments however they have issues due to cross-sensitivity with other pollutants. Several different types of metal oxide and electrochemical sensors and machine learning methods were evaluated. The objectives were to determine which type of sensor, metal oxide or electrochemical, is better at measuring traffic-related air pollution and whether deep neural networks (DNN) and recurrent neural networks (RNN) improve sensor performance. Three devices were deployed across three sites, two in Toronto and one in Beijing to evaluate the performance of calibration. Calibration was performed with two weeks of data from only one site and evaluated with the remaining data. The combination of metal oxide and electrochemical sensors were more accurate when measuring NOx. When targets were normalized, the RNN performed better than DNN and linear calibration, however, not when applied to measuring data well outside the range for calibration |
| Electronic reproduction. Ann Arbor, Mich. : ProQuest, 2019 |
| Mode of access: World Wide Web |
主題 | Chemical engineering |
| Computational chemistry |
| Environmental engineering |
| Electronic books. |
| 0542 |
| 0219 |
| 0775 |
ISBN/ISSN | 9780438183735 |