Faculté de Technologie

Permanent URI for this community

http://dspace.univ-skikda.dz:4000/handle/123456789/22

Browse

Browsing Faculté de Technologie by Author "ADDA Bachir"

Now showing 1 - 1 of 1
  • Thumbnail Image
    Item
    Optimisation des performances mécaniques des biocomposites à matrice époxyde renforcée par des fibres naturelles : Modélisation par RSM et ANN
    (université 20aout1955 skikda, 2024) ADDA Bachir
    In recent years, natural fiber biocomposites have attracted considerable importance as a potential structural material. The attractive characteristics of natural fibers like jute, sisal, coir and banana have been their low cost, light weight, high specific modulus, renewable and biodegradable. Natural fibers are lignocellulosic in nature. These biocomposites are gaining importance due to their non-carcinogenic and biodegradable nature. Natural fiber biocomposites can be very cost-effective materials, especially for the building and construction industry. In many cases the current use of jute is primarily as fuel in sugar mill kilns; it is believed that the value of this agricultural residue can be increased by bonding with resin to produce composites suitable for building materials. With this in mind, current work has been undertaken to develop a polymer matrix composite (epoxy resin) using jute fiber as reinforcement and to study its mechanical properties and environmental performance. The composites are prepared with different rates of jute fibres. This thesis work focuses on the modeling and optimization of machining processes of a laminated structure consisting of an epoxy matrix reinforced with bidirectional jute fibers by response surface methodology (RSM) and artificial neural networks. (ANN). Classical methods are particularly expensive in terms of computation time and appear divergent for problems involving many parameters to manage. Therefore, our contribution is to develop a faster and more efficient modeling and optimization method. During this study, it appeared that the ANN approach has a great potential to model complex relationships to be described with simple mathematical models. Moreover, this approach is particularly fast and can be generated in parallel. Furthermore, the results obtained are of satisfactory quality. This work opens more promising perspectives that can be exploited in various applications.