Intelligent machine control using neural fuzzy algorithms.

Shih, C.-H.V., 1996. Intelligent machine control using neural fuzzy algorithms. PhD, Nottingham Trent University.

10183054.pdf - Published version

Download (94MB) | Preview


Conventional machine control typically requires the measurement of numerous parameters in order to define the system state. These various parameters, from sensors such as shaft encoders, accelerometers, etc., are combined through the control system to provide output actuation. For many problems, the vast number of input transducers required and the complexity of control algorithms mean that such problems are not economically addressable. This is particularly true when non-linearity of the process and time variant process parameters are encountered. In contrast a skilled operator would be able to judge, for example, the quality of cut by visual inspection of the workpiece, and based on experience adjust some of the parameters in order to improve the performance. The work reported in this thesis attempts to employ Artificial Intelligence in combination with remote sensing, in order to reduce the need for feedback sensors and achieve a more effective computer control solution.

The research uses automation of lace trimming as a suitable platform for investigation. The main problems here are cutter path detection in real-time and coping with material flexibility. The system has to work with many different patterns and sizes of lace as well as tolerating misalignment. To achieve a sufficient degree of automation, the trimming path must be located without prior knowledge of the lace pattern. A Fuzzy Reasoning Rule-based technique is applied to overcome the problem of material pattern variation and distortion. Finding the river location across the lace strip must be carried out in real-time. To achieve this, a novel approach, namely the Line Mapping Method, is devised to speed up the search for the path. Experimental results indicate that the path can be successfully detected in different lace patterns in real time, whilst coping with lace distortion.

Work has been reported in using non-tactile means to cut deformable materials. Although the use of non-tactile cutters reduces material deformation, distortion due to mechanical feed misalignment persists. Changes in the lace pattern are also caused by the release of tension in the lace structure as it is cut. To tackle the problem of distortion due to material flexibility in general, a novel approach using inexact algorithms, i.e., fuzzy logic, neural networks and neural fuzzy technique, is developed. A Spring Mounted Pen is used to emulate material distortion caused by tactile cutting and feed misalignment. Using pre- and post-processing vision systems, it is possible to monitor the effects of flexibility and generate on-line information for error compensation. Applying the algorithms developed, the system can produce excellent results, much better than a human operator.

The system developed is a novel approach to flexible sheet material trimming and has further applications where modelling system behaviour characteristics is difficult. Such systems can range from controlling a robot moving on a slippery surface or piloting a boat. Furthermore, by relying on the intelligent software engine, problems such as transmission backlash, joint flexibility and stick-slip can potentially be compensated for. When characteristics of the mechanism, such as component wear and temperature, change over time, the controller can learn the new system behaviour and automatically make appropriate compensation. An industrially sponsored programme of work has just commenced to develop a commercial machine tool controller based on the developed principle.

Item Type: Thesis
Creators: Shih, C.-H.V.
Date: 1996
ISBN: 9781369313468
Divisions: Schools > School of Science and Technology
Record created by: Jeremy Silvester
Date Added: 04 Sep 2020 08:58
Last Modified: 28 Jun 2023 09:02

Actions (login required)

Edit View Edit View


Views per month over past year


Downloads per month over past year