Title: Artificial Intelligence Based Analysis of Expanded Polypropylene (EPP) Bead Foams using Computed Tomography Scans

Duration: February 2023- July 2023

Research Area: Engineering and Business

EPP bead foams have taken an important place in research and in a variety of industries, from automotive to packaging, serving roles from insulation to impact protection. Understanding their micro properties is essential for performance optimization.

This project focused on three pivotal steps for achieving 3D bead detection and extracting these micro-properties. Initially, Computed Tomography data acquisition, generation, and processing were performed.

The second phase involved 2D bead detection using the Mask R-CNN instance segmentation model, providing an effective framework for isolating individual beads in two-dimensional space. The concluding phase of the study employed 3D bead tracking, enabled by the application of bead shaping using DBSCAN clustering.

These AI algorithms successfully detected beads in both 2D and 3D spaces with high accuracy, enabling the extraction of critical micro-properties such as bead size, shape, and spatial distribution within EPP foam structures.

Aims

AI is often seen as a solution in materials analysis, but it is not yet integrated into EPP bead foam analysis approaches, particularly in the determination of micro-properties, which play a crucial role in ensuring the desired foam quality.

The AI methods developed in this project aims on being an efficient solution for improved use of the CT Imaging Technique, enabling a faster and more accurate analysis of CT scans of samples.

Technology

• volumetric CT scans
• Mask R-CNN
• Detectron2 model
• DBSCAN (Density-Based Spatial Clustering of Applications with Noise) algorithm

Outlook

The results of this project open up numerous possibilities for research and industrial application. The methodology not only serves as a robust method for automating the micro-analysis of EPP bead foams but is also adaptable to the study of other similar materials. Further extensions of this project could focus on generating annotated datasets to improve algorithmic accuracy and introduce more layers of analytical depth. Furthermore, the successful implementation of these techniques in real-time during the production process could offer an unprecedented level of quality control, transforming manufacturing practices in the industry.

Publications

• Chayma Boukthir: Artificial Intelligence Based Analysis of Expanded Polypropylene Bead Foams using Computed Tomography Scans. Graduation Project Report, SUP’com, Tunis

Team

Lead

• Prof. Dr. Maik Gude (TUD-ILK)
• Associate Prof. Rim Barrak (SUP’COM)

Team Members

• Chayma Boukthir
• Dr. Bernd Grüber
• Dr. Ilja Koch
• Dipl.-Wi.-Ing. Paul Schulz

Partners

• ILK – Institute of Lightweight Engineering and Polymer Technology, University of Technology Dresden
• Ecole Supérieure des Communications de Tunis (SUP’COM)