Expert System for Composite Winding
Expert System for Composite Winding
The project proposes to use an expert system to model the relation between the settings of the winding bench parameters and the composite material and customer requirements. Currently the knowledge on these relations is only present in the winding expert expertise. Making this expert a very critical link in the winding process and company. An expert system will capture the expert’s knowledge in a mathematical model.
Airborne International is a development company specialised in advanced composite structures and systems. One of the activities is the production of composite structures by means of filament winding.
The filament winding process is a production method for composite tubes. The quality of the product is directly linked to the craftsmanship of the manufacturer. Due to the large amount of inputs needed and the feeling of the craftsman this process is very hard to computerize.
The materials used for the production of tubes differ from thermoplastic to thermoset matrices and from carbon to glass fibres. The different properties and thus the differ inputs for production lead to a high skill of craftsmanship necessary for a good quality.
An ideal situation for composite winding would be the automation of the filament winding process. This automation would lead to less dependency of the manufacturer and thus lower production costs, optimisation of production time, quality and preservation of company expertise.
An expert system provides an effective conceptual framework for dealing with problems of knowledge representation in an environment of uncertainty and vagueness. In this project the expert system will implement the expert operator’s approximate reasoning process in the selection of a control action
- describing and modeling the system as the human expert would describe it in words, linguistic representation of the system
- incorporating uncertainty and/or lack of information about the problem, problems for which the conventional control fails due to non-linearities and enormously complex mathematical model.
Currently efforts to perform a follow up study are ungoing. Introducing learning capabilities into a fully integrated expert system in a fillament winding bench is the main challenge in this project.