Practical implementation of norm-optimal and predictive iterative learning control on a chain conveyor system
Practical implementation of norm-optimal and predictive iterative learning control on a chain conveyor system
The aim of this thesis is to examine and verify the implementation of two new ILC algorithms which are based on optimality, on a real-time application, a chain conveyor system, and also to obtain a detailed performance comparison of these algorithms. The chain conveyor application is based on a two axis system that consists of a short section of chain conveyor and dispenser. These systems are designed to accomplish two modes of operation: indexing and synchronising. When indexing, the conveyor places items it carries consecutively under the product dispenser in a stop/start function. The synchronising mode however, requires the conveyor to move at a constant speed with the dispenser accelerating to match it. The product is then dispensed when both speeds are in synchrony. An approximate linear model and a theoretical non-linear model of the system have been developed, with the linear model being used for simulation and controller design studies.
The motivation of norm-optimal and predictive algorithms came from the idea of having an ideal form of ILC algorithm that can achieve zero convergence with high performance. The first learning algorithm investigated was the norm optimal ILC approach. The known theoretical background of this algorithm is presented. Simulations with different values of weighing parameters have been carried out, and a successful practical implementation on the plant was achieved. The other learning algorithm is predictive ILC, which is considered as an extension of the previous ILC algorithm. Here the predicted errors on a number of future trials are explicitly included in the cost function for controller design. The properties include an improved rate of convergence using the prediction horizon plus the design weights as mentioned above. This advantage and its merits are clearly presented in the simulation of this algorithm. Also, a successful practical implementation has been achieved.
University of Southampton
Al-Towaim, Tarek
97ae18bf-8b77-4135-9f23-0b3a5c9f110a
2004
Al-Towaim, Tarek
97ae18bf-8b77-4135-9f23-0b3a5c9f110a
Al-Towaim, Tarek
(2004)
Practical implementation of norm-optimal and predictive iterative learning control on a chain conveyor system.
University of Southampton, Doctoral Thesis.
Record type:
Thesis
(Doctoral)
Abstract
The aim of this thesis is to examine and verify the implementation of two new ILC algorithms which are based on optimality, on a real-time application, a chain conveyor system, and also to obtain a detailed performance comparison of these algorithms. The chain conveyor application is based on a two axis system that consists of a short section of chain conveyor and dispenser. These systems are designed to accomplish two modes of operation: indexing and synchronising. When indexing, the conveyor places items it carries consecutively under the product dispenser in a stop/start function. The synchronising mode however, requires the conveyor to move at a constant speed with the dispenser accelerating to match it. The product is then dispensed when both speeds are in synchrony. An approximate linear model and a theoretical non-linear model of the system have been developed, with the linear model being used for simulation and controller design studies.
The motivation of norm-optimal and predictive algorithms came from the idea of having an ideal form of ILC algorithm that can achieve zero convergence with high performance. The first learning algorithm investigated was the norm optimal ILC approach. The known theoretical background of this algorithm is presented. Simulations with different values of weighing parameters have been carried out, and a successful practical implementation on the plant was achieved. The other learning algorithm is predictive ILC, which is considered as an extension of the previous ILC algorithm. Here the predicted errors on a number of future trials are explicitly included in the cost function for controller design. The properties include an improved rate of convergence using the prediction horizon plus the design weights as mentioned above. This advantage and its merits are clearly presented in the simulation of this algorithm. Also, a successful practical implementation has been achieved.
Text
973511.pdf
- Version of Record
More information
Published date: 2004
Identifiers
Local EPrints ID: 465557
URI: http://eprints.soton.ac.uk/id/eprint/465557
PURE UUID: d11d7882-cb6b-45ca-87fa-86b23a9945ee
Catalogue record
Date deposited: 05 Jul 2022 01:45
Last modified: 16 Mar 2024 20:15
Export record
Contributors
Author:
Tarek Al-Towaim
Download statistics
Downloads from ePrints over the past year. Other digital versions may also be available to download e.g. from the publisher's website.
View more statistics