This Website is not fully compatible with Internet Explorer.
For a more complete and secure browsing experience please consider using Microsoft Edge, Firefox, or Chrome

Einsatz von KI und maschinellem Lernen in der Simulation von Infiltrationsverfahren


David Droste, M.Sc., Dr.-Ing. Junhong Zhu, and Dr.-Ing. Tim Frerich presented their research on the "Application of AI and Machine Learning in the Simulation of Infiltration Processes". Conducted at the Faserinstitut Bremen and CTC, their work explored integrating artificial intelligence (AI) and machine learning (ML) into manufacturing process simulations, specifically focusing on infiltration methods. The presentation emphasized the challenges in maintaining component quality due to process control limitations, particularly the slow nature of real-time process simulations. To address this, the team developed surrogate models for process simulation, enhancing both offline and online applications. They identified specific applications of AI, such as in loading simulations for autoclaves and infiltration simulations. A significant portion of the presentation was dedicated to a case study on flow front detection using Convolutional Neural Networks (CNN). This study demonstrated the feasibility of monitoring flow fronts in processes based on pressure sensor data, using only simulation data for training and validating the ML model. The researchers described a model involving a flat plate with varying permeability zones, using RTM-Worx software for the simulations. The analysis focused on preparing and analyzing simulation data, particularly on flow front evaluation, which was then used to train the neural network. The design of a fully connected neural network was detailed, including its architecture, parameters, and the use of Python Tensorflow. The team explored the impact of dataset sizes on the model's Mean Relative Error (MRE) and discussed validation techniques. Another case study focused on using CNNs to evaluate permeation experiments, aiming to improve measurement accuracy by allowing longer test durations and reducing material requirements. The team compared real measurements with analytical methods and CNN predictions, showcasing the potential of AI in enhancing the accuracy of permeability determination.

Document Details

AuthorsDroste. D Zhu. J Frerich. T
Date 25th October 2023
OrganisationsFASER Institute - FIBRE CTC


Back to Previous Page