Penganggaran Pecahan Minyak Menggunakan Sistem Pintar Berbilang
Estimation of oil fraction is important to know the actual value of oil production. Artificial neural network (ANNs) are able to be used to estimate parameters of flow processes, based on electrical capacitance–sensed tomographic (ECT) data. The estimations of the parameters are done directly,...
| Main Author: | |
|---|---|
| Format: | Monograph |
| Language: | English |
| Published: |
Universiti Sains Malaysia
2006
|
| Subjects: | |
| Online Access: | http://eprints.usm.my/58756/ http://eprints.usm.my/58756/1/Penganggaran%20Pecahan%20Minyak%20Menggunakan%20Sistem%20Pintar%20Berbilang_Tuan%20Sharifah%20%40%20Tuan%20Norhasliza%20Salleh.pdf |
| Summary: | Estimation of oil fraction is important to know the actual value of oil production. Artificial
neural network (ANNs) are able to be used to estimate parameters of flow processes, based
on electrical capacitance–sensed tomographic (ECT) data. The estimations of the
parameters are done directly, without recourse to tomographic images. For this project, the
architecture of ANN that has been used is the Multilayer Perceptron (MLP). The MLP has
been trained with the simulated ECT data. The Matlab version 7 has been used to design
the MLP architecture. The simulated ECT data have been divided into 3 sets for training,
validation and testing process. Stratified and general estimator were trained with this data.
The validation condition has been adopted to stop the training process. After completion of
training process, the best network of each system will be tested with a set of testing data for
its credibility to estimate oil fraction. The performance shows that the error from the
stratified estimator is larger than the general estimator. Meaning that, the estimation made
by general estimator is more accurate. |
|---|