Aluminium extrusion is a forming process used for manufacturing straight and long aluminium products. Among all aspects of the process, extrusion die design is the key issue for producing high-quality extrudates. The approaches to design extrusion dies can be broadly classified into three groups; trial and error, empirical based approach and numerical simulation based approaches. By using the first two methodologies, the quality of extrusion die designs are intrinsically and strongly linked with designers' experience and skill. As the required forms for extrusions become more complex, these two approaches becomes less useful. Besides, since the design knowledge is held by designers, it is more often a 'black art', and the personnel movement can influence the design work significantly. On the other hand, with the advent of computers and greatly enhanced computing capability, many new approaches have been introduced for designing extrusion dies in last few decades. However, even with the current computing power, the numerical simulation approach has its limitations, particular in time required and even accuracy. Extrusion process involves complex constitutive relationships and large deformation of material. To overcome the limitations posed by current available design approaches, a new geometry based methodology has been proposed in this thesis. The new methodology combines empirical design formulae, geometry reasoning technique and optimization algorithm together. The work originates from the earlier work done by Miles et al. [1, 2, 3, 4], and Armstrong and his colleagues [5, 6, 7, 8, 9, 10, 11, 12], In this research work, a new knowledge representation scheme is developed so that historical data can be easily gathered and reused. By using empirical bearing length design formulae with historical data, a new bearing length estimation approach is introduced so that new profiles can be designed based on past good designs. A novel die layout design approach has also been developed and validated. This new method uses bearing length estimation algorithms with maximum bearing length difference to give radial or fiat layout for single/multi-hole dies. By using medial axis transform, a set of new geometry reasoning algorithms have been studied. These algorithms give a general and robust way to analyze two-dimensional geometry shapes. A brand-new die profile categories have been proposed to avoid the drawbacks held by current classification. A new algorithm and a set of new classifying criteria have been introduced. Based on medial axis transform and geometry reasoning technique, extrusion die profiles can be classified into different category correctly and efficiently. This research work shows that all the proposed approaches give several feedback paths in extrusion die design process. Therefore, not only historical data can be reused for new designs, but it is also possible to acquire and represent design knowledge and to optimize the whole design process.
| Identifer | oai:union.ndltd.org:bl.uk/oai:ethos.bl.uk:637922 |
| Date | January 2005 |
| Creators | Lin, Chao |
| Publisher | Swansea University |
| Source Sets | Ethos UK |
| Detected Language | English |
| Type | Electronic Thesis or Dissertation |
| Source | https://cronfa.swan.ac.uk/Record/cronfa42734 |
Page generated in 0.013 seconds