Artificial neural network sediment transport model for Sungai Bernam / Muhammad Syahreen Sa'adon
Artificial neural network (ANN) model is proposed as an alternative to the conventional sediment transport models for Sungai Bernam. This study evaluates the existing sediment transport equations against the local river data. Equations used in evaluations are Ackers & White, Ariffin, Engelund...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English |
| Published: |
2008
|
| Subjects: | |
| Online Access: | https://ir.uitm.edu.my/id/eprint/564/1/TM_MUHAMMAD%20SYAHREEN%20SA%27ADON%20EC%2008_5%201.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Artificial neural network (ANN) model is proposed as an alternative to the conventional
sediment transport models for Sungai Bernam. This study evaluates the existing sediment
transport equations against the local river data. Equations used in evaluations are Ackers
& White, Ariffin, Engelund & Hansen, Einstein-Brown, Graf and Yang equations.
Selections of the equations are based on their performance by previous investigator.
Accuracy of the proposed sediment model was evaluated using the discrepancy ratio.
Discrepancy ratio is the ratio of predicted to measured sediment values. From the
evaluations, Engelund and Hansen equation gave the best prediction when tested against
the local river data. In this study, an improvement on the equation was made and ANN is
used as a tool in analysis. The proposed architecture of the sediment model is a 3 layer
multi perceptron model (2:3:3:3:1) with two neurons in the input layer. The hidden layer
consisting of three slabs and each slab in the hidden layer has 3 neurons. The output
neuron is the total sediment load. The momentum rate parameter and learning rate
parameter are 0.4, and 0.3 respectively. The proposed model was trained using 36 sets of
field data and was further validated using a different set of field data. In both training and
testing phases, the proposed models yield about 90% accuracy. The accuracy of the
model was measured using the discrepancy ratio. Range of discrepancy ratio used for
measurement of accuracy is 0.5 to 2.0. |
|---|