- Title
- A novel approach to coke strength prediction using self organizing maps
- Creator
- North, L.; Blackmore, K.; Nesbitt, K.; Hockings, K.; Mahoney, M.
- Relation
- 2017 International Conference on Data Mining (DMIN'17). Proceedings of the 2017 International Conference on Data Mining (Las Vegas, NV 17-20 July, 2017) p. 17-23
- Relation
- https://csce.ucmss.com/cr/books/2017/ConferenceReport?ConferenceKey=DMI
- Publisher
- CSREA Press
- Resource Type
- conference paper
- Date
- 2017
- Description
- Vitrinite reflectance is an important property of coal, and can be used in the prediction of the quality of coke derived from coal blends. However, limited methods that incorporate the associated reflectance distributions within prediction models exist. We present a novel method of classifying vitrinite reflectance distributions using self organizing maps. This approach is shown to provide representative and repeatable distributions. In comparison to approaches that capture only the spread of the distribution, the work presented here is able to capture the often multimodal nature of coal blends. Implementation of the resulting distributions within regression based prediction models showed statistically significant improvement to model prediction, furthering the understanding and control of coal blending. More generally, the work presented here may inform other application domains where the general shape of data distributions needs to be incorporated as a single attribute into simulation models.
- Subject
- self organizing maps; pattern recognition; data distributions; coke prediction
- Identifier
- http://hdl.handle.net/1959.13/1397153
- Identifier
- uon:34200
- Identifier
- ISBN:1601324537
- Language
- eng
- Reviewed
- Hits: 3261
- Visitors: 3226
- Downloads: 0
| Thumbnail | File | Description | Size | Format |
|---|