↵

Natural fiber-reinforced plastics (NFRP) offer great potential for the development of sustainable products, but their use is limited due to their mechanical properties and other problems such as moisture. However, an improvement in mechanical properties can be achieved by using a hybrid material approach in which the NFRP is selectively reinforced with unidirectional carbon fiber tapes (CF tapes). In order to make optimum use of the additional stiffening effects resulting from the unidirectional CF tapes, optimization algorithms can be used to calculate the optimum orientation and position. However, these are very computationally intensive, particularly due to the strong non-linearity of the objectives and boundary conditions and the large number of possible solutions. Therefore, reinforcement learning approaches will be used to develop a method that enables the efficient generation of initial design proposals for hybrid NFRP.

Task:

• Systematic literature review on the use of reinforcement learning in the context of design optimization
• Development of a concept for agent-based optimization of hybrid NFVK
• Implementation of the developed approach in Pytorch
• Critical evaluation of the results and derivation of further research needs

Profile: