Title: AI Methods for the Prediction of Stress Concentrations in Fiber Reinforced Composites

Project duration: March 2023 – July 2024

Research Area: Engineering and Business

Carbon Fibre Reinforced Polymers are one of the widely used in many fields due to their low weight and high bearing capacity. Though they have high strength to weight ratio, due to the utility requirements such as the electric wiring and fuel lines, notches are introduced into them thereby leading to local high stress-strain values with stiff gradients. Over years, extensive research has been carried out on composites with several notches and interferences and an analytical tool has been developed at ILK.  In this work, an ANN is developed to predict the strain values of a uni-directional Carbon Fibre Reinforced Polymer thereby it acts as a black box which can be passed to a third-party user without the requirement to give away the analytical tool. For the developed multi regression model ANN, the obtained predictions are compared to the analytical values and the model limitations are discussed in this project.

Aims

This research project aims to develop an ANN, which is a multi-regression model, which enables the user to predict the strain values of a multilayered carbon fibre reinforced composite (CFRP) build-up of single unidirectional reinforced laminae under tension with a notch at the centre of the composite. As for training, the necessary data are provided by the use of ILK`s internal analytical software DEKEB for calculation stress-strain-displacement fields of notched multi-layered composites.

Problem

At ILK huge affords have been put into developing the analytical software DEKEB for stress concentration analysis of notched multilayered composites. The project now enables the researchers to implicitly pass the developed calculation methods to third parties without explicitly giving away the developed DEKEB program.

Practical example

• Proof of concept

Technology

• Analytical calculation program DEKEB for providing the training data
• Artificial neuronal network for regression

Outlook

• Adding of different amounts and kinds of loads, i.g. shear, compression or combined loads.
• Include elliptical notches at different angles
• Take into account composites with different angles and of different layups.

Publications

• Pamarthi, R. AI methods for the prediction of stress concentrations in fiber reinforced composites. CMS Research Project, TU Dresden 2023

Team

Lead

• Prof. Dr. Maik Gude

Team Members

• Rajesh Pamarthi
• Dr. Bernd Grüber

Partners

• ILK – Institute of Lightweight Engineering and Polymer Technology, TUD Dresden University of Technology