NEURAL NETWORK BASED PID CONTROLLER FOR AIRCRAFT PITCH ATTITUDE CONTROL SYSTEM
The application of deep learning or Artificial Neural Network (ANN) can provide a way to include empirical data from “best practice” action into the development of a controller. ANN can be employed to construct the dynamics of the controller and form the representation of the plant dynamics. The ANN...
Saved in:
| Main Author: | |
|---|---|
| Format: | Theses |
| Language: | Indonesia |
| Online Access: | https://digilib.itb.ac.id/gdl/view/64226 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Institut Teknologi Bandung |
| Language: | Indonesia |
| Summary: | The application of deep learning or Artificial Neural Network (ANN) can provide a way to include empirical data from “best practice” action into the development of a controller. ANN can be employed to construct the dynamics of the controller and form the representation of the plant dynamics. The ANN can be integrated with other schemes to create a flexible and robust controller configuration. In this thesis, a control scheme based on the ANN model is elaborated to be implemented in an aircraft flight control system. An aircraft can have characteristics during its operation/flight that may require a quite complex model for representing its dynamics. An ANN-based control scheme is then explored for providing a control authority for such aircraft. The investigated scheme exploits ANN for constructing the aircraft dynamics model, the structure of which then is integrated into a component that computes the parameters of a controller that produces proportional-integral-derivative control action. To construct the ANN, sets of training data are required. In this work, the training data is generated through some numerical simulations reproducing the controlling process of an aircraft's linear flight control system. This control scheme enables parameters updates on the controller to be executed when a certain criterion is met. In addition to that, a filter is also implemented that acts as a smoother to compensate for possible abrupt parameter changes. The scheme then is implemented in a nonlinear flight dynamic model of an aircraft and exploited for pitch attitude control case. The ANN-based controller is simulated and analyzed to investigate its performance for some cases representing various parameters control settings and to observe its robustness to plant parameters uncertainties/deviations. The results show that the control scheme can produce a good performance and can overcome the simulated condition where parameter uncertainties/deviations occur.
|
|---|