System identification of flexible plate structure
This research presented an investigation into the performance of system identification using parametric and nonparametric techniques for the identification of a two-dimensional flexible plate structure. The input and output data of the flexible system were acquired through the experimental work usin...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English |
| Published: |
2010
|
| Subjects: | |
| Online Access: | http://eprints.utm.my/id/eprint/12681/6/AliAbdulHussainMFKM2010.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | This research presented an investigation into the performance of system identification using parametric and nonparametric techniques for the identification of a two-dimensional flexible plate structure. The input and output data of the flexible system were acquired through the experimental work using National Instrumentation data acquisition system and flexible plate test rig. A sinusoidal force was applied to excite the flexible plate and the dynamic response of the system was investigated. The parametric models of the system were developed through Recursive Least Square (RLS) and Genetic Algorithms (GA) methods; whilst the nonparametric models of the system were developed using Multi-layer Perceptron Neural Networks (MLP-NN), Adaptive Elman Neural Networks (ENN) and Adaptive Neuro-Fuzzy Inference System (ANFIS). The validity of the models was investigated based on statistical measures, mean square error and one step ahead output prediction. A comparative performance of all the approaches developed in this research was presented and discussed. It has been demonstrated that the best mean squared error for RLS was 0.0095 and for GA algorithm was 0.000562. This indicates the superiority of GA as compared to RLS for the parametric modelling approaches. For the nonparametric modelling of the system, the best mean squared error for MLP-NN, ENN and ANFIS were 0.000163, 0.001700 and 0.0003978, respectively. The results demonstrated that MLP-NN shows superiority as compared to ENN and ANFIS. The investigation also revealed that, comparing to all modelling techniques, MLP-NN performed the best in terms of convergence time to an optimum solution. |
|---|